W3C

HTML 5.3

W3C Working Draft,

This version:
https://www.w3.org/TR/2018/WD-html53-20180703/
Latest published version:
https://www.w3.org/TR/html53/
Editor's Draft:
https://w3c.github.io/html/
Previous Version:
https://www.w3.org/TR/2018/WD-html53-20180426/
Test Suite:
https://w3c-test.org/html/
Editors:
(Invited Expert)
(Invited Expert)
(HM Government)
(Invited Expert)
(Invited Expert)
(W3C)
(The Paciello Group)
Former Editors:
(The Paciello Group)
(Microsoft)
(Microsoft)
(Google)
Participate:
File an issue (open issues)
Others:
Single page version

Abstract

This specification defines the 5th major version, third minor revision of the core language of the World Wide Web: the Hypertext Markup Language (HTML). In this version, new features continue to be introduced to help Web application authors, new elements continue to be introduced based on research into prevailing authoring practices, and special attention continues to be given to defining clear conformance criteria for user agents in an effort to improve interoperability.

Status of this document

This section describes the status of this document at the time of its publication. Other documents may supersede this document. A list of current W3C publications and the latest revision of this technical report can be found in the W3C technical reports index at https://www.w3.org/TR/.

This document was published by the Web Platform Working Group as a Working Draft. This document is intended to become a W3C Recommendation. Feedback and comments on this specification are welcome. Please use Github issues. Historical discussions can be found in the public-html@w3.org archives.

This document is proposed to the Working Group as a Public Working Draft for HTML 5.3, reflecting the "leading edge" of what is interoperably deployed as HTML.

This specification is ready for wide review. Review is particularly requested on significant changes made to the specification, noted in the changelog section.

The following features are considered candidates for being marked "at risk" and unless testing before or during Candidate Recommendation demonstrates interoperable implementation will be not be included in a Proposed Recommendation, per W3C Process - 6.4 Candidate Recommendation:

Publication as a Working Draft does not imply endorsement by the W3C Membership. This is a draft document and may be updated, replaced or obsoleted by other documents at any time. It is inappropriate to cite this document as other than work in progress.

This document was produced by a group operating under the W3C Patent Policy. W3C maintains a public list of any patent disclosures made in connection with the deliverables of the group; that page also includes instructions for disclosing a patent. An individual who has actual knowledge of a patent which the individual believes contains Essential Claim(s) must disclose the information in accordance with section 6 of the W3C Patent Policy.

This document is governed by the 1 February 2018 W3C Process Document.

1. Introduction

1.1. Background

This section is non-normative.

HTML is the World Wide Web’s core markup language. Originally, HTML was primarily designed as a language for semantically describing scientific documents. Its general design, however, has enabled it to be adapted, over the subsequent years, to describe a number of other types of documents and even applications.

1.2. Audience

This section is non-normative.

This specification is intended for authors of documents and scripts that use the features defined in this specification, implementors of tools that operate on pages that use the features defined in this specification, and individuals wishing to establish the correctness of documents or implementations with respect to the requirements of this specification.

This document is probably not suited to readers who do not already have at least a passing familiarity with Web technologies, as in places it sacrifices clarity for precision, and brevity for completeness. More approachable tutorials and authoring guides can provide a gentler introduction to the topic.

In particular, familiarity with the basics of DOM is necessary for a complete understanding of some of the more technical parts of this specification. An understanding of Web IDL, HTTP, XML, Unicode, character encodings, JavaScript, and CSS will also be helpful in places but is not essential.

1.3. Scope

This section is non-normative.

This specification is limited to providing a semantic-level markup language and associated semantic-level scripting APIs for authoring accessible pages on the Web ranging from static documents to dynamic applications.

The scope of this specification does not include providing mechanisms for media-specific customization of presentation (although default rendering rules for Web browsers are included at the end of this specification, and several mechanisms for hooking into CSS are provided as part of the language).

The scope of this specification is not to describe an entire operating system. In particular, hardware configuration software, image manipulation tools, and applications that users would be expected to use with high-end workstations on a daily basis are out of scope. In terms of applications, this specification is targeted specifically at applications that would be expected to be used by users on an occasional basis, or regularly but from disparate locations, with low CPU requirements. Examples of such applications include online purchasing systems, searching systems, games (especially multiplayer online games), public telephone books or address books, communications software (e-mail clients, instant messaging clients, discussion software), document editing software, etc.

1.4. History

This section is non-normative.

For its first five years (1990-1995), HTML went through a number of revisions and experienced a number of extensions, primarily hosted first at CERN, and then at the IETF.

With the creation of the W3C, HTML’s development changed venue again. A first abortive attempt at extending HTML in 1995 known as HTML 3.0 then made way to a more pragmatic approach known as HTML 3.2, which was completed in 1997. HTML 4.01 quickly followed later that same year.

The following year, the W3C membership decided to stop evolving HTML and instead begin work on an XML-based equivalent, called XHTML. This effort started with a reformulation of HTML 4.01 in XML, known as XHTML 1.0, which added no new features except the new serialization, and which was completed in 2000. After XHTML 1.0, the W3C’s focus turned to making it easier for other working groups to extend XHTML, under the banner of XHTML Modularization. In parallel with this, the W3C also worked on a new language that was not compatible with the earlier HTML and XHTML languages, calling it XHTML 2.0.

Around the time that HTML’s evolution was stopped in 1998, parts of the API for HTML developed by browser vendors were specified and published under the name DOM Level 1 (in 1998) and DOM Level 2 Core and DOM Level 2 HTML (starting in 2000 and culminating in 2003). These efforts then petered out, with some DOM Level 3 specifications published in 2004 but the working group being closed before all the Level 3 drafts were completed.

In 2003, the publication of XForms, a technology which was positioned as the next generation of Web forms, sparked a renewed interest in evolving HTML itself, rather than finding replacements for it. This interest was borne from the realization that XML’s deployment as a Web technology was limited to entirely new technologies (like RSS and later Atom), rather than as a replacement for existing deployed technologies (like HTML).

A proof of concept to show that it was possible to extend HTML 4.01’s forms to provide many of the features that XForms 1.0 introduced, without requiring browsers to implement rendering engines that were incompatible with existing HTML Web pages, was the first result of this renewed interest. At this early stage, while the draft was already publicly available, and input was already being solicited from all sources, the specification was only under Opera Software’s copyright.

The idea that HTML’s evolution should be reopened was tested at a W3C workshop in 2004, where some of the principles that underlie the HTML work (described below), as well as the aforementioned early draft proposal covering just forms-related features, were presented to the W3C jointly by Mozilla and Opera. The proposal was rejected on the grounds that the proposal conflicted with the previously chosen direction for the Web’s evolution; the W3C staff and membership voted to continue developing XML-based replacements instead.

Shortly thereafter, Apple, Mozilla, and Opera jointly announced their intent to continue working on the effort under the umbrella of a new venue called the WHATWG. A public mailing list was created, and the draft was moved to the WHATWG site. The copyright was subsequently amended to be jointly owned by all three vendors, and to allow reuse of the specification.

The WHATWG was based on several core principles, in particular that technologies need to be backwards compatible, that specifications and implementations need to match even if this means changing the specification rather than the implementations, and that specifications need to be detailed enough that implementations can achieve complete interoperability without reverse-engineering each other.

The latter requirement in particular required that the scope of the HTML specification include what had previously been specified in three separate documents: HTML 4.01, XHTML 1.1, and DOM Level 2 HTML. It also meant including significantly more detail than had previously been considered the norm.

In 2006, the W3C indicated an interest to participate in the development of HTML 5.0 after all, and in 2007 formed a working group chartered to work with the WHATWG on the development of the HTML specification. Apple, Mozilla, and Opera allowed the W3C to publish the specification under the W3C copyright, while keeping a version with the less restrictive license on the WHATWG site.

For a number of years, both groups then worked together under the same editor: Ian Hickson. In 2011, the groups came to the conclusion that they had different goals: the W3C wanted to draw a line in the sand for features for a HTML 5.0 Recommendation, while the WHATWG wanted to continue working on a Living Standard for HTML, continuously maintaining the specification and adding new features. In mid 2012, a new editing team was introduced at the W3C to take care of creating a HTML 5.0 Recommendation and prepare a Working Draft for the next HTML version.

Since then, the W3C Web Platform WG has been cherry picking patches from the WHATWG that resolved bugs registered on the W3C HTML specification or more accurately represented implemented reality in user agents. At time of publication of this document, patches from the WHATWG HTML specification have been merged until January 12, 2016. The W3C HTML editors have also added patches that resulted from discussions and decisions made by the W3C Web Platform WG as well a bug fixes from bugs not shared by the WHATWG.

A separate document is published to document the differences between the HTML specified in this document and the language described in the HTML 4.01 specification. [HTML5-DIFF]

1.5. Design notes

This section is non-normative.

It must be admitted that many aspects of HTML appear at first glance to be nonsensical and inconsistent.

HTML, its supporting DOM APIs, as well as many of its supporting technologies, have been developed over a period of several decades by a wide array of people with different priorities who, in many cases, did not know of each other’s existence.

Features have thus arisen from many sources, and have not always been designed in especially consistent ways. Furthermore, because of the unique characteristics of the Web, implementation bugs have often become de-facto, and now de-jure, standards, as content is often unintentionally written in ways that rely on them before they can be fixed.

Despite all this, efforts have been made to adhere to certain design goals. These are described in the next few subsections.

1.5.1. Serializability of script execution

This section is non-normative.

To avoid exposing Web authors to the complexities of multithreading, the HTML and DOM APIs are designed such that no script can ever detect the simultaneous execution of other scripts. Even with workers, the intent is that the behavior of implementations can be thought of as completely serializing the execution of all scripts in all browsing contexts.

1.5.2. Compliance with other specifications

This section is non-normative.

This specification interacts with and relies on a wide variety of other specifications. In certain circumstances, unfortunately, conflicting needs have led to this specification violating the requirements of these other specifications. Whenever this has occurred, the transgressions have each been noted as a "willful violation", and the reason for the violation has been noted.

1.5.3. Extensibility

This section is non-normative.

HTML has a wide array of extensibility mechanisms that can be used for adding semantics or behaviours in a way that will not conflict with future development of the Web Platform:

When extending HTML authors should consider whether the new functionality is accessible to users with disabilities, and whether it risks degrading the privacy and security of users. In addition, considering internationalization is important wherever users provide data. These best-practice design considerations are part of the development of the HTML specification.

1.6. HTML vs XML Syntax

This section is non-normative.

This specification defines an abstract language for describing documents and applications, and some APIs for interacting with in-memory representations of resources that use this language.

The in-memory representation is known as "DOM HTML", or "the DOM" for short.

There are various concrete syntaxes that can be used to transmit resources that use this abstract language, two of which are defined in this specification.

The first such concrete syntax is the HTML syntax. This is the format suggested for most authors. It is compatible with most legacy Web browsers. If a document is transmitted with the text/html MIME type, then it will be processed as an HTML document by Web browsers. This specification defines the latest version of the HTML syntax, known simply as "HTML".

The second concrete syntax is the XHTML syntax, which is an application of XML. When a document is transmitted with an XML MIME type, such as application/xhtml+xml, then it is treated as an XML document by Web browsers, to be parsed by an XML processor. Authors are reminded that the processing for XML and HTML differs; in particular, even minor syntax errors will prevent a document labeled as XML from being rendered fully, whereas they would be ignored in the HTML syntax. This specification defines the latest version of the XHTML syntax, known simply as "XHTML".

The DOM, the HTML syntax, and the XHTML syntax cannot all represent the same content. For example, namespaces cannot be represented using the HTML syntax, but they are supported in the DOM and in the XHTML syntax. Similarly, documents that use the noscript feature can be represented using the HTML syntax, but cannot be represented with the DOM or in the XHTML syntax. Comments that contain the string "-->" can only be represented in the DOM, not in the HTML and XHTML syntaxes.

1.7. Structure of this specification

This section is non-normative.

This specification is divided into the following major sections:

§1 Introduction

Non-normative materials providing a context for the HTML specification.

§2 Common infrastructure

The conformance classes, algorithms, definitions, and the common underpinnings of the rest of the specification.

§3 Semantics, structure, and APIs of HTML documents

Documents are built from elements. These elements form a tree using the DOM. This section defines the features of this DOM, as well as introducing the features common to all elements, and the concepts used in defining elements.

§4 The elements of HTML

Each element has a predefined meaning, which is explained in this section. Rules for authors on how to use the element, along with user agent requirements for how to handle each element, are also given. This includes large signature features of HTML such as video playback and subtitles, form controls and form submission, and a 2D graphics API known as the HTML canvas.

§5 User interaction

HTML documents can provide a number of mechanisms for users to interact with and modify content, which are described in this section, such as how focus works, and drag-and-drop.

§6 Loading Web pages

HTML documents do not exist in a vacuum — this section defines many of the features that affect environments that deal with multiple pages, such as Web browsers and offline caching of Web applications.

§7 Web application APIs

This section introduces basic features for scripting of applications in HTML.

§8 The HTML syntax
§9 The XML syntax

All of these features would be for naught if they couldn’t be represented in a serialized form and sent to other people, and so these sections define the syntaxes of HTML and XHTML, along with rules for how to parse content using those syntaxes.

§10 Rendering

This section defines the default rendering rules for Web browsers.

There are also some appendices, listing §11 Obsolete features and §12 IANA considerations, and several indices.

1.7.1. How to read this specification

This specification should be read like all other specifications. First, it should be read cover-to-cover, multiple times. Then, it should be read backwards at least once. Then it should be read by picking random sections from the contents list and following all the cross-references.

As described in the conformance requirements section below, this specification describes conformance criteria for a variety of conformance classes. In particular, there are conformance requirements that apply to producers, for example authors and the documents they create, and there are conformance requirements that apply to consumers, for example Web browsers. They can be distinguished by what they are requiring: a requirement on a producer states what is allowed, while a requirement on a consumer states how software is to act.

For example, "the foo attribute’s value must be a valid integer" is a requirement on producers, as it lays out the allowed values; in contrast, the requirement "the foo attribute’s value must be parsed using the rules for parsing integers" is a requirement on consumers, as it describes how to process the content.

Requirements on producers have no bearing whatsoever on consumers.

Continuing the above example, a requirement stating that a particular attribute’s value is constrained to being a valid integer emphatically does not imply anything about the requirements on consumers. It might be that the consumers are in fact required to treat the attribute as an opaque string, completely unaffected by whether the value conforms to the requirements or not. It might be (as in the previous example) that the consumers are required to parse the value using specific rules that define how invalid (non-numeric in this case) values are to be processed.

1.7.2. Typographic conventions

This is a definition, requirement, or explanation.

This is a note.

This is an example.

This is an open issue.

This is a warning.

interface Example {
  // this is an IDL definition
};
variable = object . method( [ optionalArgument ] )
This is a note to authors describing the usage of an interface.
/* this is a CSS fragment */

The defining instance of a term is marked up like this. Uses of that term are marked up like this or like this.

The defining instance of an element, attribute, or API is marked up like this. References to that element, attribute, or API are marked up like this.

Other code fragments are marked up like this.

Byte sequences with bytes in the range 0x00 to 0x7F, inclusive, are marked up like this.

Variables are marked up like this.

In an algorithm, steps in synchronous sections are marked with ⌛.

In some cases, requirements are given in the form of lists with conditions and corresponding requirements. In such cases, the requirements that apply to a condition are always the first set of requirements that follow the condition, even in the case of there being multiple sets of conditions for those requirements. Such cases are presented as follows:

This is a condition
This is another condition
This is the requirement that applies to the conditions above.
This is a third condition
This is the requirement that applies to the third condition.

1.8. Privacy concerns

This section is non-normative.

Some features of HTML trade user convenience for a measure of user privacy.

In general, due to the Internet’s architecture, a user can be distinguished from another by the user’s IP address. IP addresses do not perfectly match to a user; as a user moves from device to device, or from network to network, their IP address will change; similarly, NAT routing, proxy servers, and shared computers enable packets that appear to all come from a single IP address to actually map to multiple users. Technologies such as onion routing can be used to further anonymize requests so that requests from a single user at one node on the Internet appear to come from many disparate parts of the network.

However, the IP address used for a user’s requests is not the only mechanism by which a user’s requests could be related to each other. Cookies, for example, are designed specifically to enable this, and are the basis of most of the Web’s session features that enable you to log into a site with which you have an account.

There are other mechanisms that are more subtle. Certain characteristics of a user’s system can be used to distinguish groups of users from each other; by collecting enough such information, an individual user’s browser’s "digital fingerprint" can be computed, which can be as good, if not better, than an IP address in ascertaining which requests are from the same user.

Grouping requests in this manner, especially across multiple sites, can be used for both benign (and even arguably positive) purposes, as well as for malevolent purposes. An example of a reasonably benign purpose would be determining whether a particular person seems to prefer sites with dog illustrations as opposed to sites with cat illustrations (based on how often they visit the sites in question) and then automatically using the preferred illustrations on subsequent visits to participating sites. Malevolent purposes, however, could include governments combining information such as the person’s home address (determined from the addresses they use when getting driving directions on one site) with their apparent political affiliations (determined by examining the forum sites that they participate in) to determine whether the person should be prevented from voting in an election.

Since the malevolent purposes can be remarkably evil, user agent implementors are encouraged to consider how to provide their users with tools to minimize leaking information that could be used to fingerprint a user.

Unfortunately, as the first paragraph in this section implies, sometimes there is great benefit to be derived from exposing the very information that can also be used for fingerprinting purposes, so it’s not as easy as simply blocking all possible leaks. For instance, the ability to log into a site to post under a specific identity requires that the user’s requests be identifiable as all being from the same user. More subtly, though, information such as how wide text is, which is necessary for many effects that involve drawing text onto a canvas (e.g., any effect that involves drawing a border around the text) also leaks information that can be used to group a user’s requests. (In this case, by potentially exposing, via a brute force search, which fonts a user has installed, information which can vary considerably from user to user.)

Features in this specification which can be used to fingerprint the user are marked as this paragraph is. (This is a fingerprinting vector.)

Other features in the platform can be used for the same purpose, though, including, though not limited to:

1.9. A quick introduction to HTML

This section is non-normative.

A basic HTML document looks like this:

<!DOCTYPE html>
<html>
  <head>
    <title>Document title</title>
  </head>
  <body>
    <h1>Document heading</h1>
    <p>This is a paragraph of text.</p>
    <p>
      <a href="another-html-document.html">
        Link text for a link to another-html-document.html
      </a>
    </p>
    <!-- this is a comment -->
  </body>
</html>

HTML documents consist of a tree of elements and text. Each element is denoted in the source by a start tag, such as "body", and an end tag, such as "/body". (Certain start tags and end tags can in certain cases be omitted and are implied by other tags.)

Tags have to be nested such that elements are all completely within each other, without overlapping:

<!-- incorrect tag nesting -->
<h1>
  <a href="this-html-document.html">Document heading that is a link to this document
</h1>
  </a>

<!-- appropriate tag nesting -->
<h1>
  <a href="this-html-document.html">Document heading that is a link to this document</a>
</h1>

This specification defines a set of elements that can be used in HTML, along with rules about the ways in which the elements can be nested.

Elements can have attributes, which control how the elements work. In the example below, there is a hyperlink, formed using the a element and its href attribute:

<a href="demo.html">demo link</a>

Attributes are placed inside the start tag, and consist of a name and a value, separated by an "=" character. The attribute value can remain unquoted if it doesn’t contain space characters or any of " ' ` = < or >. Otherwise, it has to be quoted using either single or double quotes. The value, along with the "=" character, can be omitted altogether if the value is the empty string.

<!-- empty attributes -->
<input name=address disabled>
<input name="address" disabled="">

<!-- attributes with a value -->
<input name=address maxlength=200>
<input name='address' maxlength='200'>
<input name="address" maxlength="200">

HTML user agents (e.g., Web browsers) then parse this markup, turning it into a DOM (Document Object Model) tree. A DOM tree is an in-memory representation of a document.

DOM trees contain several kinds of nodes, in particular a DocumentType node, Element nodes, Text nodes, Comment nodes, and in some cases ProcessingInstruction nodes.

The basic HTML document example, at the top of this section would be turned into the following DOM tree:

The document element of this tree is the html element, which is the element always found in that position in HTML documents. It contains two elements, head and body, as well as a Text node between them.

There are many more Text nodes in the DOM tree than one would initially expect, because the source contains a number of spaces (represented here by "␣") and line breaks ("⏎") that all end up as Text nodes in the DOM. However, for historical reasons not all of the spaces and line breaks in the original markup appear in the DOM. In particular, all the white space before head start tag ends up being dropped silently, and all the white space after the body end tag ends up placed at the end of the body.

The head element contains a title element, which itself contains a Text node with the text "Document title". Similarly, the body element contains an h1 element, a p element, and a comment.


This DOM tree can be manipulated from scripts in the page. Scripts (typically in JavaScript) are small programs that can be embedded using the script element or using event handler content attributes. For example, here is a form with a script that sets the value of the form’s output element to say "Hello World".

<form name="main">
  Result: <output name="result"></output>
  <script>
    document.forms.main.elements.result.value = 'Hello World';
  </script>
</form>

Each element in the DOM tree is represented by an object, and these objects have APIs so that they can be manipulated. For instance, a link (e.g., the a element in the tree above) can have its "href" attribute changed in several ways:

var a = document.links[0]; // obtain the first link in the document
a.href = 'sample.html'; // change the destination URL of the link
a.protocol = 'https'; // change just the scheme part of the URL
a.setAttribute('href', 'http://example.com/'); // change the content attribute directly

Since DOM trees are used as the way to represent HTML documents when they are processed and presented by implementations (especially interactive implementations like Web browsers), this specification is mostly phrased in terms of DOM trees, instead of the markup described above.


HTML documents represent a media-independent description of interactive content. HTML documents might be rendered to a screen, or through a speech synthesizer, or on a braille display. To influence exactly how such rendering takes place, authors can use a styling language such as CSS.

In the following example, the page has been made yellow-on-blue using CSS.

<!DOCTYPE html>
<html>
  <head>
    <title>Document title</title>
    <style>
      body {
        background: blue;
        color: yellow;
      }
    </style>
  </head>
  <body>
    <h1>Styled page</h1>
    <p>The document has yellow text and a blue background.</p>
  </body>
</html>

For more details on how to use HTML, authors are encouraged to consult tutorials and guides. Some of the examples included in this specification might also be of use, but the novice author is cautioned that this specification, by necessity, defines the language with a level of detail that might be difficult to understand at first.

1.9.1. Writing secure applications with HTML

This section is non-normative.

When HTML is used to create interactive sites, care needs to be taken to avoid introducing vulnerabilities through which attackers can compromise the integrity of the site itself or of the site’s users.

A comprehensive study of this matter is beyond the scope of this document, and authors are strongly encouraged to study the matter in more detail. However, this section attempts to provide a quick introduction to some common pitfalls in HTML application development.

The security model of the Web is based on the concept of "origins", and correspondingly many of the potential attacks on the Web involve cross-origin actions. [ORIGIN]

Not validating user input
Cross-site scripting (XSS)
SQL injection

When accepting untrusted input, e.g., user-generated content such as text comments, values in URL parameters, messages from third-party sites, etc, it is imperative that the data be validated before use, and properly escaped when displayed. Failing to do this can allow a hostile user to perform a variety of attacks. These attacks may range from the potentially benign, such as providing bogus user information like a negative age, to the serious, such as running scripts every time a user looks at a page that includes the information, potentially propagating the attack in the process, to the catastrophic, such as deleting all data in the server.

When writing filters to validate user input, it is imperative that filters always be safelist-based, allowing known-safe constructs and disallowing all other input. Blocklist-based filters that disallow known-bad inputs and allow everything else are not secure, as not everything that is bad is yet known (for example, because it might be invented in the future).

For example, suppose a page looked at its URL’s query string to determine what to display, and the site then redirected the user to that page to display a message, such as:
<ul>
  <li><a href="message.cgi?say=Hello">Say Hello</a></li>
  <li><a href="message.cgi?say=Welcome">Say Welcome</a></li>
  <li><a href="message.cgi?say=Kittens">Say Kittens</a></li>
</ul>

If the message was just displayed to the user without escaping, a hostile attacker could then craft a URL that contained a script element:

http://example.com/message.cgi?say=%3Cscript%3Ealert%28%27Oh%20no%21%27%29%3C/script%3E

If the attacker then convinced a victim user to visit this page, a script of the attacker’s choosing would run on the page. Such a script could do any number of hostile actions, limited only by what the site offers: if the site is an e-commerce shop, for instance, such a script could cause the user to unknowingly make many unwanted purchases.

This is called a cross-site scripting attack.

There are many constructs that can be used to try to trick a site into executing code. Here are some that authors are encouraged to consider when writing safelist filters:

  • When allowing harmless-seeming elements like img, it is important to safelist any provided attributes as well. If one allowed all attributes then an attacker could, for instance, use the onload attribute to run arbitrary script.

  • When allowing URLs to be provided (e.g., for links), the scheme of each URL also needs to be explicitly safelisted, as there are many schemes that can be abused. The most prominent example is "javascript:", but user agents can implement (and indeed, have historically implemented) others.

  • Allowing a base element to be inserted means any script elements in the page with relative links can be hijacked, and similarly that any form submissions can get redirected to a hostile site.

Cross-site request forgery (CSRF)

If a site allows a user to make form submissions with user-specific side-effects, for example posting messages on a forum under the user’s name, making purchases, or applying for a passport, it is important to verify that the request was made by the user intentionally, rather than by another site tricking the user into making the request unknowingly.

This problem exists because HTML forms can be submitted to other origins.

Sites can prevent such attacks by populating forms with user-specific hidden tokens, or by checking Origin headers on all requests.

Clickjacking

A page that provides users with an interface to perform actions that the user might not wish to perform needs to be designed so as to avoid the possibility that users can be tricked into activating the interface.

One way that a user could be so tricked is if a hostile site places the victim site in a small iframe and then convinces the user to click, for instance by having the user play a reaction game. Once the user is playing the game, the hostile site can quickly position the iframe under the mouse cursor just as the user is about to click, thus tricking the user into clicking the victim site’s interface.

To avoid this, sites that do not expect to be used in frames are encouraged to only enable their interface if they detect that they are not in a frame (e.g., by comparing the window object to the value of the top attribute).

1.9.2. Common pitfalls to avoid when using the scripting APIs

This section is non-normative.

Scripts in HTML have "run-to-completion" semantics, meaning that the browser will generally run the script uninterrupted before doing anything else, such as firing further events or continuing to parse the document.

On the other hand, parsing of HTML files happens incrementally, meaning that the parser can pause at any point to let scripts run. This is generally a good thing, but it does mean that authors need to be careful to avoid hooking event handlers after the events could have possibly fired.

There are two techniques for doing this reliably: use event handler content attributes, or create the element and add the event handlers in the same script. The latter is safe because, as mentioned earlier, scripts are run to completion before further events can fire.

One way this could manifest itself is with img elements and the load event. The event could fire as soon as the element has been parsed, especially if the image has already been cached (which is common).

Here, the author uses the onload handler on an img element to catch the load event:

<img src="games.png" alt="Games" onload="gamesLogoHasLoaded(event)">

If the element is being added by script, then so long as the event handlers are added in the same script, the event will still not be missed:

<script>
  var img = new Image();
  img.src = 'games.png';
  img.alt = 'Games';
  img.onload = gamesLogoHasLoaded;

  // The following would also work
  // img.addEventListener('load', gamesLogoHasLoaded, false);
</script>

However, if the author first created the img element and then in a separate script added the event listeners, there’s a chance that the load event would be fired in between, leading it to be missed:

<!-- Do not use this style, it has a race condition! -->
<img id="games" src="games.png" alt="Games">
<!--
  The 'load' event might fire here while the parser is taking
  a break, in which case you will not see it!
-->
<script>
  var img = document.getElementById('games');
  img.onload = gamesLogoHasLoaded; // might never fire!
</script>

1.9.3. How to catch mistakes when writing HTML: validators and conformance checkers

This section is non-normative.

Authors are encouraged to make use of conformance checkers (also known as validators) to catch common mistakes. The W3C provides a number of online validation services, including the Nu Markup Validation Service.

1.10. Conformance requirements for authors

This section is non-normative.

Unlike previous versions of the HTML specification, this specification defines in some detail the required processing for invalid documents as well as valid documents.

However, even though the processing of invalid content is in most cases well-defined, conformance requirements for documents are still important: in practice, interoperability (the situation in which all implementations process particular content in a reliable and identical or equivalent way) is not the only goal of document conformance requirements. This section details some of the more common reasons for still distinguishing between a conforming document and one with errors.

1.10.1. Presentational markup

This section is non-normative.

The majority of presentational features from previous versions of HTML are no longer allowed. Presentational markup in general has been found to have a number of problems:

The use of presentational elements leads to poorer accessibility

While it is possible to use presentational markup in a way that provides users of assistive technologies (ATs) with an acceptable experience (e.g., using ARIA), doing so is significantly more difficult than doing so when using semantically-appropriate markup. Furthermore, even using such techniques doesn’t help make pages accessible for non-AT, non-graphical users, such as users of text-mode browsers.

Using media-independent markup, on the other hand, provides an easy way for documents to be authored in such a way that they are "accessible" for more users (e.g., users of text browsers).

Higher cost of maintenance

It is significantly easier to maintain a site written in such a way that the markup is style-independent. For example, changing the color of a site that uses <font color=""> throughout requires changes across the entire site, whereas a similar change to a site based on CSS can be done by changing a single file.

Larger document sizes

Presentational markup tends to be much more redundant, and thus results in larger document sizes.

For those reasons, presentational markup has been removed from HTML in this version. This change should not come as a surprise; HTML 4.0 deprecated presentational markup many years ago and provided a mode (HTML Transitional) to help authors move away from presentational markup; later, XHTML 1.1 went further and obsoleted those features altogether.

The only remaining presentational markup features in HTML are the style attribute and the style element. Use of the style attribute is somewhat discouraged in production environments, but it can be useful for rapid prototyping (where its rules can be directly moved into a separate style sheet later) and for providing specific styles in unusual cases where a separate style sheet would be inconvenient. Similarly, the style element can be useful in syndication or for page-specific styles, but in general an external style sheet is likely to be more convenient when the styles apply to multiple pages.

It is also worth noting that some elements that were previously presentational have been redefined in this specification to be media-independent: b, i, hr, s, small, and u.

1.10.2. Syntax errors

This section is non-normative.

The syntax of HTML is constrained to avoid a wide variety of problems.

Unintuitive error-handling behavior

Certain invalid syntax constructs, when parsed, result in DOM trees that are highly unintuitive.

For example, the following markup fragment results in a DOM with an hr element that is an earlier sibling of the corresponding table element:
<table><hr>...
Errors with optional error recovery

To allow user agents to be used in controlled environments without having to implement the more bizarre and convoluted error handling rules, user agents are permitted to fail whenever encountering a parse error.

Errors where the error-handling behavior is not compatible with streaming user agents

Some error-handling behavior, such as the behavior for the tablehr... example mentioned above, are incompatible with streaming user agents (user agents that process HTML files in one pass, without storing state). To avoid interoperability problems with such user agents, any syntax resulting in such behavior is considered invalid.

Errors that can result in infoset coercion

When a user agent based on XML is connected to an HTML parser, it is possible that certain invariants that XML enforces, such as element or attribute names never contain multiple colons, will be violated by an HTML file. Handling this can require that the parser coerce the HTML DOM into an XML-compatible infoset. Most syntax constructs that require such handling are considered invalid. (Comments containing two consecutive hyphens, or ending with a hyphen, are exceptions that are allowed in the HTML syntax.)

Errors that result in disproportionately poor performance

Certain syntax constructs can result in disproportionately poor performance. To discourage the use of such constructs, they are typically made non-conforming.

For example, the following markup results in poor performance, since all the unclosed i elements have to be reconstructed in each paragraph, resulting in progressively more elements in each paragraph:
<p><i>He dreamt.
<p><i>He dreamt that he ate breakfast.
<p><i>Then lunch.
<p><i>And finally dinner.

The resulting DOM for this fragment would be:

  • p
    • i
      • #text: He dreamt.
  • p
    • i
      • i
        • #text: He dreamt that he ate breakfast.
  • p
    • i
      • i
        • i
          • #text: Then lunch.
  • p
    • i
      • i
        • i
          • i
            • #text: And finally dinner.
Errors involving fragile syntax constructs

There are syntax constructs that, for historical reasons, are relatively fragile. To help reduce the number of users who accidentally run into such problems, they are made non-conforming.

For example, the parsing of certain named character references in attributes happens even with the closing semicolon being omitted. It is safe to include an ampersand followed by letters that do not form a named character reference, but if the letters are changed to a string that does form a named character reference, they will be interpreted as that character instead.

In this fragment, the attribute’s value is "?bill&ted":

<a href="?bill&ted">Bill and Ted</a>

However, in the following fragment the attribute’s value is actually "?art©", not the intended "?art&copy". This is because even without the final semicolon, "&copy" is handled the same as "&copy;" and is interpreted as "©":

<a href="?art&copy">Art and Copy</a>

To avoid this problem, all named character references are required to end with a semicolon. Uses of named character references without a semicolon are flagged as errors.

The correct way to express the above cases are as follows:

<a href="?bill&amp;ted">Bill and Ted</a>
<!--
  While &ted is not a named character, providing consistency in escaping ampersands will remove ambiguity over best practice, and will ensure that if &ted ever becomes a named character, it will not break such fragments.
-->
<a href="?art&amp;copy">Art and Copy</a>
<!-- The & has to be escaped, since &copy is a named character reference -->
Errors involving known interoperability problems in legacy user agents

Certain syntax constructs are known to cause especially subtle or serious problems in legacy user agents, and are therefore marked as non-conforming to help authors avoid them.

For example, this is why the U+0060 GRAVE ACCENT character (`) is not allowed in unquoted attributes. In certain legacy user agents, it is sometimes treated as a quote character.

Another example of this is the DOCTYPE, which is required to trigger no-quirks mode, because the behavior of legacy user agents in quirks mode is often largely undocumented.

Errors that risk exposing authors to security attacks

Certain restrictions exist purely to avoid known security problems.

For example, the restriction on using UTF-7 exists purely to avoid authors falling prey to a known cross-site-scripting attack using UTF-7. [RFC2152]

Cases where the author’s intent is unclear

Markup where the author’s intent is very unclear is often made non-conforming. Correcting these errors early makes later maintenance easier.

For example, it is unclear whether the author intended the following to be an h1 heading or an h2 heading:
<h2>Contact details</h1>
Cases that are likely to be typos

When a user makes a simple typo, it is helpful if the error can be caught early, as this can save the author a lot of debugging time. This specification therefore usually considers it an error to use element names, attribute names, and so forth, that do not match the names defined in this specification.

For example, if the author typed <capton> instead of caption, this would be flagged as an error and the author could correct the typo immediately.

Errors that could interfere with new syntax in the future

In order to allow the language syntax to be extended in the future, certain otherwise harmless features are disallowed.

For example, attributes in end tags are ignored currently, but they are invalid, in case a future change to the language makes use of that syntax feature without conflicting with already-deployed (and valid!) content.

Some authors find it helpful to be in the practice of always quoting all attributes and always including all optional tags, preferring the consistency derived from such custom over the minor benefits of terseness afforded by making use of the flexibility of the HTML syntax. To aid such authors, conformance checkers can provide modes of operation wherein such conventions are enforced.

1.10.3. Restrictions on content models and on attribute values

This section is non-normative.

Beyond the syntax of the language, this specification also places restrictions on how elements and attributes can be specified. These restrictions are present for similar reasons:

Errors involving content with dubious semantics

To avoid misuse of elements with defined meanings, content models are defined that restrict how elements can be nested when such nestings would be of dubious value.

For example, this specification disallows nesting a section element inside a kbd element, since it is highly unlikely for an author to indicate that an entire section should be keyed in.

Errors that involve a conflict in expressed semantics

Similarly, to draw the author’s attention to mistakes in the use of elements, clear contradictions in the semantics expressed are also considered conformance errors.

In the fragments below, for example, the semantics are nonsensical: a separator cannot simultaneously be a cell, nor can a radio button be a progress bar.
<hr role="cell">
<input type="radio" role="progressbar">

Another example is the restrictions on the content models of the ul element, which only allows li element children. Lists, by definition, consist of zero or more list items, so if a ul element contains something other than an li element, it’s not clear what was meant.

Cases where the default styles are likely to lead to confusion

Certain elements have default styles or behaviors that make certain combinations likely to lead to confusion. Where these have equivalent alternatives without this problem, the confusing combinations are disallowed.

For example, div elements are rendered as block boxes, and span elements as inline boxes. Putting a block box in an inline box is unnecessarily confusing; since either nesting just div elements, or nesting just span elements, or nesting span elements inside div elements all serve the same purpose as nesting a div element in a span element, but only the latter involves a block box in an inline box, the latter combination is disallowed.

Some transparent elements, such as a, del, and ins, are inline box elements, but allow for nesting of block box elements.

<a href="https://example.com">
  <h2>An interesting news story</h2>
  <p>You’re going to want to read this...</p>
</a>

<ins>
  <p>My new paragraph.</p>
</ins>

<del>
  <ul>
    <li>My old list.</li>
  </ul>
</del>

The above shows how a, ins, and del elements may contain block box elements. For additional examples and information, review these elements and other transparent elements.

Another example would be the way interactive content cannot be nested. For example, a button element cannot contain a descendant textarea element. This is because the default behavior of nesting interactive elements would be highly confusing to users. Rather than nesting these elements, they could instead be styled with CSS to be visually placed side by side.
<!-- Invalid -->
<button type="submit">
  <textarea>Confused?</textarea>
</button>

<!-- Valid -->
<textarea>Type Here</textarea>
<button type="submit">
  Submit Here
</button>
Errors that indicate a likely misunderstanding of the specification

Sometimes, something is disallowed because allowing it would likely cause author confusion.

For example, setting the disabled attribute to the value "false" is disallowed, because despite the appearance of meaning that the element is enabled, it in fact means that the element is disabled (what matters for implementations is the presence of the attribute, not its value).

Errors involving limits that have been imposed merely to simplify the language

Some conformance errors simplify the language that authors need to learn.

For example, the area element’s shape attribute, despite accepting both "circ" and "circle" values in practice as synonyms, disallows the use of the "circ" value, so as to simplify tutorials and other learning aids. There would be no benefit to allowing both, and it would cause extra confusion when teaching the language.

Errors that involve peculiarities of the parser

Certain elements are parsed in somewhat eccentric ways (typically for historical reasons), and their content model restrictions are intended to avoid exposing the author to these issues.

For example, a form element isn’t allowed inside phrasing content, because when parsed as HTML, a form element’s start tag will imply a p element’s end tag. Thus, the following markup results in two paragraphs, not one:
<p>Welcome. <form><label>Name:</label> <input></form>

It is parsed exactly like the following:

<p>Welcome. </p>
<form><label>Name:</label> <input></form>
Errors that would likely result in scripts failing in hard-to-debug ways

Some errors are intended to help prevent script problems that would be hard to debug.

This is why, for instance, it is non-conforming to have two id attributes with the same value. Duplicate IDs lead to the wrong element being selected, with sometimes disastrous effects whose cause is hard to determine.

Errors that waste authoring time

Some constructs are disallowed because historically they have been the cause of a lot of wasted authoring time, and by encouraging authors to avoid making them, authors can save time in future efforts.

For example, a script element’s src attribute causes the element’s contents to be ignored. However, this isn’t obvious, especially if the element’s contents appear to be executable script — which can lead to authors spending a lot of time trying to debug the inline script without realizing that it is not executing. To reduce this problem, this specification makes it non-conforming to have executable script in a script element when the src attribute is present. This means that authors who are validating their documents are less likely to waste time with this kind of mistake.

Errors that involve areas that affect authors migrating to and from XHTML

Some authors like to write files that can be interpreted as both XML and HTML with similar results. Though this practice is discouraged in general due to the myriad of subtle complications involved (especially when involving scripting, styling, or any kind of automated serialization), this specification has a few restrictions intended to at least somewhat mitigate the difficulties. This makes it easier for authors to use this as a transitionary step when migrating between HTML and XHTML.

For example, there are somewhat complicated rules surrounding the lang and xml:lang attributes intended to keep the two synchronized.

Another example would be the restrictions on the values of xmlns attributes in the HTML serialization, which are intended to ensure that elements in conforming documents end up in the same namespaces whether processed as HTML or XML.

Errors that involve areas reserved for future expansion

As with the restrictions on the syntax intended to allow for new syntax in future revisions of the language, some restrictions on the content models of elements and values of attributes are intended to allow for future expansion of the HTML vocabulary.

For example, limiting the values of the target attribute that start with an U+005F LOW LINE character (_) to only specific predefined values allows new predefined values to be introduced at a future time without conflicting with author-defined values.

Errors that indicate a mis-use of other specifications

Certain restrictions are intended to support the restrictions made by other specifications.

For example, requiring that attributes that take media query lists use only valid media query lists reinforces the importance of following the conformance rules of that specification.

1.11. Suggested reading

This section is non-normative.

The following documents might be of interest to readers of this specification.

Character Model for the World Wide Web 1.0: Fundamentals [CHARMOD]
This Architectural Specification provides authors of specifications, software developers, and content developers with a common reference for interoperable text manipulation on the World Wide Web, building on the Universal Character Set, defined jointly by the Unicode specification and ISO/IEC 10646. Topics addressed include use of the terms "character", "encoding" and "string", a reference processing model, choice and identification of character encodings, character escaping, and string indexing.
Unicode Security Considerations [UNICODE-SECURITY]
Because Unicode contains such a large number of characters and incorporates the varied writing systems of the world, incorrect usage can expose programs or systems to possible security attacks. This is especially important as more and more products are internationalized. This document describes some of the security considerations that programmers, system analysts, standards developers, and users should take into account, and provides specific recommendations to reduce the risk of problems.
Web Content Accessibility Guidelines (WCAG) 2.1 [WCAG21]
Web Content Accessibility Guidelines (WCAG) 2.1 covers a wide range of recommendations for making Web content more accessible. Following these guidelines will make content more accessible to a wider range of people with disabilities, including blindness and low vision, deafness and hearing loss, limited movement, speech disabilities, photosensitivity, and combinations of these, and some accommodation for learning disabilities and cognitive limitations; but will not address every user need for people with these disabilities. These guidelines address accessibility of web content on desktops, laptops, tablets, and mobile devices. Following these guidelines will also often make Web content more usable to users in general.

WCAG 2.1 extends Web Content Accessibility Guidelines 2.0 [WCAG20], which was published as a W3C Recommendation December 2008. Content that conforms to WCAG 2.1 also conforms to WCAG 2.0, and therefore to policies that reference WCAG 2.0.

Authoring Tool Accessibility Guidelines (ATAG) 2.0 [ATAG20]
This specification provides guidelines for designing Web content authoring tools that are more accessible for people with disabilities. An authoring tool that conforms to these guidelines will promote accessibility by providing an accessible user interface to authors with disabilities as well as by enabling, supporting, and promoting the production of accessible Web content by all authors.
User Agent Accessibility Guidelines (UAAG) 2.0 [UAAG20]
This document provides guidelines for designing user agents that lower barriers to Web accessibility for people with disabilities. User agents include browsers and other types of software that retrieve and render Web content. A user agent that conforms to these guidelines will promote accessibility through its own user interface and through other internal facilities, including its ability to communicate with other technologies (especially assistive technologies). Furthermore, all users, not just users with disabilities, should find conforming user agents to be more usable.
HTML Accessibility APIs Mappings 1.0 [html-aam-1.0]
Defines how user agents map HTML 5.1 elements and attributes to platform accessibility APIs. Documenting these mappings promotes interoperable exposure of roles, states, properties, and events implemented by accessibility APIs and helps to ensure that this information appears in a manner consistent with author intent.

2. Common infrastructure

2.1. Terminology

This specification refers to both HTML and XML attributes and IDL attributes, often in the same context. When it is not clear which is being referred to, they are referred to as content attributes for HTML and XML attributes, and IDL attributes for those defined on IDL interfaces. Similarly, the term "properties" is used for both JavaScript object properties and CSS properties. When these are ambiguous they are qualified as object properties and CSS properties respectively.

Generally, when the specification states that a feature applies to the HTML syntax or the XHTML syntax, it also includes the other. When a feature specifically only applies to one of the two languages, it is called out by explicitly stating that it does not apply to the other format, as in "for HTML, ... (this does not apply to XHTML)".

This specification uses the term document to refer to any use of HTML, ranging from short static documents to long essays or reports with rich multimedia, as well as to fully-fledged interactive applications. The term is used to refer both to Document objects and their descendant DOM trees, and to serialized byte streams using the HTML syntax or XHTML syntax, depending on context.

In the context of the DOM structures, the terms HTML document and XML document are used as defined in the DOM specification, and refer specifically to two different modes that Document objects can find themselves in. [DOM] (Such uses are always hyperlinked to their definition.)

In the context of byte streams, the term HTML document refers to resources labeled as text/html, and the term XML document refers to resources labeled with an XML MIME type.

The term XHTML document is used to refer to both Documents in the XML document mode that contains element nodes in the HTML namespace, and byte streams labeled with an XML MIME type that contain elements from the HTML namespace, depending on context.


For simplicity, terms such as shown, displayed, and visible might sometimes be used when referring to the way a document is rendered to the user. These terms are not meant to imply a visual medium; they must be considered to apply to other media in equivalent ways.

When an algorithm B says to return to another algorithm A, it implies that A called B. Upon returning to A, the implementation must continue from where it left off in calling B. Some algorithms run in parallel; this means that the algorithm’s subsequent steps are to be run, one after another, at the same time as other logic in the specification (e.g., at the same time as the event loop). This specification does not define the precise mechanism by which this is achieved, be it time-sharing cooperative multitasking, fibers, threads, processes, using different hyperthreads, cores, CPUs, machines, etc. By contrast, an operation that is to run immediately must interrupt the currently running task, run itself, and then resume the previously running task.

The term "transparent black" refers to the color with red, green, blue, and alpha channels all set to zero.

2.1.1. Resources

The specification uses the term supported when referring to whether a user agent has an implementation capable of decoding the semantics of an external resource. A format or type is said to be supported if the implementation can process an external resource of that format or type without critical aspects of the resource being ignored. Whether a specific resource is supported can depend on what features of the resource’s format are in use.

For example, a PNG image would be considered to be in a supported format if its pixel data could be decoded and rendered, even if, unbeknownst to the implementation, the image also contained animation data.

An MPEG-4 video file would not be considered to be in a supported format if the compression format used was not supported, even if the implementation could determine the dimensions of the movie from the file’s metadata.

What some specifications, in particular the HTTP specification, refer to as a representation is referred to in this specification as a resource. [HTTP]

The term MIME type is used to refer to what is sometimes called an Internet media type in protocol literature. The term media type in this specification is used to refer to the type of media intended for presentation, as used by the CSS specifications. [RFC2046] [MEDIAQ]

A string is a valid MIME type if it matches the media-type rule. In particular, a valid MIME type may include MIME type parameters. [HTTP]

A string is a valid MIME type with no parameters if it matches the media-type rule, but does not contain any U+003B SEMICOLON characters (;). In other words, if it consists only of a type and subtype, with no MIME Type parameters. [HTTP]

The term HTML MIME type is used to refer to the MIME type text/html.

A resource’s critical subresources are those that the resource needs to have available to be correctly processed. Which resources are considered critical or not is defined by the specification that defines the resource’s format.

2.1.2. XML compatibility

To ease migration from HTML to XHTML, user agents conforming to this specification will place elements in HTML in the http://www.w3.org/1999/xhtml namespace, at least for the purposes of the DOM and CSS. The term "HTML elements", when used in this specification, refers to any element in that namespace, and thus refers to both HTML and XHTML elements.

Except where otherwise stated, all elements defined or mentioned in this specification are in the HTML namespace ("http://www.w3.org/1999/xhtml"), and all attributes defined or mentioned in this specification have no namespace.

The term element type is used to refer to the set of elements that have a given local name and namespace. For example, button elements are elements with the element type button, meaning they have the local name "button" and (implicitly as defined above) the HTML namespace.

Attribute names are said to be XML-compatible if they match the Name production defined in XML and they contain no U+003A COLON characters (:). [XML]

The term XML MIME type is used to refer to the MIME types text/xml, application/xml, and any MIME type whose subtype ends with the four characters "+xml". [RFC7303]

2.1.3. DOM trees

When it is stated that some element or attribute is ignored, or treated as some other value, or handled as if it was something else, this refers only to the processing of the node after it is in the DOM. A user agent must not mutate the DOM in such situations.

A content attribute is said to change value only if its new value is different than its previous value; setting an attribute to a value it already has does not change it.

The term empty, when used for an attribute value, Text node, or string means that the length of the text is zero (i.e., not even containing spaces or control characters).

An element’s child text content is the concatenation of the data of all the Text nodes that are children of the element (ignoring any other nodes such as comments or elements), in tree order.

A node A is inserted into a node B when the insertion steps are invoked with A as the argument and A’s new parent is B. Similarly, a node A is removed from a node B when the removing steps are invoked with A as the removedNode argument and B as the oldParent argument.

A node is inserted into a document when the insertion steps are invoked with it as the argument and it is now in a document tree. Analogously, a node is removed from a document when the removing steps are invoked with it as the argument and it is now no longer in a document tree.

2.1.4. Scripting

The construction "a Foo object", where Foo is actually an interface, is sometimes used instead of the more accurate "an object implementing the interface Foo".

An IDL attribute is said to be getting when its value is being retrieved (e.g., by author script), and is said to be setting when a new value is assigned to it.

If a DOM object is said to be live, then the attributes and methods on that object must operate on the actual underlying data, not a snapshot of the data.

In the contexts of events, the terms fire and dispatch are used as defined in the DOM specification: firing an event means to create and dispatch it, and dispatching an event means to follow the steps that propagate the event through the tree. The term trusted event is used to refer to events whose isTrusted attribute is initialized to true. [DOM]

2.1.5. Plugins

The term plugin refers to a user-agent defined set of content handlers that can be used by the user agent. The content handlers can take part in the user agent’s rendering of a Document object, but that neither act as child browsing contexts of the Document nor introduce any Node objects to the Document's DOM.

Typically such content handlers are provided by third parties, though a user agent can also designate built-in content handlers as plugins.

A user agent must not consider the types text/plain and application/octet-stream as having a registered plugin.

One example of a plugin would be a PDF viewer that is instantiated in a browsing context when the user navigates to a PDF file. This would count as a plugin regardless of whether the party that implemented the PDF viewer component was the same as that which implemented the user agent itself. However, a PDF viewer application that launches separate from the user agent (as opposed to using the same interface) is not a plugin by this definition.

This specification does not define a mechanism for interacting with plugins, as it is expected to be user-agent- and platform-specific. Some user agents might opt to support a plugin mechanism such as the Netscape Plugin API; others might use remote content converters or have built-in support for certain types. Indeed, this specification doesn’t require user agents to support plugins at all. [NPAPI]

A plugin can be secured if it honors the semantics of the sandbox attribute.

For example, a secured plugin would prevent its contents from creating pop-up windows when the plugin is instantiated inside a sandboxed iframe.

Browsers should take extreme care when interacting with external content intended for plugins. When third-party software is run with the same privileges as the user agent itself, vulnerabilities in the third-party software become as dangerous as if they were vulnerabilities of the user agent itself.

Since different users having different sets of plugins provides a fingerprinting vector that increases the chances of users being uniquely identified, user agents are encouraged to support the exact same set of plugins for each user. (This is a fingerprinting vector.)

2.1.6. Character encodings

A character encoding, or just encoding where that is not ambiguous, is a defined way to convert between byte streams and Unicode strings, as defined in the WHATWG Encoding specification. An encoding has an encoding name and one or more encoding labels, referred to as the encoding’s name and labels in the Encoding specification. [ENCODING]

A UTF-16 encoding is UTF-16BE or UTF-16LE. [ENCODING]

An ASCII-compatible encoding is any encoding that is not a UTF-16 encoding. [ENCODING]

Since support for encodings that are not defined in the WHATWG Encoding specification is prohibited, UTF-16 encodings are the only encodings that this specification needs to treat as not being ASCII-compatible encodings.

The term code unit is used as defined in the Web IDL specification: a 16 bit unsigned integer, the smallest atomic component of a DOMString. (This is a narrower definition than the one used in Unicode, and is not the same as a code point.) [WEBIDL]

The term Unicode code point means a Unicode scalar value where possible, and an isolated surrogate code point when not. When a conformance requirement is defined in terms of characters or Unicode code points, a pair of code units consisting of a high surrogate followed by a low surrogate must be treated as the single code point represented by the surrogate pair, but isolated surrogates must each be treated as the single code point with the value of the surrogate. [UNICODE]

In this specification, the term character, when not qualified as Unicode character, is synonymous with the term Unicode code point.

The term Unicode character is used to mean a Unicode scalar value (i.e. any Unicode code point that is not a surrogate code point). [UNICODE]

The code-unit length of a string is the number of code units in that string.

This complexity results from the historical decision to define the DOM API in terms of 16 bit (UTF-16) code units, rather than in terms of Unicode characters.

2.2. Conformance requirements

All diagrams, examples, and notes in this specification are non-normative, as are all sections explicitly marked non-normative. Everything else in this specification is normative.

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in the normative parts of this document are to be interpreted as described in RFC2119. The key word "OPTIONALLY" in the normative parts of this document is to be interpreted with the same normative meaning as "MAY" and "OPTIONAL". For readability, these words do not appear in all uppercase letters in this specification. [RFC2119]

Requirements phrased in the imperative as part of algorithms (such as "strip any leading space characters" or "return false and abort these steps") are to be interpreted with the meaning of the key word ("must", "should", "may", etc) used in introducing the algorithm.

For example, were the spec to say:
To eat an orange, the user must:
1. Peel the orange.
2. Separate each slice of the orange.
3. Eat the orange slices.

...it would be equivalent to the following:

To eat an orange:
1. The user must peel the orange.
2. The user must separate each slice of the orange.
3. The user must eat the orange slices.

Here the key word is "must".

The former (imperative) style is generally preferred in this specification for stylistic reasons.

Conformance requirements phrased as algorithms or specific steps may be implemented in any manner, so long as the end result is equivalent. (In particular, the algorithms defined in this specification are intended to be easy to follow, and not intended to be performant.)

2.2.1. Conformance classes

This specification describes the conformance criteria for user agents (relevant to implementors) and documents (relevant to authors and authoring tool implementors).

Conforming documents are those that comply with all the conformance criteria for documents. For readability, some of these conformance requirements are phrased as conformance requirements on authors; such requirements are implicitly requirements on documents: by definition, all documents are assumed to have had an author. (In some cases, that author may itself be a user agent — such user agents are subject to additional rules, as explained below.)

For example, if a requirement states that "authors must not use the foobar element", it would imply that documents are not allowed to contain elements named foobar.

There is no implied relationship between document conformance requirements and implementation conformance requirements. User agents are not free to handle non-conformant documents as they please; the processing model described in this specification applies to implementations regardless of the conformity of the input documents.

User agents fall into several (overlapping) categories with different conformance requirements.

Web browsers and other interactive user agents

Web browsers that support the XHTML syntax must process elements and attributes from the HTML namespace found in XML documents as described in this specification, so that users can interact with them, unless the semantics of those elements have been overridden by other specifications.

A conforming XHTML processor would, upon finding an XHTML script element in an XML document, execute the script contained in that element. However, if the element is found within a transformation expressed in XSLT (assuming the user agent also supports XSLT), then the processor would instead treat the script element as an opaque element that forms part of the transform.

Web browsers that support the HTML syntax must process documents labeled with an HTML MIME type as described in this specification, so that users can interact with them.

User agents that support scripting must also be conforming implementations of the IDL fragments in this specification, as described in the Web IDL specification. [WEBIDL]

Unless explicitly stated, specifications that override the semantics of HTML elements do not override the requirements on DOM objects representing those elements. For example, the script element in the example above would still implement the HTMLScriptElement interface.

Non-interactive presentation user agents

User agents that process HTML and XHTML documents purely to render non-interactive versions of them must comply to the same conformance criteria as Web browsers, except that they are exempt from requirements regarding user interaction.

Typical examples of non-interactive presentation user agents are printers (static user agents) and overhead displays (dynamic user agents). It is expected that most static non-interactive presentation user agents will also opt to lack scripting support.

A non-interactive but dynamic presentation user agent would still execute scripts, allowing forms to be dynamically submitted, and so forth. However, since the concept of "focus" is irrelevant when the user cannot interact with the document, the user agent would not need to support any of the focus-related DOM APIs.

Visual user agents that support the suggested default rendering

User agents, whether interactive or not, may be designated (possibly as a user option) as supporting the suggested default rendering defined by this specification.

This is not required. In particular, even user agents that do implement the suggested default rendering are encouraged to offer settings that override this default to improve the experience for the user, e.g., changing the color contrast, using different focus styles, or otherwise making the experience more accessible and usable to the user.

User agents that are designated as supporting the suggested default rendering must, while so designated, implement the rules in §10 Rendering. That section defines the behavior that user agents are expected to implement.

User agents with no scripting support

Implementations that do not support scripting (or which have their scripting features disabled entirely) are exempt from supporting the events and DOM interfaces mentioned in this specification. For the parts of this specification that are defined in terms of an events model or in terms of the DOM, such user agents must still act as if events and the DOM were supported.

Scripting can form an integral part of an application. Web browsers that do not support scripting, or that have scripting disabled, might be unable to fully convey the author’s intent.

Conformance checkers

Conformance checkers must verify that a document conforms to the applicable conformance criteria described in this specification. Automated conformance checkers are exempt from detecting errors that require interpretation of the author’s intent (for example, while a document is non-conforming if the content of a blockquote element is not a quote, conformance checkers running without the input of human judgement do not have to check that blockquote elements only contain quoted material).

Conformance checkers must check that the input document conforms when parsed without a browsing context (meaning that no scripts are run, and that the parser’s scripting flag is disabled), and should also check that the input document conforms when parsed with a browsing context in which scripts execute, and that the scripts never cause non-conforming states to occur other than transiently during script execution itself. (This is only a "SHOULD" and not a "MUST" requirement because it has been proven to be impossible. [COMPUTABLE])

The term "HTML validator" can be used to refer to a conformance checker that itself conforms to the applicable requirements of this specification.

XML DTDs cannot express all the conformance requirements of this specification. Therefore, a validating XML processor and a DTD cannot constitute a conformance checker. Also, since neither of the two authoring formats defined in this specification are applications of SGML, a validating SGML system cannot constitute a conformance checker either.

To put it another way, there are three types of conformance criteria:

  1. Criteria that can be expressed in a DTD.

  2. Criteria that cannot be expressed by a DTD, but can still be checked by a machine.

  3. Criteria that can only be checked by a human.

A conformance checker must check for the first two. A simple DTD-based validator only checks for the first class of errors and is therefore not a conforming conformance checker according to this specification.

Data mining tools

Applications and tools that process HTML and XHTML documents for reasons other than to either render the documents or check them for conformance should act in accordance with the semantics of the documents that they process.

A tool that generates document outlines but increases the nesting level for each paragraph and does not increase the nesting level for each section would not be conforming.

Authoring tools and markup generators

Authoring tools and markup generators must generate conforming documents. Conformance criteria that apply to authors also apply to authoring tools, where appropriate.

Authoring tools are exempt from the strict requirements of using elements only for their specified purpose, but only to the extent that authoring tools are not yet able to determine author intent. However, authoring tools must not automatically misuse elements or encourage their users to do so.

For example, it is not conforming to use an address element for arbitrary contact information; that element can only be used for marking up contact information for the author of the document or section. However, since an authoring tool is likely unable to determine the difference, an authoring tool is exempt from that requirement. This does not mean, though, that authoring tools can use address elements for any block of italics text (for instance); it just means that the authoring tool doesn’t have to verify, if a user inserts contact information for a section or something else.

In terms of conformance checking, an editor has to output documents that conform to the same extent that a conformance checker will verify.

When an authoring tool is used to edit a non-conforming document, it may preserve the conformance errors in sections of the document that were not edited during the editing session (i.e., an editing tool is allowed to round-trip erroneous content). However, an authoring tool must not claim that the output is conformant if errors have been so preserved.

Authoring tools are expected to come in two broad varieties: tools that work from structure or semantic data, and tools that work on a What-You-See-Is-What-You-Get media-specific editing basis (WYSIWYG).

The former is the preferred mechanism for tools that author HTML, since the structure in the source information can be used to make informed choices regarding which HTML elements and attributes are most appropriate.

However, WYSIWYG tools are legitimate. WYSIWYG tools should use elements they know are appropriate, and should not use elements that they do not know to be appropriate. This might in certain extreme cases mean limiting the use of flow elements to just a few elements, like div, b, i, and span and making liberal use of the style attribute.

All authoring tools, whether WYSIWYG or not, should make a best effort attempt at enabling users to create well-structured, semantically rich, media-independent content.

User agents may impose implementation-specific limits on otherwise unconstrained inputs, e.g., to prevent denial of service attacks, to guard against running out of memory, or to work around platform-specific limitations. (This is a fingerprinting vector.)

For compatibility with existing content and prior specifications, this specification describes two authoring formats: one based on XML (referred to as the XHTML syntax), and one using a custom format inspired by SGML (referred to as the HTML syntax). Implementations must support at least one of these two formats, although supporting both is encouraged.

Some conformance requirements are phrased as requirements on elements, attributes, methods or objects. Such requirements fall into two categories: those describing content model restrictions, and those describing implementation behavior. Those in the former category are requirements on documents and authoring tools. Those in the second category are requirements on user agents. Similarly, some conformance requirements are phrased as requirements on authors; such requirements are to be interpreted as conformance requirements on the documents that authors produce. (In other words, this specification does not distinguish between conformance criteria on authors and conformance criteria on documents.)

2.2.2. Dependencies

This specification relies on several other underlying specifications.

Unicode and Encoding

The Unicode character set is used to represent textual data, and the Encoding specification defines requirements around character encodings. [UNICODE]

This specification introduces terminology based on the terms defined in those specifications, as described earlier.

The following terms are used as defined in the Encoding specification: [ENCODING]

  • Getting an encoding

  • Get an output encoding

  • The generic decode algorithm which takes a byte stream and an encoding and returns a character stream

  • The UTF-8 decode algorithm which takes a byte stream and returns a character stream, additionally stripping one leading UTF-8 Byte Order Mark (BOM), if any

  • The UTF-8 decode without BOM algorithm which is identical to UTF-8 decode except that it does not strip one leading UTF-8 Byte Order Mark (BOM)

  • The UTF-8 decode without BOM or fail algorithm which is identical to UTF-8 decode without BOM except that it returns failure upon encountering an error

  • The encode algorithm which takes a character stream and an encoding and returns a byte stream

  • The UTF-8 encode algorithm which takes a character stream and returns a byte stream.

XML and related specifications

Implementations that support the XHTML syntax must support some version of XML, as well as its corresponding namespaces specification, because that syntax uses an XML serialization with namespaces. [XML] [XML-NAMES]

The attribute with the tag name xml:space in the XML namespace is defined by the XML specification. [XML]

This specification also references the <?xml-stylesheet?> processing instruction, defined in the Associating Style Sheets with XML documents specification. [XML-STYLESHEET]

This specification also non-normatively mentions the XSLTProcessor interface and its transformToFragment() and transformToDocument() methods. [XSLTP]

URLs

The following terms are defined in the WHATWG URL specification: [URL]

A number of schemes and protocols are referenced by this specification also:

Media fragment syntax is defined in the Media Fragments URI specification. [MEDIA-FRAGS]

HTTP and related specifications

The following terms are defined in the HTTP specifications: [HTTP]

The following terms are defined in the Cookie specification: [COOKIES]

The following term is defined in the Web Linking specification: [RFC8288]

Fetch

The following terms are defined in the WHATWG Fetch specification: [FETCH]

The following terms are defined in Referrer Policy [REFERRERPOLICY]

  • referrer policy

  • The Referrer-Policy HTTP header

  • The parse a referrer policy from a Referrer-Policy header algorithm

  • The "no-referrer", "no-referrer-when-downgrade", and "unsafe-url" referrer policies

Web IDL

The IDL fragments in this specification must be interpreted as required for conforming IDL fragments, as described in the Web IDL specification. [WEBIDL]

The following terms are defined in the Web IDL specification:

The Web IDL specification also defines the following types that are used in Web IDL fragments in this specification:

The term throw in this specification is used as defined in the WebIDL specification. The following exception names are defined by WebIDL and used by this specification:

When this specification requires a user agent to create a Date object representing a particular time (which could be the special value Not-a-Number), the milliseconds component of that time, if any, must be truncated to an integer, and the time value of the newly created Date object must represent the resulting truncated time.

For instance, given the time 23045 millionths of a second after 01:00 UTC on January 1st 2000, i.e., the time 2000-01-01T00:00:00.023045Z, then the Date object created representing that time would represent the same time as that created representing the time 2000-01-01T00:00:00.023Z, 45 millionths earlier. If the given time is NaN, then the result is a Date object that represents a time value NaN (indicating that the object does not represent a specific instant of time).

JavaScript

Some parts of the language described by this specification only support JavaScript as the underlying scripting language. [ECMA-262]

The term "JavaScript" is used to refer to ECMA262, rather than the official term ECMAScript, since the term JavaScript is more widely known. Similarly, the MIME type used to refer to JavaScript in this specification is text/javascript, since that is the most commonly used type, despite it being an officially obsoleted type according to RFC 4329. [RFC4329]

The following terms are defined in the JavaScript specification and used in this specification [ECMA-262]:

DOM

The Document Object Model (DOM) is a representation — a model — of a document and its content. The DOM is not just an API; the conformance criteria of HTML implementations are defined, in this specification, in terms of operations on the DOM. [DOM]

Implementations must support DOM and the events defined in UI Events, because this specification is defined in terms of the DOM, and some of the features are defined as extensions to the DOM interfaces. [DOM] [UIEVENTS]

In particular, the following features are defined in the DOM specification: [DOM]

For example, to throw a TimeoutError exception, a user agent would construct a DOMException object whose type was the string "TimeoutError" (and whose code was the number 23, for legacy reasons) and actually throw that object as an exception.

The following features are defined in the UI Events specification: [UIEVENTS]

The following features are defined in the Touch Events specification: [TOUCH-EVENTS]

This specification sometimes uses the term name to refer to the event’s type; as in, "an event named click" or "if the event name is keypress". The terms "name" and "type" for events are synonymous.

The following features are defined in the DOM Parsing and Serialization specification: [DOM-PARSING]

The Selection interface is defined in the Selection API specification. [SELECTION-API]

User agents are also encouraged to implement the features described in the HTML Editing APIs and UndoManager and DOM Transaction specifications. [EDITING] [UNDO]

The following parts of the Fullscreen specification are referenced from this specification, in part to define the rendering of dialog elements, and also to define how the Fullscreen API interacts with the sandboxing features in HTML: [FULLSCREEN]

The High Resolution Time specification provides the DOMHighResTimeStamp typedef and the Performance object’s now() method. [HR-TIME-2]

File API

This specification uses the following features defined in the File API specification: [FILEAPI]

Indexed Database API

This specification uses cleanup Indexed Database transactions defined by the Indexed Database API specification. [INDEXEDDB]

Media Source Extensions

The following terms are defined in the Media Source Extensions specification: [MEDIA-SOURCE]

Media Capture and Streams

The following term is defined in the Media Capture and Streams specification: [MEDIACAPTURE-STREAMS]

XMLHttpRequest

This specification references the XMLHttpRequest specification to describe how the two specifications interact and to use its ProgressEvent features. The following features and terms are defined in the XMLHttpRequest specification: [XHR]

Server-Sent Events

This specification references EventSource which is specified in the Server-Sent Events specification [EVENTSOURCE]

Media Queries

Implementations must support the Media Queries language. [MEDIAQ]

<media-condition>

CSS modules

While support for CSS as a whole is not required of implementations of this specification (though it is encouraged, at least for Web browsers), some features are defined in terms of specific CSS requirements.

In particular, some features require that a string be parsed as a CSS <color> value. When parsing a CSS value, user agents are required by the CSS specifications to apply some error handling rules. These apply to this specification also. [CSS3COLOR] [CSS-2015]

For example, user agents are required to close all open constructs upon finding the end of a style sheet unexpectedly. Thus, when parsing the string "rgb(0,0,0" (with a missing close-parenthesis) for a color value, the close parenthesis is implied by this error handling rule, and a value is obtained (the color black). However, the similar construct "rgb(0,0," (with both a missing parenthesis and a missing "blue" value) cannot be parsed, as closing the open construct does not result in a viable value.

The following terms and features are defined in the CSS specification: [CSS-2015]

  • viewport

  • replaced element

  • intrinsic dimensions

The term named color is defined in the CSS Color specification. [CSS3COLOR]

The terms intrinsic width and intrinsic height refer to the width dimension and the height dimension, respectively, of intrinsic dimensions.

The term paint source is used as defined in the CSS Image Values and Replaced Content specification to define the interaction of certain HTML elements with the CSS 'element()' function. [CSS3-IMAGES]

The term default object size is also defined in the CSS Image Values and Replaced Content specification. [CSS3-IMAGES]

Implementations that support scripting must support the CSS Object Model. The following features and terms are defined in the CSSOM specifications: [CSSOM] [CSSOM-VIEW]

The following features and terms are defined in the CSS Syntax specifications: [CSS-SYNTAX-3]

The following terms are defined in the Selectors specification: [SELECTORS4]

The feature <length> is defined in the CSS Values and Units specification. [CSS-VALUES]

The term style attribute is defined in the CSS Style Attributes specification. [CSS-STYLE-ATTR]

The term used value is defined in the CSS Cascading and Inheritance specification. [CSS-CASCADE-4]

The CanvasRenderingContext2D object’s use of fonts depends on the features described in the CSS Fonts and Font Loading specifications, including in particular FontFace objects and the font source concept. [CSS-FONTS-3] [CSS-FONT-LOADING-3]

The following interface is defined in the Geometry Interfaces Module specification: [GEOMETRY-1]

SVG

The CanvasRenderingContext2D object’s use of fonts depends on the features described in the CSS Fonts and Font Loading specifications, including in particular FontFace objects and the font source concept. [CSS-FONTS-3] [CSS-FONT-LOADING-3]

The following interface is defined in the SVG specification: [SVG11]

WebGL

The following interface is defined in the WebGL specification: [WEBGL-1]

WebGL2

The following interface is defined in the WebGL2 specification: [webgl-2]

WebVTT

Implementations may support WebVTT as a text track format for subtitles, captions, chapter titles, metadata, etc, for media resources. [WEBVTT]

The following terms, used in this specification, are defined in the WebVTT specification:

  • WebVTT file

  • WebVTT file using cue text

  • WebVTT file using chapter title text

  • WebVTT file using only nested cues

  • WebVTT parser

  • The rules for updating the display of WebVTT text tracks

  • The rules for interpreting WebVTT cue text

  • The WebVTT text track cue writing direction

The WebSocket protocol

The following terms are defined in the WebSocket protocol specification: [RFC6455]

  • establish a WebSocket connection

  • the WebSocket connection is established

  • validate the server’s response

  • extensions in use

  • subprotocol in use

  • headers to send appropriate cookies

  • cookies set during the server’s opening handshake

  • a WebSocket message has been received

  • send a WebSocket Message

  • fail the WebSocket connection

  • close the WebSocket connection

  • start the WebSocket closing handshake

  • the WebSocket closing handshake is started

  • the WebSocket connection is closed (possibly cleanly)

  • the WebSocket connection close code

  • the WebSocket connection close reason

  • Sec-WebSocket-Protocol field

ARIA

The role attribute is defined in the ARIA specification, as are the following roles: [wai-aria-1.1]

In addition, the following aria-* content attributes are defined in the ARIA specification: [wai-aria-1.1]

Content Security Policy

The following terms are defined in Content Security Policy: [CSP3]

The following terms are defined in Content Security Policy: Document Features

Service Workers

The following terms are defined in Service Workers: [SERVICE-WORKERS]

  • client message queue

  • match service worker registration

  • ServiceWorkerContainer

Secure Contexts

The following term is defined in Secure Contexts: [SECURE-CONTEXTS]

  • Is environment settings object a secure context?

Payment Request API

The following term is defined in the Payment Request API specification: [PAYMENT-REQUEST]

MathML

While support for MathML as a whole is not required by this specification (though it is encouraged, at least for Web browsers), certain features depend upon small parts of MathML being implemented. [MATHML]

The following features are defined in the MathML specification:

  • MathML annotation-xml element

  • MathML math element

  • MathML merror element

  • MathML mi element

  • MathML mn element

  • MathML mo element

  • MathML ms element

  • MathML mtext element

SVG

While support for SVG as a whole is not required by this specification (though it is encouraged, at least for Web browsers), certain features depend upon parts of SVG being implemented.

Also, the SVG specifications do not reflect implementation reality. Implementations implement subsets of SVG 1.1 and SVG Tiny 1.2. Although it is hoped that the in-progress SVG 2 specification is a more realistic target for implementations, until that specification is ready, user agents that implement SVG must do so with the following willful violations and additions. [SVG11] [SVGTINY12] [SVG2]

User agents that implement SVG must not implement the following features from SVG 1.1:

  • The tref element

  • The cursor element (use CSS’s cursor property instead)

  • The font-defining SVG elements: font, glyph, missing-glyph, hkern, vkern, font-face, font-face-src, font-face-uri, font-face-format, and font-face-name (use CSS’s @font-face instead)

  • The externalResourcesRequired attribute

  • The enable-background property

  • The contentScriptType and contentStyleType attributes (use the type attribute on the SVG script and style elements instead)

User agents that implement SVG must implement the following features from SVG Tiny 1.2:

  • The non-scaling-stroke value for the vector-effect property

  • The class attribute is allowed on all SVG elements

  • The tabindex attribute is allowed on visible SVG elements

  • The ARIA accessibility attributes are allowed on all SVG elements

The following features are defined in the SVG specifications:

Filter Effects

The following feature is defined in the Filter Effects specification:

  • <filter-function-list>


This specification does not require support of any particular network protocol, style sheet language, scripting language, or any of the DOM specifications beyond those required in the list above. However, the language described by this specification is biased towards CSS as the styling language, JavaScript as the scripting language, and HTTP as the network protocol, and several features assume that those languages and protocols are in use.

A user agent that implements the HTTP protocol must implement the Web Origin Concept specification and the HTTP State Management Mechanism specification (Cookies) as well. [HTTP] [ORIGIN] [COOKIES]

This specification might have certain additional requirements on character encodings, image formats, audio formats, and video formats in the respective sections.

2.2.3. Extensibility

Vendor-specific proprietary user agent extensions to this specification are strongly discouraged. Documents must not use such extensions, as doing so reduces interoperability and fragments the user base, allowing only users of specific user agents to access the content in question.

If such extensions are nonetheless needed, e.g., for experimental purposes, then vendors are strongly urged to use one of the following extension mechanisms:

Attribute names beginning with the two characters "x-" are reserved for user agent use and are guaranteed to never be formally added to the HTML language. For flexibility, attributes names containing underscores (the U+005F LOW LINE character) are also reserved for experimental purposes and are guaranteed to never be formally added to the HTML language.

Pages that use such attributes are by definition non-conforming.

For DOM extensions, e.g., new methods and IDL attributes, the new members should be prefixed by vendor-specific strings to prevent clashes with future versions of this specification.

For events, experimental event types should be prefixed with vendor-specific strings.

For example, if a user agent called "Pleasold" were to add an event to indicate when the user is going up in an elevator, it could use the prefix "pleasold" and thus name the event "pleasoldgoingup", possibly with an event handler attribute named "onpleasoldgoingup".

All extensions must be defined so that the use of extensions neither contradicts nor causes the non-conformance of functionality defined in the specification.

For example, while strongly discouraged from doing so, an implementation "Foo Browser" could add a new IDL attribute "fooTypeTime" to a control’s DOM interface that returned the time it took the user to select the current value of a control (say). On the other hand, defining a new control that appears in a form’s elements array would be in violation of the above requirement, as it would violate the definition of elements given in this specification.

When adding new reflecting IDL attributes corresponding to content attributes of the form "x-vendor-feature", the IDL attribute should be named "vendorFeature" (i.e., the "x" is dropped from the IDL attribute’s name).


When vendor-neutral extensions to this specification are needed, either this specification can be updated accordingly, or an extension specification can be written that overrides the requirements in this specification. When someone applying this specification to their activities decides that they will recognize the requirements of such an extension specification, it becomes an applicable specification for the purposes of conformance requirements in this specification.

Someone could write a specification that defines any arbitrary byte stream as conforming, and then claim that their random junk is conforming. However, that does not mean that their random junk actually is conforming for everyone’s purposes: if someone else decides that the specification does not apply to their work, then they can quite legitimately say that the aforementioned random junk is just that, junk, and not conforming at all. As far as conformance goes, what matters in a particular community is what that community agrees is applicable.

applicable specification.

The conformance terminology for documents depends on the nature of the changes introduced by such applicable specifications, and on the content and intended interpretation of the document. Applicable specifications MAY define new document content (e.g., a foobar element), MAY prohibit certain otherwise conforming content (e.g., prohibit use of <table>s), or MAY change the semantics, DOM mappings, or other processing rules for content defined in this specification. Whether a document is or is not a conforming HTML document does not depend on the use of applicable specifications: if the syntax and semantics of a given conforming HTML document is unchanged by the use of applicable specification(s), then that document remains a conforming HTML document. If the semantics or processing of a given (otherwise conforming) document is changed by use of applicable specification(s), then it is not a conforming HTML document. For such cases, the applicable specifications SHOULD define conformance terminology.

As a suggested but not required convention, such specifications might define conformance terminology such as: "Conforming HTML+XXX document", where XXX is a short name for the applicable specification. (Example: "Conforming HTML+AutomotiveExtensions document").

a consequence of the rule given above is that certain syntactically correct HTML documents may not be conforming HTML documents in the presence of applicable specifications. (Example: the applicable specification defines <table> to be a piece of furniture — a document written to that specification and containing a <table> element is NOT a conforming HTML document, even if the element happens to be syntactically correct HTML.)


User agents must treat elements and attributes that they do not understand as semantically neutral; leaving them in the DOM (for DOM processors), and styling them according to CSS (for CSS processors), but not inferring any meaning from them.

When support for a feature is disabled (e.g., as an emergency measure to mitigate a security problem, or to aid in development, or for performance reasons), user agents must act as if they had no support for the feature whatsoever, and as if the feature was not mentioned in this specification. For example, if a particular feature is accessed via an attribute in a Web IDL interface, the attribute itself would be omitted from the objects that implement that interface — leaving the attribute on the object but making it return null or throw an exception is insufficient.

2.2.4. Interactions with XPath and XSLT

Implementations of XPath 1.0 that operate on HTML documents parsed or created in the manners described in this specification (e.g., as part of the document.evaluate() API) must act as if the following edit was applied to the XPath 1.0 specification.

First, remove this paragraph:

A QName in the node test is expanded into an expanded-name using the namespace declarations from the expression context. This is the same way expansion is done for element type names in start and end-tags except that the default namespace declared with xmlns is not used: if the QName does not have a prefix, then the namespace URI is null (this is the same way attribute names are expanded). It is an error if the QName has a prefix for which there is no namespace declaration in the expression context.

Then, insert in its place the following:

A QName in the node test is expanded into an expanded-name using the namespace declarations from the expression context. If the QName has a prefix, then there must be a namespace declaration for this prefix in the expression context, and the corresponding namespace URI is the one that is associated with this prefix. It is an error if the QName has a prefix for which there is no namespace declaration in the expression context.

If the QName has no prefix and the principal node type of the axis is element, then the default element namespace is used. Otherwise if the QName has no prefix, the namespace URI is null. The default element namespace is a member of the context for the XPath expression. The value of the default element namespace when executing an XPath expression through the DOM3 XPath API is determined in the following way:

  1. If the context node is from an HTML DOM, the default element namespace is "http://www.w3.org/1999/xhtml".

  2. Otherwise, the default element namespace URI is null.

This is equivalent to adding the default element namespace feature of XPath 2.0 to XPath 1.0, and using the HTML namespace as the default element namespace for HTML documents. It is motivated by the desire to have implementations be compatible with legacy HTML content while still supporting the changes that this specification introduces to HTML regarding the namespace used for HTML elements, and by the desire to use XPath 1.0 rather than XPath 2.0.

This change is a willful violation of the XPath 1.0 specification, motivated by desire to have implementations be compatible with legacy content while still supporting the changes that this specification introduces to HTML regarding which namespace is used for HTML elements. [XPATH]


XSLT 1.0 processors outputting to a DOM when the output method is "html" (either explicitly or via the defaulting rule in XSLT 1.0) are affected as follows:

If the transformation program outputs an element in no namespace, the processor must, prior to constructing the corresponding DOM element node, change the namespace of the element to the HTML namespace, ASCII-lowercase the element’s local name, and ASCII-lowercase the names of any non-namespaced attributes on the element.

This requirement is a willful violation of the XSLT 1.0 specification, required because this specification changes the namespaces and case-sensitivity rules of HTML in a manner that would otherwise be incompatible with DOM-based XSLT transformations. (Processors that serialize the output are unaffected.) [XSLT]


This specification does not specify precisely how XSLT processing interacts with the HTML parser infrastructure (for example, whether an XSLT processor acts as if it puts any elements into a stack of open elements). However, XSLT processors must stop parsing if they successfully complete, and must set the current document readiness first to "interactive" and then to "complete" if they are aborted.


This specification does not specify how XSLT interacts with the navigation algorithm, how it fits in with the event loop, nor how error pages are to be handled (e.g., whether XSLT errors are to replace an incremental XSLT output, or are rendered inline, etc).

There are also additional non-normative comments regarding the interaction of XSLT and HTML in the script element section, and of XSLT, XPath, and HTML in the template element section.

2.3. Case-sensitivity and string comparison

Comparing two strings in a case-sensitive manner means comparing them exactly, code point for code point.

Comparing two strings in an ASCII case-insensitive manner means comparing them exactly, code point for code point, except that the characters in the range U+0041 to U+005A (i.e., LATIN CAPITAL LETTER A to LATIN CAPITAL LETTER Z) and the corresponding characters in the range U+0061 to U+007A (i.e., LATIN SMALL LETTER A to LATIN SMALL LETTER Z) are considered to also match.

Except where otherwise stated, string comparisons must be performed in a case-sensitive manner.

Converting a string to ASCII uppercase means replacing all characters in the range U+0061 to U+007A (i.e., LATIN SMALL LETTER A to LATIN SMALL LETTER Z) with the corresponding characters in the range U+0041 to U+005A (i.e., LATIN CAPITAL LETTER A to LATIN CAPITAL LETTER Z).

Converting a string to ASCII lowercase means replacing all characters in the range U+0041 to U+005A (i.e., LATIN CAPITAL LETTER A to LATIN CAPITAL LETTER Z) with the corresponding characters in the range U+0061 to U+007A (i.e., LATIN SMALL LETTER A to LATIN SMALL LETTER Z).

A string pattern is a prefix match for a string s when pattern is not longer than s and truncating s to pattern’s length leaves the two strings as matches of each other.

2.4. Common microsyntaxes

There are various places in HTML that accept particular data types, such as dates or numbers. This section describes what the conformance criteria for content in those formats is, and how to parse them.

Implementors are strongly urged to carefully examine any third-party libraries they might consider using to implement the parsing of syntaxes described below. For example, date libraries are likely to implement error handling behavior that differs from what is required in this specification, since error-handling behavior is often not defined in specifications that describe date syntaxes similar to those used in this specification, and thus implementations tend to vary greatly in how they handle errors.

2.4.1. Common parser idioms

The space characters, for the purposes of this specification, are U+0020 SPACE, U+0009 CHARACTER TABULATION (tab), U+000A LINE FEED (LF), U+000C FORM FEED (FF), and U+000D CARRIAGE RETURN (CR).

The White_Space characters are those that have the Unicode property "White_Space" in the Unicode PropList.txt data file. [UNICODE]

This should not be confused with the "White_Space" value (abbreviated "WS") of the "Bidi_Class" property in the Unicode.txt data file.

The control characters are those whose Unicode "General_Category" property has the value "Cc" in the Unicode UnicodeData.txt data file. [UNICODE]

The uppercase ASCII letters are the characters in the range U+0041 LATIN CAPITAL LETTER A to U+005A LATIN CAPITAL LETTER Z.

The lowercase ASCII letters are the characters in the range U+0061 LATIN SMALL LETTER A to U+007A LATIN SMALL LETTER Z.

The ASCII letters are the characters that are either uppercase ASCII letters or lowercase ASCII letters.

The ASCII digits are the characters in the range U+0030 DIGIT ZERO (0) to U+0039 DIGIT NINE (9).

The alphanumeric ASCII characters are those that are either uppercase ASCII letters, lowercase ASCII letters, or ASCII digits.

The ASCII hex digits are the characters in the ranges U+0030 DIGIT ZERO (0) to U+0039 DIGIT NINE (9), U+0041 LATIN CAPITAL LETTER A to U+0046 LATIN CAPITAL LETTER F, and U+0061 LATIN SMALL LETTER A to U+0066 LATIN SMALL LETTER F.

The uppercase ASCII hex digits are the characters in the ranges U+0030 DIGIT ZERO (0) to U+0039 DIGIT NINE (9) and U+0041 LATIN CAPITAL LETTER A to U+0046 LATIN CAPITAL LETTER F only.

The lowercase ASCII hex digits are the characters in the ranges U+0030 DIGIT ZERO (0) to U+0039 DIGIT NINE (9) and U+0061 LATIN SMALL LETTER A to U+0066 LATIN SMALL LETTER F only.

Some of the micro-parsers described below follow the pattern of having an input variable that holds the string being parsed, and having a position variable pointing at the next character to parse in input.

For parsers based on this pattern, a step that requires the user agent to collect a sequence of characters means that the following algorithm must be run, with characters being the set of characters that can be collected:

  1. Let input and position be the same variables as those of the same name in the algorithm that invoked these steps.

  2. Let result be the empty string.

  3. While position doesn’t point past the end of input and the character at position is one of the characters, append that character to the end of result and advance position to the next character in input.

  4. Return result.

The step skip white space means that the user agent must collect a sequence of characters that are space characters. The collected characters are not used.

When a user agent is to strip line breaks from a string, the user agent must remove any U+000A LINE FEED (LF) and U+000D CARRIAGE RETURN (CR) characters from that string.

When a user agent is to strip leading and trailing white space from a string, the user agent must remove all space characters that are at the start or end of the string.

When a user agent is to strip and collapse white space in a string, it must replace any sequence of one or more consecutive space characters in that string with a single U+0020 SPACE character, and then strip leading and trailing white space from that string.

When a user agent has to strictly split a string on a particular delimiter character delimiter, it must use the following algorithm:

  1. Let input be the string being parsed.

  2. Let position be a pointer into input, initially pointing at the start of the string.

  3. Let tokens be an ordered list of tokens, initially empty.

  4. While position is not past the end of input:

    1. Collect a sequence of characters that are not the delimiter character.

    2. Append the string collected in the previous step to tokens.

    3. Advance position to the next character in input.

  5. Return tokens.

For the special cases of splitting a string on spaces and on commas, this algorithm does not apply (those algorithms also perform white space trimming).

2.4.2. Boolean attributes

A number of attributes are boolean attributes. The presence of a boolean attribute on an element represents the true value, and the absence of the attribute represents the false value.

If the attribute is present, its value must either be the empty string or a value that is an ASCII case-insensitive match for the attribute’s canonical name, with no leading or trailing white space.

A boolean attribute without a value assigned to it (e.g. checked) is implicitly equivalent to one that has the empty string assigned to it (i.e. checked=""). As a consequence, it represents the true value.

The values "true" and "false" are not allowed on boolean attributes. To represent a false value, the attribute has to be omitted altogether.

Here is an example of a checkbox that is checked and disabled. The checked and disabled attributes are the boolean attributes.
<label><input type="checkbox" checked name="cheese" disabled> Cheese</label>

This could be equivalently written as this:

<label><input type="checkbox" checked="checked" name="cheese" disabled="disabled"> Cheese</label>

You can also mix styles:

<label><input type='checkbox' checked name=cheese disabled=""> Cheese</label>

2.4.3. Keywords and enumerated attributes

Some attributes are defined as taking one of a finite set of keywords. Such attributes are called enumerated attributes. The keywords are each defined to map to a particular state (several keywords might map to the same state, in which case some of the keywords are synonyms of each other; additionally, some of the keywords can be said to be non-conforming, and are only in the specification for historical reasons). In addition, two default states can be given. The first is the invalid value default, the second is the missing value default.

If an enumerated attribute is specified, the attribute’s value must be an ASCII case-insensitive match for one of the given keywords that are not said to be non-conforming, with no leading or trailing white space.

When the attribute is specified, if its value is an ASCII case-insensitive match for one of the given keywords then that keyword’s state is the state that the attribute represents. If the attribute value matches none of the given keywords, but the attribute has an invalid value default, then the attribute represents that state. Otherwise, if the attribute value matches none of the keywords but there is a missing value default state defined, then that is the state represented by the attribute. Otherwise, there is no default, and invalid values mean that there is no state represented.

When the attribute is not specified, if there is a missing value default state defined, then that is the state represented by the (missing) attribute. Otherwise, the absence of the attribute means that there is no state represented.

The empty string can be a valid keyword.

2.4.4. Numbers

2.4.4.1. Signed integers

A string is a valid integer if it consists of one or more ASCII digits, optionally prefixed with a U+002D HYPHEN-MINUS character (-).

A valid integer without a U+002D HYPHEN-MINUS (-) prefix represents the number that is represented in base ten by that string of digits. A valid integer with a U+002D HYPHEN-MINUS (-) prefix represents the number represented in base ten by the string of digits that follows the U+002D HYPHEN-MINUS, subtracted from zero.

The rules for parsing integers are as given in the following algorithm. When invoked, the steps must be followed in the order given, aborting at the first step that returns a value. This algorithm will return either an integer or an error.

  1. Let input be the string being parsed.

  2. Let position be a pointer into input, initially pointing at the start of the string.

  3. Let sign have the value "positive".

  4. Skip white space.

  5. If position is past the end of input, return an error.

  6. If the character indicated by position (the first character) is a U+002D HYPHEN-MINUS character (-):

    1. Let sign be "negative".

    2. Advance position to the next character.

    3. If position is past the end of input, return an error.

    Otherwise, if the character indicated by position (the first character) is a U+002B PLUS SIGN character (+):

    1. Advance position to the next character. (The "+" is ignored, but it is not conforming.)

    2. If position is past the end of input, return an error.

  7. If the character indicated by position is not an ASCII digit, then return an error.

  8. Collect a sequence of characters that are ASCII digits, and interpret the resulting sequence as a base-ten integer. Let value be that integer.

  9. If sign is "positive", return value, otherwise return the result of subtracting value from zero.

2.4.4.2. Non-negative integers

A string is a valid non-negative integer if it consists of one or more ASCII digits.

A valid non-negative integer represents the number that is represented in base ten by that string of digits.

The rules for parsing non-negative integers are as given in the following algorithm. When invoked, the steps must be followed in the order given, aborting at the first step that returns a value. This algorithm will return either zero, a positive integer, or an error.

  1. Let input be the string being parsed.

  2. Let value be the result of parsing input using the rules for parsing integers.

  3. If value is an error, return an error.

  4. If value is less than zero, return an error.

  5. Return value.

2.4.4.3. Floating-point numbers

A string is a valid floating-point number if it consists of:

  1. Optionally, a U+002D HYPHEN-MINUS character (-).

  2. One or both of the following, in the given order:

    1. A series of one or more ASCII digits.

    2. Both of the following, in the given order:

      1. A single U+002E FULL STOP character (.).

      2. A series of one or more ASCII digits.

  3. Optionally:

    1. Either a U+0065 LATIN SMALL LETTER E character (e) or a U+0045 LATIN CAPITAL LETTER E character (E).

    2. Optionally, a U+002D HYPHEN-MINUS character (-) or U+002B PLUS SIGN character (+).

    3. A series of one or more ASCII digits.

A valid floating-point number represents the number obtained by multiplying the significand by ten raised to the power of the exponent, where the significand is the first number, interpreted as base ten (including the decimal point and the number after the decimal point, if any, and interpreting the significand as a negative number if the whole string starts with a U+002D HYPHEN-MINUS character (-) and the number is not zero), and where the exponent is the number after the E, if any (interpreted as a negative number if there is a U+002D HYPHEN-MINUS character (-) between the E and the number and the number is not zero, or else ignoring a U+002B PLUS SIGN character (+) between the E and the number if there is one). If there is no E, then the exponent is treated as zero.

The Infinity and Not-a-Number (NaN) values are not valid floating-point numbers.

The best representation of the number n as a floating-point number is the string obtained from running ToString(n). The abstract operation ToString is not uniquely determined. When there are multiple possible strings that could be obtained from ToString for a particular value, the user agent must always return the same string for that value (though it may differ from the value used by other user agents).

The rules for parsing floating-point number values are as given in the following algorithm. This algorithm must be aborted at the first step that returns something. This algorithm will return either a number or an error.

  1. Let input be the string being parsed.

  2. Let position be a pointer into input, initially pointing at the start of the string.

  3. Let value have the value 1.

  4. Let divisor have the value 1.

  5. Let exponent have the value 1.

  6. Skip white space.

  7. If position is past the end of input, return an error.

  8. If the character indicated by position is a U+002D HYPHEN-MINUS character (-):

    1. Change value and divisor to -1.

    2. Advance position to the next character.

    3. If position is past the end of input, return an error.

    Otherwise, if the character indicated by position (the first character) is a U+002B PLUS SIGN character (+):

    1. Advance position to the next character. (The "+" is ignored, but it is not conforming.)

    2. If position is past the end of input, return an error.

  9. If the character indicated by position is a U+002E FULL STOP (.), and that is not the last character in input, and the character after the character indicated by position is an ASCII digit, then set value to zero and jump to the step labeled fraction.

  10. If the character indicated by position is not an ASCII digit, then return an error.

  11. Collect a sequence of characters that are ASCII digits, and interpret the resulting sequence as a base-ten integer. Multiply value by that integer.

  12. If position is past the end of input, jump to the step labeled conversion.

  13. Fraction: If the character indicated by position is a U+002E FULL STOP (.), run these substeps:

    1. Advance position to the next character.

    2. If position is past the end of input, or if the character indicated by position is not an ASCII digit, U+0065 LATIN SMALL LETTER E (e), or U+0045 LATIN CAPITAL LETTER E (E), then jump to the step labeled conversion.

    3. If the character indicated by position is a U+0065 LATIN SMALL LETTER E character (e) or a U+0045 LATIN CAPITAL LETTER E character (E), skip the remainder of these substeps.

    4. Fraction loop: Multiply divisor by ten.

    5. Add the value of the character indicated by position, interpreted as a base-ten digit (0..9) and divided by divisor, to value.

    6. Advance position to the next character.

    7. If position is past the end of input, then jump to the step labeled conversion.

    8. If the character indicated by position is an ASCII digit, jump back to the step labeled fraction loop in these substeps.

  14. If the character indicated by position is a U+0065 LATIN SMALL LETTER E character (e) or a U+0045 LATIN CAPITAL LETTER E character (E), run these substeps:

    1. Advance position to the next character.

    2. If position is past the end of input, then jump to the step labeled conversion.

    3. If the character indicated by position is a U+002D HYPHEN-MINUS character (-):

      1. Change exponent to -1.

      2. Advance position to the next character.

      3. If position is past the end of input, then jump to the step labeled conversion.

      Otherwise, if the character indicated by position is a U+002B PLUS SIGN character (+):

      1. Advance position to the next character.

      2. If position is past the end of input, then jump to the step labeled conversion.

    4. If the character indicated by position is not an ASCII digit, then jump to the step labeled conversion.

    5. Collect a sequence of characters that are ASCII digits, and interpret the resulting sequence as a base-ten integer. Multiply exponent by that integer.

    6. Multiply value by ten raised to the exponentth power.

  15. Conversion: Let S be the set of finite IEEE 754 double-precision floating-point values except -0, but with two special values added: 21024 and -21024.

  16. Let rounded-value be the number in S that is closest to value, selecting the number with an even significand if there are two equally close values. (The two special values 21024 and -21024 are considered to have even significands for this purpose.)

  17. If rounded-value is 21024 or -21024, return an error.

  18. Return rounded-value.

2.4.4.4. Percentages and lengths

The rules for parsing dimension values are as given in the following algorithm. When invoked, the steps must be followed in the order given, aborting at the first step that returns a value. This algorithm will return either a number greater than or equal to 0.0, or an error; if a number is returned, then it is further categorized as either a percentage or a length.

  1. Let input be the string being parsed.

  2. Let position be a pointer into input, initially pointing at the start of the string.

  3. Skip white space.

  4. If position is past the end of input, return an error.

  5. If the character indicated by position is a U+002B PLUS SIGN character (+), advance position to the next character.

  6. If position is past the end of input, return an error.

  7. If the character indicated by position is not an ASCII digit, then return an error.

  8. Collect a sequence of characters that are ASCII digits, and interpret the resulting sequence as a base-ten integer. Let value be that number.

  9. If position is past the end of input, return value as a length.

  10. If the character indicated by position is a U+002E FULL STOP character (.):

    1. Advance position to the next character.

    2. If position is past the end of input, or if the character indicated by position is not an ASCII digit, then return value as a length.

    3. Let divisor have the value 1.

    4. Fraction loop: Multiply divisor by ten.

    5. Add the value of the character indicated by position, interpreted as a base-ten digit (0..9) and divided by divisor, to value.

    6. Advance position to the next character.

    7. If position is past the end of input, then return value as a length.

    8. If the character indicated by position is an ASCII digit, return to the step labeled fraction loop in these substeps.

  11. If position is past the end of input, return value as a length.

  12. If the character indicated by position is a U+0025 PERCENT SIGN character (%), return value as a percentage.

  13. Return value as a length.

2.4.4.5. Non-zero percentages and lengths

The rules for parsing non-zero dimension values are as given in the following algorithm. When invoked, the steps must be followed in the order given, aborting at the first step that returns a value. This algorithm will return either a number greater than 0.0, or an error; if a number is returned, then it is further categorized as either a percentage or a length.

  1. Let input be the string being parsed.

  2. Let value be the result of parsing input using the rules for parsing dimension values.

  3. If value is an error, return an error.

  4. If value is zero, return an error.

  5. If value is a percentage, return value as a percentage.

  6. Return value as a length.

2.4.4.6. Lists of floating-point numbers

A valid list of floating-point numbers is a number of valid floating-point numbers separated by U+002C COMMA characters, with no other characters (e.g. no space characters). In addition, there might be restrictions on the number of floating-point numbers that can be given, or on the range of values allowed.

The rules for parsing a list of floating-point numbers are as follows:

  1. Let input be the string being parsed.

  2. Let position be a pointer into input, initially pointing at the start of the string.

  3. Let numbers be an initially empty list of floating-point numbers. This list will be the result of this algorithm.

  4. Collect a sequence of characters that are space characters, U+002C COMMA, or U+003B SEMICOLON characters. This skips past any leading delimiters.

  5. While position is not past the end of input:

    1. Collect a sequence of characters that are not space characters, U+002C COMMA, U+003B SEMICOLON, ASCII digits, U+002E FULL STOP, or U+002D HYPHEN-MINUS characters. This skips past leading garbage.

    2. Collect a sequence of characters that are not space characters, U+002C COMMA, or U+003B SEMICOLON characters, and let unparsed number be the result.

    3. Let number be the result of parsing unparsed number using the rules for parsing floating-point number values.

    4. If number is an error, set number to zero.

    5. Append number to numbers.

    6. Collect a sequence of characters that are space characters, U+002C COMMA, or U+003B SEMICOLON characters. This skips past the delimiter.

  6. Return numbers.

2.4.4.7. Lists of dimensions

The rules for parsing a list of dimensions are as follows. These rules return a list of zero or more pairs consisting of a number and a unit, the unit being one of percentage, relative, and absolute.

  1. Let raw input be the string being parsed.

  2. If the last character in raw input is a U+002C COMMA character (,), then remove that character from raw input.

  3. Split the string raw input on commas. Let raw tokens be the resulting list of tokens.

  4. Let result be an empty list of number/unit pairs.

  5. For each token in raw tokens, run the following substeps:

    1. Let input be the token.

    2. Let position be a pointer into input, initially pointing at the start of the string.

    3. Let value be the number 0.

    4. Let unit be absolute.

    5. If position is past the end of input, set unit to relative and jump to the last substep.

    6. If the character at position is an ASCII digit, collect a sequence of characters that are ASCII digits, interpret the resulting sequence as an integer in base ten, and increment value by that integer.

    7. If the character at position is a U+002E FULL STOP character (.), run these substeps:

      1. Collect a sequence of characters consisting of space characters and ASCII digits. Let s be the resulting sequence.

      2. Remove all space characters in s.

      3. If s is not the empty string, run these subsubsteps:

        1. Let length be the number of characters in s (after the spaces were removed).

        2. Let fraction be the result of interpreting s as a base-ten integer, and then dividing that number by 10length.

        3. Increment value by fraction.

    8. Skip white space.

    9. If the character at position is a U+0025 PERCENT SIGN character (%), then set unit to percentage.

      Otherwise, if the character at position is a U+002A ASTERISK character (*), then set unit to relative.

    10. Add an entry to result consisting of the number given by value and the unit given by unit.

  6. Return the list result.

2.4.5. Dates and times

This specification encodes dates and times according to a common subset of the [ISO8601] standard for dates.

This means that encoded dates will look like 1582-03-01, 0033-03-27, or 2016-03-01, and date-times will look like 1929-11-13T19:00Z, 0325-06-03T00:21+10:30. The format is approximately YYYY-MM-DDTHH:MM:SS.DD±HH:MM, although some parts are optional, for example to express a month and day as in a birthday, a time without time-zone information, and the like.

Times are expressed using the 24-hour clock, and it is an error to express leap seconds.

Dates are expressed in the proleptic Gregorian calendar between the proleptic year 0001, and the year 9999 inclusive. Other years cannot be encoded.

The proleptic Gregorian calendar is the calendar most common globally since around 1950, and is likely to be understood by almost everyone for dates between the years 1950 and 9999, and for many people for dates in the last few decades or centuries.

The Gregorian calendar was adopted officially in different countries at different times, between the years 1582 when it was proposed by Pope Gregory XIII as a replacement for the Julian calendar, and 1949 when it was adopted by the People’s republic of China.

For most practical purposes, dealing with the present, recent past, or the next few thousand years, this will work without problems. For dates before the adoption of the Gregorian Calendar - for example prior to 1917 in Russia or Turkey, prior to 1752 in Britain or the then British colonies of America, or prior to 1582 in Spain, the Spanish colonies in America, and the rest of the world, dates will not match those written at the time.

The use of the Gregorian calendar as an underlying encoding is a somewhat arbitrary choice. Many other calendars were or are in use, and the interested reader should look for information on the Web.

See also the discussion of date, time, and number formats in forms (for authors), implementation notes regarding localization of form controls, and the time element.

In the algorithms below, the number of days in month month of year year is: 31 if month is 1, 3, 5, 7, 8, 10, or 12; 30 if month is 4, 6, 9, or 11; 29 if month is 2 and year is a number divisible by 400, or if year is a number divisible by 4 but not by 100; and 28 otherwise. This takes into account leap years in the Gregorian calendar. [GREGORIAN]

When ASCII digits are used in the date and time syntaxes defined in this section, they express numbers in base ten.

While the formats described here are intended to be subsets of the corresponding ISO8601 formats, this specification defines parsing rules in much more detail than ISO8601. Implementors are therefore encouraged to carefully examine any date parsing libraries before using them to implement the parsing rules described below; ISO8601 libraries might not parse dates and times in exactly the same manner. [ISO8601]

Where this specification refers to the proleptic Gregorian calendar, it means the modern Gregorian calendar, extrapolated backwards to year 0001. A date in the proleptic Gregorian calendar, sometimes explicitly referred to as a proleptic-Gregorian date, is one that is described using that calendar even if that calendar was not in use at the time (or place) in question. [GREGORIAN]

2.4.5.1. Months

A month consists of a specific proleptic-Gregorian date with no time-zone information and no date information beyond a year and a month. [GREGORIAN]

A string is a valid month string representing a year year and month month if it consists of the following components in the given order:

  1. Four ASCII digits, representing year, where year >= 1

  2. A U+002D HYPHEN-MINUS character (-)

  3. Two ASCII digits, representing the month month, in the range 1 ≤ month ≤ 12

For example, February 2005 is encoded 2005-02, and March of the year 33AD (as a proleptic gregorian date) is encoded 0033-03. The expression 325-03 does not mean March in the year 325, it is an error, because it does not have 4 digits for the year.

The rules to parse a month string are as follows. This will return either a year and month, or nothing. If at any point the algorithm says that it "fails", this means that it is aborted at that point and returns nothing.

  1. Let input be the string being parsed.

  2. Let position be a pointer into input, initially pointing at the start of the string.

  3. Parse a month component to obtain year and month. If this returns nothing, then fail.

  4. If position is not beyond the end of input, then fail.

  5. Return year and month.

The rules to parse a month component, given an input string and a position, are as follows. This will return either a year and a month, or nothing. If at any point the algorithm says that it "fails", this means that it is aborted at that point and returns nothing.

  1. Collect a sequence of characters that are ASCII digits. If the collected sequence is not at least four characters long, then fail. Otherwise, interpret the resulting sequence as a base-ten integer. Let that number be the year.

  2. If year is not a number greater than zero, then fail.

  3. If position is beyond the end of input or if the character at position is not a U+002D HYPHEN-MINUS character, then fail. Otherwise, move position forwards one character.

  4. Collect a sequence of characters that are ASCII digits. If the collected sequence is not exactly two characters long, then fail. Otherwise, interpret the resulting sequence as a base-ten integer. Let that number be the month.

  5. If month is not a number in the range 1 ≤ month ≤ 12, then fail.

  6. Return year and month.

2.4.5.2. Dates

A date consists of a specific proleptic-Gregorian date with no time-zone information, consisting of a year, a month, and a day. [GREGORIAN]

A string is a valid date string representing a year year, month month, and day day if it consists of the following components in the given order:

  1. A valid month string, representing year and month

  2. A U+002D HYPHEN-MINUS character (-)

  3. Two ASCII digits, representing day, in the range 1 ≤ day ≤ maxday where maxday is the number of days in the month month and year year

For example, 29 February 2016 is encoded 2016-02-29, and 3 March of the year 33AD (as a proleptic gregorian date) is encoded 0033-03-03. The expression 325-03-03 does not mean 3 March in the year 325, it is an error, because it does not have 4 digits for the year.

The rules to parse a date string are as follows. This will return either a date, or nothing. If at any point the algorithm says that it "fails", this means that it is aborted at that point and returns nothing.

  1. Let input be the string being parsed.

  2. Let position be a pointer into input, initially pointing at the start of the string.

  3. Parse a date component to obtain year, month, and day. If this returns nothing, then fail.

  4. If position is not beyond the end of input, then fail.

  5. Let date be the date with year year, month month, and day day.

  6. Return date.

The rules to parse a date component, given an input string and a position, are as follows. This will return either a year, a month, and a day, or nothing. If at any point the algorithm says that it "fails", this means that it is aborted at that point and returns nothing.

  1. Parse a month component to obtain year and month. If this returns nothing, then fail.

  2. Let maxday be the number of days in month month of year year.

  3. If position is beyond the end of input or if the character at position is not a U+002D HYPHEN-MINUS character, then fail. Otherwise, move position forwards one character.

  4. Collect a sequence of characters that are ASCII digits. If the collected sequence is not exactly two characters long, then fail. Otherwise, interpret the resulting sequence as a base-ten integer. Let that number be the day.

  5. If day is not a number in the range 1 ≤ day ≤ maxday, then fail.

  6. Return year, month, and day.

2.4.5.3. Yearless dates

A yearless date consists of a Gregorian month and a day within that month, but with no associated year. [GREGORIAN]

A string is a valid yearless date string representing a month month and a day day if it consists of the following components in the given order:

  1. Optionally, two U+002D HYPHEN-MINUS characters (-)

  2. Two ASCII digits, representing the month month, in the range 1 ≤ month ≤ 12

  3. A U+002D HYPHEN-MINUS character (-)

  4. Two ASCII digits, representing day, in the range 1 ≤ day ≤ maxday where maxday is the number of days in the month month and any arbitrary leap year (e.g., 4 or 2000)

In other words, if the month is "02", meaning February, then the day can be 29, as if the year was a leap year.

For example, 29 February is encoded 02-29, and 3 March is encoded 03-03.

The rules to parse a yearless date string are as follows. This will return either a month and a day, or nothing. If at any point the algorithm says that it "fails", this means that it is aborted at that point and returns nothing.

  1. Let input be the string being parsed.

  2. Let position be a pointer into input, initially pointing at the start of the string.

  3. Parse a yearless date component to obtain month and day. If this returns nothing, then fail.

  4. If position is not beyond the end of input, then fail.

  5. Return month and day.

The rules to parse a yearless date component, given an input string and a position, are as follows. This will return either a month and a day, or nothing. If at any point the algorithm says that it "fails", this means that it is aborted at that point and returns nothing.

  1. Collect a sequence of characters that are U+002D HYPHEN-MINUS characters (-). If the collected sequence is not exactly zero or two characters long, then fail.

  2. Collect a sequence of characters that are ASCII digits. If the collected sequence is not exactly two characters long, then fail. Otherwise, interpret the resulting sequence as a base-ten integer. Let that number be the month.

  3. If month is not a number in the range 1 ≤ month ≤ 12, then fail.

  4. Let maxday be the number of days in month month of any arbitrary leap year (e.g., 4 or 2000).

  5. If position is beyond the end of input or if the character at position is not a U+002D HYPHEN-MINUS character, then fail. Otherwise, move position forwards one character.

  6. Collect a sequence of characters that are ASCII digits. If the collected sequence is not exactly two characters long, then fail. Otherwise, interpret the resulting sequence as a base-ten integer. Let that number be the day.

  7. If day is not a number in the range 1 ≤ day ≤ maxday, then fail.

  8. Return month and day.

2.4.5.4. Times

A time consists of a specific time with no time-zone information, consisting of an hour, a minute, a second, and a fraction of a second.

A string is a valid time string representing an hour hour, a minute minute, and a second second if it consists of the following components in the given order:

  1. Two ASCII digits, representing hour, in the range 0 ≤ hour ≤ 23

  2. A U+003A COLON character (:)

  3. Two ASCII digits, representing minute, in the range 0 ≤ minute ≤ 59

  4. If second is non-zero, or optionally if second is zero:

    1. A U+003A COLON character (:)

    2. Two ASCII digits, representing the integer part of second, in the range 0 ≤ s ≤ 59

    3. If second is not an integer, or optionally if second is an integer:

      1. A 002E FULL STOP character (.)

      2. One, two, or three ASCII digits, representing the fractional part of second

The second component cannot be 60 or 61; leap seconds cannot be represented.

Times are encoded using the 24 hour clock, with optional seconds, and optional decimal fractions of seconds. Thus 7.45pm is encoded as 19:45. Note that parsing that time will return 19:45:00, or 7.45pm and zero seconds. 19:45:45.456 is 456 thousandths of a second after 7.45pm and 45 seconds.

The rules to parse a time string are as follows. This will return either a time, or nothing. If at any point the algorithm says that it "fails", this means that it is aborted at that point and returns nothing.

  1. Let input be the string being parsed.

  2. Let position be a pointer into input, initially pointing at the start of the string.

  3. Parse a time component to obtain hour, minute, and second. If this returns nothing, then fail.

  4. If position is not beyond the end of input, then fail.

  5. Let time be the time with hour hour, minute minute, and second second.

  6. Return time.

The rules to parse a time component, given an input string and a position, are as follows. This will return either an hour, a minute, and a second, or nothing. If at any point the algorithm says that it "fails", this means that it is aborted at that point and returns nothing.

  1. Collect a sequence of characters that are ASCII digits. If the collected sequence is not exactly two characters long, then fail. Otherwise, interpret the resulting sequence as a base-ten integer. Let that number be the hour.

  2. If hour is not a number in the range 0 ≤ hour ≤ 23, then fail.

  3. If position is beyond the end of input or if the character at position is not a U+003A COLON character, then fail. Otherwise, move position forwards one character.

  4. Collect a sequence of characters that are ASCII digits. If the collected sequence is not exactly two characters long, then fail. Otherwise, interpret the resulting sequence as a base-ten integer. Let that number be the minute.

  5. If minute is not a number in the range 0 ≤ minute ≤ 59, then fail.

  6. Let second be a string with the value "0".

  7. If position is not beyond the end of input and the character at position is a U+003A COLON, then run these substeps:

    1. Advance position to the next character in input.

    2. If position is beyond the end of input, or at the last character in input, or if the next two characters in input starting at position are not both ASCII digits, then fail.

    3. Collect a sequence of characters that are either ASCII digits or U+002E FULL STOP characters. If the collected sequence is three characters long, or if it is longer than three characters long and the third character is not a U+002E FULL STOP character, or if it has more than one U+002E FULL STOP character, then fail. Otherwise, let second be the collected string.

  8. Interpret second as a base-ten number (possibly with a fractional part). Let second be that number instead of the string version.

  9. If second is not a number in the range 0 ≤ second < 60, then fail.

  10. Return hour, minute, and second.

2.4.5.5. Floating dates and times

A floating date and time consists of a specific proleptic-Gregorian date, consisting of a year, a month, and a day, and a time, consisting of an hour, a minute, a second, and a fraction of a second, but expressed without a time zone. [GREGORIAN]

A string is a valid floating date and time string representing a date and time if it consists of the following components in the given order:

  1. A valid date string representing the date

  2. A U+0054 LATIN CAPITAL LETTER T character (T) or a U+0020 SPACE character

  3. A valid time string representing the time

A string is a valid normalized floating date and time string representing a date and time if it consists of the following components in the given order:

  1. A valid date string representing the date

  2. A U+0054 LATIN CAPITAL LETTER T character (T)

  3. A valid time string representing the time, expressed as the shortest possible string for the given time (e.g., omitting the seconds component entirely if the given time is zero seconds past the minute)

The rules to parse a floating date and time string are as follows. This will return either a date and time, or nothing. If at any point the algorithm says that it "fails", this means that it is aborted at that point and returns nothing.

  1. Let input be the string being parsed.

  2. Let position be a pointer into input, initially pointing at the start of the string.

  3. Parse a date component to obtain year, month, and day. If this returns nothing, then fail.

  4. If position is beyond the end of input or if the character at position is neither a U+0054 LATIN CAPITAL LETTER T character (T) nor a U+0020 SPACE character, then fail. Otherwise, move position forwards one character.

  5. Parse a time component to obtain hour, minute, and second. If this returns nothing, then fail.

  6. If position is not beyond the end of input, then fail.

  7. Let date be the date with year year, month month, and day day.

  8. Let time be the time with hour hour, minute minute, and second second.

  9. Return date and time.

2.4.5.6. Time zones

A time-zone offset consists of a signed number of hours and minutes.

A string is a valid time-zone offset string representing a time-zone offset if it consists of either:

This format allows for time-zone offsets from -23:59 to +23:59. In practice, however, right now the range of offsets of actual time zones is -12:00 to +14:00, and the minutes component of offsets of actual time zones is always either 00, 30, or 45. There is no guarantee that this will remain so forever, however; time zones are changed by countries at will and do not follow a standard.

See also the usage notes and examples in the global date and time section below for details on using time-zone offsets with historical times that predate the formation of formal time zones.

The rules to parse a time-zone offset string are as follows. This will return either a time-zone offset, or nothing. If at any point the algorithm says that it "fails", this means that it is aborted at that point and returns nothing.

  1. Let input be the string being parsed.

  2. Let position be a pointer into input, initially pointing at the start of the string.

  3. Parse a time-zone offset component to obtain timezonehours and timezoneminutes. If this returns nothing, then fail.

  4. If position is not beyond the end of input, then fail.

  5. Return the time-zone offset that is timezonehours hours and timezoneminutes minutes from UTC.

The rules to parse a time-zone offset component, given an input string and a position, are as follows. This will return either time-zone hours and time-zone minutes, or nothing. If at any point the algorithm says that it "fails", this means that it is aborted at that point and returns nothing.

  1. If the character at position is a U+005A LATIN CAPITAL LETTER Z character (Z), then:

    1. Let timezonehours be 0.

    2. Let timezoneminutes be 0.

    3. Advance position to the next character in input.

    Otherwise, if the character at position is either a U+002B PLUS SIGN (+) or a U+002D HYPHEN-MINUS (-), then:

    1. If the character at position is a U+002B PLUS SIGN (+), let sign be "positive". Otherwise, it’s a U+002D HYPHEN-MINUS (-); let sign be "negative".

    2. Advance position to the next character in input.

    3. Collect a sequence of characters that are ASCII digits. Let s be the collected sequence.

    4. If s is exactly two characters long, then run these substeps:

      1. Interpret s as a base-ten integer. Let that number be the timezonehours.

      2. If position is beyond the end of input or if the character at position is not a U+003A COLON character, then fail. Otherwise, move position forwards one character.

      3. Collect a sequence of characters that are ASCII digits. If the collected sequence is not exactly two characters long, then fail. Otherwise, interpret the resulting sequence as a base-ten integer. Let that number be the timezoneminutes.

      If s is exactly four characters long, then run these substeps:

      1. Interpret the first two characters of s as a base-ten integer. Let that number be the timezonehours.

      2. Interpret the last two characters of s as a base-ten integer. Let that number be the timezoneminutes.

      Otherwise, fail.

    5. If timezonehours is not a number in the range 0 ≤ timezonehours ≤ 23, then fail.

    6. If sign is "negative", then negate timezonehours.

    7. If timezoneminutes is not a number in the range 0 ≤ timezoneminutes ≤ 59, then fail.

    8. If sign is "negative", then negate timezoneminutes.

    Otherwise, fail.

  2. Return timezonehours and timezoneminutes.

2.4.5.7. Global dates and times

A global date and time consists of a specific proleptic-Gregorian date, consisting of a year, a month, and a day, and a time, consisting of an hour, a minute, a second, and a fraction of a second, expressed with a time-zone offset, consisting of a signed number of hours and minutes. [GREGORIAN]

A string is a valid global date and time string representing a date, time, and a time-zone offset if it consists of the following components in the given order:

  1. A valid date string representing the date

  2. A U+0054 LATIN CAPITAL LETTER T character (T) or a U+0020 SPACE character

  3. A valid time string representing the time

  4. A valid time-zone offset string representing the time-zone offset

Times in dates before the formation of UTC in the mid twentieth century must be expressed and interpreted in terms of UT1 (contemporary Earth mean solar time at the 0° longitude), not UTC (the approximation of UT1 that ticks in SI seconds). Time before the formation of time zones must be expressed and interpreted as UT1 times with explicit time zones that approximate the contemporary difference between the appropriate local time and the time observed at the location of Greenwich, London.

The following are some examples of dates written as valid global date and time strings.
"0037-12-13 00:00Z"

Midnight "London time" (UTC) on the birthday of the Roman Emperor Nero. See below for further discussion on which date this actually corresponds to.

"1979-10-14T12:00:00.001-04:00"

One millisecond after noon on October 14th 1979, in the time zone in use on the east coast of the USA during daylight saving time.

"8592-01-01T02:09+02:09"

Midnight UTC on the 1st of January, 8592. The time zone associated with that time is two hours and nine minutes ahead of UTC, which is not currently a real time zone, but is nonetheless allowed.

Several things are notable about these dates:

  • Years with fewer than four digits have to be zero-padded. The date "37-12-13" is not a valid date.

  • If the "T" is replaced by a space, it must be a single space character. The string "2001-12-21  12:00Z" (with two spaces between the components) would not be parsed successfully.

  • To unambiguously identify a date it has to be first converted to the Gregorian calendar (e.g., from the Hijri, Jewish, Julian or other calendar). For example, the Roman Emperor Nero was born on the 15th of December 37 in the Julian Calendar, which is the 13th of December 37 in the proleptic Gregorian calendar.

  • The time and time-zone offset components are not optional.

  • Dates before the year one or after the year 9999 in the Gregorian calendar cannot be represented as a datetime in this version of HTML.

  • Time-zone offsets for a place may vary, for example due to daylight savings time.

The zone offset is not a complete time zone specification. When working with real date and time values, consider using a separate field for time zone, perhaps using IANA time zone IDs. [TIMEZONE]

A string is a valid normalized global date and time string representing a date, time, and a time-zone offset if it consists of the following components in the given order:

  1. A valid date string representing the date converted to the UTC time zone

  2. A U+0054 LATIN CAPITAL LETTER T character (T)

  3. A valid time string representing the time converted to the UTC time zone and expressed as the shortest possible string for the given time (e.g., omitting the seconds component entirely if the given time is zero seconds past the minute)

  4. A U+005A LATIN CAPITAL LETTER Z character (Z)

The rules to parse a global date and time string are as follows. This will return either a time in UTC, with associated time-zone offset information for round-tripping or display purposes, or nothing. If at any point the algorithm says that it "fails", this means that it is aborted at that point and returns nothing.

  1. Let input be the string being parsed.

  2. Let position be a pointer into input, initially pointing at the start of the string.

  3. Parse a date component to obtain year, month, and day. If this returns nothing, then fail.

  4. If position is beyond the end of input or if the character at position is neither a U+0054 LATIN CAPITAL LETTER T character (T) nor a U+0020 SPACE character, then fail. Otherwise, move position forwards one character.

  5. Parse a time component to obtain hour, minute, and second. If this returns nothing, then fail.

  6. If position is beyond the end of input, then fail.

  7. Parse a time-zone offset component to obtain timezonehours and timezoneminutes. If this returns nothing, then fail.

  8. If position is not beyond the end of input, then fail.

  9. Let time be the moment in time at year year, month month, day day, hours hour, minute minute, second second, subtracting timezonehours hours and timezoneminutes minutes. That moment in time is a moment in the UTC time zone.

  10. Let timezone be timezonehours hours and timezoneminutes minutes from UTC.

  11. Return time and timezone.

2.4.5.8. Weeks

A week consists of a week-year number and a week number representing a seven-day period starting on a Monday. Each week-year in this calendaring system has either 52 or 53 such seven-day periods, as defined below. The seven-day period starting on the Gregorian date Monday December 29th 1969 (1969-12-29) is defined as week number 1 in week-year 1970. Consecutive weeks are numbered sequentially. The week before the number 1 week in a week-year is the last week in the previous week-year, and vice versa. [GREGORIAN]

A week-year with a number year has 53 weeks if it corresponds to either a year year in the proleptic Gregorian calendar that has a Thursday as its first day (January 1st), or a year year in the proleptic Gregorian calendar that has a Wednesday as its first day (January 1st) and where year is a number divisible by 400, or a number divisible by 4 but not by 100. All other week-years have 52 weeks.

The week number of the last day of a week-year with 53 weeks is 53; the week number of the last day of a week-year with 52 weeks is 52.

The week-year number of a particular day can be different than the number of the year that contains that day in the proleptic Gregorian calendar. The first week in a week-year y is the week that contains the first Thursday of the Gregorian year y.

For modern purposes, a week as defined here is equivalent to ISO weeks as defined in ISO 8601. [ISO8601]

A string is a valid week string representing a week-year year and week week if it consists of the following components in the given order:

  1. Four ASCII digits, representing year, where year >= 1

  2. A U+002D HYPHEN-MINUS character (-)

  3. A U+0057 LATIN CAPITAL LETTER W character (W)

  4. Two ASCII digits, representing the week week, in the range 1 ≤ week ≤ maxweek, where maxweek is the week number of the last day of week-year year

The rules to parse a week string are as follows. This will return either a week-year number and week number, or nothing. If at any point the algorithm says that it "fails", this means that it is aborted at that point and returns nothing.

  1. Let input be the string being parsed.

  2. Let position be a pointer into input, initially pointing at the start of the string.

  3. Collect a sequence of characters that are ASCII digits. If the collected sequence is not at least four characters long, then fail. Otherwise, interpret the resulting sequence as a base-ten integer. Let that number be the year.

  4. If year is not a number greater than zero, then fail.

  5. If position is beyond the end of input or if the character at position is not a U+002D HYPHEN-MINUS character, then fail. Otherwise, move position forwards one character.

  6. If position is beyond the end of input or if the character at position is not a U+0057 LATIN CAPITAL LETTER W character (W), then fail. Otherwise, move position forwards one character.

  7. Collect a sequence of characters that are ASCII digits. If the collected sequence is not exactly two characters long, then fail. Otherwise, interpret the resulting sequence as a base-ten integer. Let that number be the week.

  8. Let maxweek be the week number of the last day of year year.

  9. If week is not a number in the range 1 ≤ week ≤ maxweek, then fail.

  10. If position is not beyond the end of input, then fail.

  11. Return the week-year number year and the week number week.

2.4.5.9. Durations

A duration consists of a number of seconds.

Since months and seconds are not comparable (a month is not a precise number of seconds, but is instead a period whose exact length depends on the precise day from which it is measured) a duration as defined in this specification cannot include months (or years, which are equivalent to twelve months). Only durations that describe a specific number of seconds can be described.

A string is a valid duration string representing a duration t if it consists of either of the following:

The rules to parse a duration string are as follows. This will return either a duration or nothing. If at any point the algorithm says that it "fails", this means that it is aborted at that point and returns nothing.

  1. Let input be the string being parsed.

  2. Let position be a pointer into input, initially pointing at the start of the string.

  3. Let months, seconds, and component count all be zero.

  4. Let M-disambiguator be minutes.

    This flag’s other value is months. It is used to disambiguate the "M" unit in ISO8601 durations, which use the same unit for months and minutes. Months are not allowed, but are parsed for future compatibility and to avoid misinterpreting ISO8601 durations that would be valid in other contexts.

  5. Skip white space.

  6. If position is past the end of input, then fail.

  7. If the character in input pointed to by position is a U+0050 LATIN CAPITAL LETTER P character, then advance position to the next character, set M-disambiguator to months, and skip white space.

  8. Run the following substeps in a loop, until a step requiring the loop to be broken or the entire algorithm to fail is reached:

    1. Let units be undefined. It will be assigned one of the following values: years, months, weeks, days, hours, minutes, and seconds.

    2. Let next character be undefined. It is used to process characters from the input.

    3. If position is past the end of input, then break the loop.

    4. If the character in input pointed to by position is a U+0054 LATIN CAPITAL LETTER T character, then advance position to the next character, set M-disambiguator to minutes, skip white space, and return to the top of the loop.

    5. Set next character to the character in input pointed to by position.

    6. If next character is a U+002E FULL STOP character (.), then let N equal zero. (Do not advance position. That is taken care of below.)

      Otherwise, if next character is an ASCII digit, then collect a sequence of characters that are ASCII digits, interpret the resulting sequence as a base-ten integer, and let N be that number.

      Otherwise next character is not part of a number; fail.

    7. If position is past the end of input, then fail.

    8. Set next character to the character in input pointed to by position, and this time advance position to the next character. (If next character was a U+002E FULL STOP character (.) before, it will still be that character this time.)

    9. If next character is a U+002E FULL STOP character (.), then run these substeps:

      1. Collect a sequence of characters that are ASCII digits. Let s be the resulting sequence.

      2. If s is the empty string, then fail.

      3. Let length be the number of characters in s.

      4. Let fraction be the result of interpreting s as a base-ten integer, and then dividing that number by 10length.

      5. Increment N by fraction.

      6. Skip white space.

      7. If position is past the end of input, then fail.

      8. Set next character to the character in input pointed to by position, and advance position to the next character.

      9. If next character is neither a U+0053 LATIN CAPITAL LETTER S character nor a U+0073 LATIN SMALL LETTER S character, then fail.

      10. Set units to seconds.

      Otherwise, run these substeps:

      1. If next character is a space character, then skip white space, set next character to the character in input pointed to by position, and advance position to the next character.

      2. If next character is a U+0059 LATIN CAPITAL LETTER Y character, or a U+0079 LATIN SMALL LETTER Y character, set units to years and set M-disambiguator to months.

        If next character is a U+004D LATIN CAPITAL LETTER M character or a U+006D LATIN SMALL LETTER M character, and M-disambiguator is months, then set units to months.

        If next character is a U+0057 LATIN CAPITAL LETTER W character or a U+0077 LATIN SMALL LETTER W character, set units to weeks and set M-disambiguator to minutes.

        If next character is a U+0044 LATIN CAPITAL LETTER D character or a U+0064 LATIN SMALL LETTER D character, set units to days and set M-disambiguator to minutes.

        If next character is a U+0048 LATIN CAPITAL LETTER H character or a U+0068 LATIN SMALL LETTER H character, set units to hours and set M-disambiguator to minutes.

        If next character is a U+004D LATIN CAPITAL LETTER M character or a U+006D LATIN SMALL LETTER M character, and M-disambiguator is minutes, then set units to minutes.

        If next character is a U+0053 LATIN CAPITAL LETTER S character or a U+0073 LATIN SMALL LETTER S character, set units to seconds and set M-disambiguator to minutes.

        Otherwise if next character is none of the above characters, then fail.

    10. Increment component count.

    11. Let multiplier be 1.

    12. If units is years, multiply multiplier by 12 and set units to months.

    13. If units is months, add the product of N and multiplier to months.

      Otherwise, run these substeps:

      1. If units is weeks, multiply multiplier by 7 and set units to days.

      2. If units is days, multiply multiplier by 24 and set units to hours.

      3. If units is hours, multiply multiplier by 60 and set units to minutes.

      4. If units is minutes, multiply multiplier by 60 and set units to seconds.

      5. Forcibly, units is now seconds. Add the product of N and multiplier to seconds.

    14. Skip white space.

  9. If component count is zero, fail.

  10. If months is not zero, fail.

  11. Return the duration consisting of seconds seconds.

2.4.5.10. Vaguer moments in time

A string is a valid date string with optional time if it is also one of the following:


The rules to parse a date or time string are as follows. The algorithm will return either a date, a time, a global date and time, or nothing. If at any point the algorithm says that it "fails", this means that it is aborted at that point and returns nothing.

  1. Let input be the string being parsed.

  2. Let position be a pointer into input, initially pointing at the start of the string.

  3. Set start position to the same position as position.

  4. Set the date present and time present flags to true.

  5. Parse a date component to obtain year, month, and day. If this fails, then set the date present flag to false.

  6. If date present is true, and position is not beyond the end of input, and the character at position is either a U+0054 LATIN CAPITAL LETTER T character (T) or a U+0020 SPACE character, then advance position to the next character in input.

    Otherwise, if date present is true, and either position is beyond the end of input or the character at position is neither a U+0054 LATIN CAPITAL LETTER T character (T) nor a U+0020 SPACE character, then set time present to false.

    Otherwise, if date present is false, set position back to the same position as start position.

  7. If the time present flag is true, then parse a time component to obtain hour, minute, and second. If this returns nothing, then fail.

  8. If the date present and time present flags are both true, but position is beyond the end of input, then fail.

  9. If the date present and time present flags are both true, parse a time-zone offset component to obtain timezonehours and timezoneminutes. If this returns nothing, then fail.

  10. If position is not beyond the end of input, then fail.

  11. If the date present flag is true and the time present flag is false, then let date be the date with year year, month month, and day day, and return date.

    Otherwise, if the time present flag is true and the date present flag is false, then let time be the time with hour hour, minute minute, and second second, and return time.

    Otherwise, let time be the moment in time at year year, month month, day day, hours hour, minute minute, second second, subtracting timezonehours hours and timezoneminutes minutes, that moment in time being a moment in the UTC time zone; let timezone be timezonehours hours and timezoneminutes minutes from UTC; and return time and timezone.

2.4.6. Colors

A simple color consists of three 8-bit numbers in the range 0..255, representing the red, green, and blue components of the color respectively, in the sRGB color space. [SRGB]

A string is a valid simple color if it is exactly seven characters long, and the first character is a U+0023 NUMBER SIGN character (#), and the remaining six characters are all ASCII hex digits, with the first two digits representing the red component, the middle two digits representing the green component, and the last two digits representing the blue component, in hexadecimal.

A string is a valid lowercase simple color if it is a valid simple color and doesn’t use any characters in the range U+0041 LATIN CAPITAL LETTER A to U+0046 LATIN CAPITAL LETTER F.

The rules for parsing simple color values are as given in the following algorithm. When invoked, the steps must be followed in the order given, aborting at the first step that returns a value. This algorithm will return either a simple color or an error.

  1. Let input be the string being parsed.

  2. If input is not exactly seven characters long, then return an error.

  3. If the first character in input is not a U+0023 NUMBER SIGN character (#), then return an error.

  4. If the last six characters of input are not all ASCII hex digits, then return an error.

  5. Let result be a simple color.

  6. Interpret the second and third characters as a hexadecimal number and let the result be the red component of result.

  7. Interpret the fourth and fifth characters as a hexadecimal number and let the result be the green component of result.

  8. Interpret the sixth and seventh characters as a hexadecimal number and let the result be the blue component of result.

  9. Return result.

The rules for serializing simple color values given a simple color are as given in the following algorithm:

  1. Let result be a string consisting of a single U+0023 NUMBER SIGN character (#).

  2. Convert the red, green, and blue components in turn to two-digit hexadecimal numbers using lowercase ASCII hex digits, zero-padding if necessary, and append these numbers to result, in the order red, green, blue.

  3. Return result, which will be a valid lowercase simple color.


Some obsolete legacy attributes parse colors in a more complicated manner, using the rules for parsing a legacy color value, which are given in the following algorithm. When invoked, the steps must be followed in the order given, aborting at the first step that returns a value. This algorithm will return either a simple color or an error.

  1. Let input be the string being parsed.

  2. If input is the empty string, then return an error.

  3. Strip leading and trailing white space from input.

  4. If input is an ASCII case-insensitive match for the string "transparent", then return an error.

  5. If input is an ASCII case-insensitive match for one of the named colors, then return the simple color corresponding to that keyword. [CSS3COLOR]

    CSS2 System Colors are not recognized.

  6. If input is four characters long, and the first character in input is a U+0023 NUMBER SIGN character (#), and the last three characters of input are all ASCII hex digits, then run these substeps:

    1. Let result be a simple color.

    2. Interpret the second character of input as a hexadecimal digit; let the red component of result be the resulting number multiplied by 17.

    3. Interpret the third character of input as a hexadecimal digit; let the green component of result be the resulting number multiplied by 17.

    4. Interpret the fourth character of input as a hexadecimal digit; let the blue component of result be the resulting number multiplied by 17.

    5. Return result.

  7. Replace any characters in input that have a Unicode code point greater than U+FFFF (i.e., any characters that are not in the basic multilingual plane) with the two-character string "00".

  8. If input is longer than 128 characters, truncate input, leaving only the first 128 characters.

  9. If the first character in input is a U+0023 NUMBER SIGN character (#), remove it.

  10. Replace any character in input that is not an ASCII hex digit with the character U+0030 DIGIT ZERO (0).

  11. While input’s length is zero or not a multiple of three, append a U+0030 DIGIT ZERO (0) character to input.

  12. Split input into three strings of equal length, to obtain three components. Let length be the length of those components (one third the length of input).

  13. If length is greater than 8, then remove the leading length-8 characters in each component, and let length be 8.

  14. While length is greater than two and the first character in each component is a U+0030 DIGIT ZERO (0) character, remove that character and reduce length by one.

  15. If length is still greater than two, truncate each component, leaving only the first two characters in each.

  16. Let result be a simple color.

  17. Interpret the first component as a hexadecimal number; let the red component of result be the resulting number.

  18. Interpret the second component as a hexadecimal number; let the green component of result be the resulting number.

  19. Interpret the third component as a hexadecimal number; let the blue component of result be the resulting number.

  20. Return result.

2.4.7. Space-separated tokens

A set of space-separated tokens is a string containing zero or more words (known as tokens) separated by one or more space characters, where words consist of any string of one or more characters, none of which are space characters.

A string containing a set of space-separated tokens may have leading or trailing space characters.

An unordered set of unique space-separated tokens is a set of space-separated tokens where none of the tokens are duplicated.

An ordered set of unique space-separated tokens is a set of space-separated tokens where none of the tokens are duplicated but where the order of the tokens is meaningful.

Sets of space-separated tokens sometimes have a defined set of allowed values. When a set of allowed values is defined, the tokens must all be from that list of allowed values; other values are non-conforming. If no such set of allowed values is provided, then all values are conforming.

How tokens in a set of space-separated tokens are to be compared (e.g., case-sensitively or not) is defined on a per-set basis.

When a user agent has to split a string on spaces, it must use the following algorithm:

  1. Let input be the string being parsed.

  2. Let position be a pointer into input, initially pointing at the start of the string.

  3. Let tokens be an ordered list of tokens, initially empty.

  4. Skip white space

  5. While position is not past the end of input:

    1. Collect a sequence of characters that are not space characters.

    2. Append the string collected in the previous step to tokens.

    3. Skip white space

  6. Return tokens.

2.4.8. Comma-separated tokens

A set of comma-separated tokens is a string containing zero or more tokens each separated from the next by a single U+002C COMMA character (,), where tokens consist of any string of zero or more characters, neither beginning nor ending with space characters, nor containing any U+002C COMMA characters (,), and optionally surrounded by space characters.

For instance, the string " a ,b, ,d d " consists of four tokens: "a", "b", the empty string, and "d d". Leading and trailing white space around each token doesn’t count as part of the token, and the empty string can be a token.

Sets of comma-separated tokens sometimes have further restrictions on what consists a valid token. When such restrictions are defined, the tokens must all fit within those restrictions; other values are non-conforming. If no such restrictions are specified, then all values are conforming.

When a user agent has to split a string on commas, it must use the following algorithm:

  1. Let input be the string being parsed.

  2. Let position be a pointer into input, initially pointing at the start of the string.

  3. Let tokens be an ordered list of tokens, initially empty.

  4. Token: If position is past the end of input, jump to the last step.

  5. Collect a sequence of characters that are not U+002C COMMA characters (,). Let s be the resulting sequence (which might be the empty string).

  6. Strip leading and trailing white space from s.

  7. Append s to tokens.

  8. If position is not past the end of input, then the character at position is a U+002C COMMA character (,); advance position past that character.

  9. Jump back to the step labeled token.

  10. Return tokens.

2.4.9. References

A valid hash-name reference to an element of type type is a string consisting of a U+0023 NUMBER SIGN character (#) followed by a string which exactly matches the value of the name attribute of an element with type type in the document.

The rules for parsing a hash-name reference to an element of type type, are as follows:

  1. If the string being parsed does not contain a U+0023 NUMBER SIGN character, or if the first such character in the string is the last character in the string, then return null and abort these steps.

  2. Let s be the string from the character immediately after the first U+0023 NUMBER SIGN character in the string being parsed up to the end of that string.

  3. Return the first element of type type in tree order that has an id attribute or a name attribute whose value is s.

2.4.10. Media queries

A string is a valid media query list if it matches the <media-query-list> production of the Media Queries specification. [MEDIAQ]

A string matches the environment of the user if it is the empty string, a string consisting of only space characters, or is a media query list that matches the user’s environment according to the definitions given in the Media Queries specification. [MEDIAQ]

2.5. URLs

2.5.1. Terminology

A valid URL is a string that conforms to the authoring requirements for a valid url string. [URL]

Typically, this means an absolute URL of the form scheme://host(port)/path or a relative URL which is a path with optional fragment and query sections, but this depends on the scheme used.

A host(port) is typically a domain or IP address, otionally followed by ":" and a number. Not all URL schemes require or even allow a host.

A path is typically a series of one or more Unicode strings separated by "/" characters.

The following are all valid URLs:

  • this (relative URL)
  • that/over#there (relative URL with fragment)
  • https://example.com:8899/some/thing.there?perhaps (absolute https: URL)
  • irc://irc.w3.org/#html (absolute irc: url)
  • data:text/html;charset=utf-8;base64,PCFET0NUWVBFIGh0bWw+DQo8dGl0bGU+aGVsbG88L3RpdGxlPg== (a data: URL)
  • mailto:user@ (a mailto: URL)
  • The empty string

A string is a valid non-empty URL if it is a valid URL but it is not the empty string.

A string is a valid URL potentially surrounded by spaces if, after stripping leading and trailing white space from it, it is a valid URL.

A string is a valid non-empty URL potentially surrounded by spaces if, after stripping leading and trailing white space from it, it is a valid non-empty URL.

This specification defines the URL about:legacy-compat as a reserved, though unresolvable, about: URL, for use in DOCTYPEs in HTML documents when needed for compatibility with XML tools. [RFC6694]

This specification defines the URL about:html-kind as a reserved, though unresolvable, about: URL, that is used as an identifier for kinds of media tracks. [RFC6694]

This specification defines the URL about:srcdoc as a reserved, though unresolvable, about: URL, that is used as the document’s URL of iframe srcdoc documents. [RFC6694]

The fallback base URL of a Document object is the absolute URL obtained by running these substeps:

  1. If document is an iframe srcdoc document, then return the document base URL of the Document’s browsing context’s browsing context container’s node document.

  2. If document’s URL is about:blank, and the Document’s browsing context has a creator browsing context, then return the creator base URL.

  3. Return document’s URL.

The document base URL of a Document object is the absolute URL obtained by running these substeps:

  1. If there is no base element that has an href attribute in the Document, then the document base URL is the Document's fallback base URL; abort these steps.

  2. Otherwise, the document base URL is the frozen base URL of the first base element in the Document that has an href attribute, in tree order.

2.5.2. Parsing URLs

Parsing a URL is the process of taking a URL string and obtaining the URL record that it implies. While this process is defined in the WHATWG URL specification, this specification defines a wrapper for convenience. [URL]

This wrapper is only useful when the character encoding for the URL parser has to match that of the document or environment settings object for legacy reasons. When that is not the case the URL parser can be used directly.

To parse a URL url, relative to either a document or environment settings object, the user agent must use the following steps. Parsing a URL either results in failure or a resulting URL string and resulting URL record.

  1. Let encoding be document’s character encoding, if document was given, and environment settings object’s API URL character encoding otherwise.

  2. Let baseURL be document’s base URL, if document was given, and environment settings object’s API base URL otherwise.

  3. Let urlRecord be the result of applying the URL parser to url, with baseURL and encoding.

  4. If urlRecord is failure, then abort these steps with an error.

  5. Let urlString be the result of applying the URL serializer to urlRecord.

  6. Return urlString as the resulting URL string and urlRecord as the resulting URL record.

2.5.3. Dynamic changes to base URLs

When a document’s document base URL changes, all elements in that document are affected by a base URL change.

The following are base URL change steps, which run when an element is affected by a base URL change (as defined by the DOM specification):

If the element creates a hyperlink
If the URL identified by the hyperlink is being shown to the user, or if any data derived from that URL is affecting the display, then the href attribute should be reparsed relative to the element’s node document and the UI updated appropriately.

For example, the CSS :link/:visited pseudo-classes might have been affected.

If the element is a q, blockquote, ins, or del element with a cite attribute
If the URL identified by the cite attribute is being shown to the user, or if any data derived from that URL is affecting the display, then the URL should be reparsed relative to the element’s node document and the UI updated appropriately.
Otherwise
The element is not directly affected.

For instance, changing the base URL doesn’t affect the image displayed by img elements, although subsequent accesses of the src IDL attribute from script will return a new absolute URL that might no longer correspond to the image being shown.

2.6. Fetching resources

2.6.1. Terminology

User agents can implement a variety of transfer protocols, but this specification mostly defines behavior in terms of HTTP. [HTTP]

The HTTP GET method is equivalent to the default retrieval action of the protocol. For example, RETR in FTP. Such actions are idempotent and safe, in HTTP terms.

The HTTP response codes are equivalent to statuses in other protocols that have the same basic meanings. For example, a "file not found" error is equivalent to a 404 code, a server error is equivalent to a 5xx code, and so on.

The HTTP headers are equivalent to fields in other protocols that have the same basic meaning. For example, the HTTP authentication headers are equivalent to the authentication aspects of the FTP protocol.

A referrer source is either a Document or a URL.

To create a potential-CORS request, given a url, corsAttributeState, and an optional same-origin fallback flag, run these steps:

  1. Let mode be "no-cors" if corsAttributeState is No CORS, and "cors" otherwise.

  2. If same-origin fallback flag is set and mode is "no-cors", set mode to "same-origin".

  3. Let credentialsMode be "include".

  4. If corsAttributeState is Anonymous, set credentialsMode to "same-origin".

  5. Let request be a new request whose URL is url, destination is "subresource", mode is mode, credentials mode is credentialsMode, and whose use-URL-credentials flag is set.

2.6.2. Processing model

When a user agent is to fetch a resource or URL, optionally from an origin origin, optionally using a specific referrer source as an override referrer source, and optionally with any of a synchronous flag, a manual redirect flag, a force same-origin flag, and a block cookies flag, the following steps must be run. (When a URL is to be fetched, the URL identifies a resource to be obtained.)

  1. If there is a specific override referrer source, and it is a URL, then let referrer be the override referrer source, and jump to the step labeled clean referrer.

  2. Let document be the appropriate Document as given by the following list:

    If there is a specific override referrer source
    The override referrer source.
    When navigating
    The active document of the source browsing context.
    When fetching resources for an element
    The element’s Document.
  3. While document is an iframe srcdoc document, let document be document’s browsing context’s browsing context container’s Document instead.

  4. If the origin of Document is not a scheme/host/port tuple, then set referrer to the empty string and jump to the step labeled Clean referrer.

  5. Let referrer be document’s URL.

  6. Clean referrer: Apply the URL parser to referrer and let parsed referrer be the resulting URL record.

  7. Let referrer be the result of applying the URL serializer to parsed referrer, with the exclude fragment flag set.

  8. If referrer is not the empty string, is not a data: URL, and is not the URL "about:blank", then generate the address of the resource from which Request-URIs are obtained as required by HTTP for the Referer (sic) header from referrer. [HTTP]

    Otherwise, the Referer (sic) header must be omitted, regardless of its value.

  9. If the algorithm was not invoked with the synchronous flag, perform the remaining steps in parallel.

  10. If the Document with which any tasks queued by this algorithm would be associated doesn’t have an associated browsing context, then abort these steps.

  11. This is the main step.

    If the resource is identified by an absolute URL, and the resource is to be obtained using an idempotent action (such as an HTTP GET or equivalent), and it is already being downloaded for other reasons (e.g., another invocation of this algorithm), and this request would be identical to the previous one (e.g., same Accept and Origin headers), and the user agent is configured such that it is to reuse the data from the existing download instead of initiating a new one, then use the results of the existing download instead of starting a new one.

    Otherwise, if the resource is identified by an absolute URL with a scheme that does not define a mechanism to obtain the resource (e.g., it is a mailto: URL) or that the user agent does not support, then act as if the resource was an HTTP 204 No Content response with no other metadata.

    Otherwise, if the resource is identified by the URL about:blank, then the resource is immediately available and consists of the empty string, with no metadata.

    Otherwise, at a time convenient to the user and the user agent, download (or otherwise obtain) the resource, applying the semantics of the relevant specifications (e.g., performing an HTTP GET or POST operation, or reading the file from disk, or expanding data: URLs, etc).

    For the purposes of the Referer (sic) header, use the address of the resource from which Request-URIs are obtained generated in the earlier step.

    For the purposes of the Origin header, if the fetching algorithm was explicitly initiated from an origin, then the origin that initiated the HTTP request is origin. Otherwise, this is a request from a "privacy-sensitive" context. [ORIGIN]

  12. If the algorithm was not invoked with the block cookies flag, and there are cookies to be set, update the cookies. [COOKIES] (This is a fingerprinting vector.)

  13. If the fetched resource is an HTTP redirect or equivalent, then:

    If the force same-origin flag is set and the URL of the target of the redirect does not have the same origin as the URL for which the fetch algorithm was invoked
    Abort these steps and return failure from this algorithm, as if the remote host could not be contacted.
    If the manual redirect flag is set
    Continue, using the fetched resource (the redirect) as the result of the algorithm. If the calling algorithm subsequently requires the user agent to transparently follow the redirect, then the user agent must resume this algorithm from the main step, but using the target of the redirect as the resource to fetch, rather than the original resource.
    Otherwise
    First, apply any relevant requirements for redirects (such as showing any appropriate prompts). Then, redo main step, but using the target of the redirect as the resource to fetch, rather than the original resource. For HTTP requests, the new request must include the same headers as the original request, except for headers for which other requirements are specified (such as the Host header). [HTTP]

    The HTTP specification requires that 301, 302, and 307 redirects, when applied to methods other than the safe methods, not be followed without user confirmation. That would be an appropriate prompt for the purposes of the requirement in the paragraph above. [HTTP]

  14. If the algorithm was not invoked with the synchronous flag: When the resource is available, or if there is an error of some description, queue a task that uses the resource as appropriate. If the resource can be processed incrementally, as, for instance, with a progressively interlaced JPEG or an HTML file, additional tasks may be queued to process the data as it is downloaded. The task source for these tasks is the networking task source.

    Otherwise, return the resource or error information to the calling algorithm.

If the user agent can determine the actual length of the resource being fetched for an instance of this algorithm, and if that length is finite, then that length is the file’s size. Otherwise, the subject of the algorithm (that is, the resource being fetched) has no known size. (For example, the HTTP Content-Length header might provide this information.)

The user agent must also keep track of the number of bytes downloaded for each instance of this algorithm. This number must exclude any out-of-band metadata, such as HTTP headers.

The navigation processing model handles redirects itself, overriding the redirection handling that would be done by the fetching algorithm.

Whether the type sniffing rules apply to the fetched resource depends on the algorithm that invokes the rules — they are not always applicable.

Anything in this specification that refers to HTTP also applies to HTTP-over-TLS, as represented by URLs representing the https scheme. [HTTP]

User agents should report certificate errors to the user and must either refuse to download resources sent with erroneous certificates or must act as if such resources were in fact served with no encryption.

User agents should warn the user that there is a potential problem whenever the user visits a page that the user has previously visited, if the page uses less secure encryption on the second visit.

Not doing so can result in users not noticing man-in-the-middle attacks.

If a user connects to a server with a self-signed certificate, the user agent could allow the connection but just act as if there had been no encryption. If the user agent instead allowed the user to override the problem and then displayed the page as if it was fully and safely encrypted, the user could be easily tricked into accepting man-in-the-middle connections.

If a user connects to a server with full encryption, but the page then refers to an external resource that has an expired certificate, then the user agent will act as if the resource was unavailable, possibly also reporting the problem to the user. If the user agent instead allowed the resource to be used, then an attacker could just look for "secure" sites that used resources from a different host and only apply man-in-the-middle attacks to that host, for example taking over scripts in the page.

If a user bookmarks a site that uses a CA-signed certificate, and then later revisits that site directly but the site has started using a self-signed certificate, the user agent could warn the user that a man-in-the-middle attack is likely underway, instead of simply acting as if the page was not encrypted.

2.6.4. Determining the type of a resource

The Content-Type metadata of a resource must be obtained and interpreted in a manner consistent with the requirements of the MIME Sniffing specification. [MIMESNIFF]

The computed type of a resource must be found in a manner consistent with the requirements given in the MIME Sniffing specification for finding the computed media type of the relevant sequence of octets. [MIMESNIFF]

The rules for sniffing images specifically and the rules for distinguishing if a resource is text or binary are also defined in the MIME Sniffing specification. Both sets of rules return a MIME type as their result. [MIMESNIFF]

It is imperative that the rules in the MIME Sniffing specification be followed exactly. When a user agent uses different heuristics for content type detection than the server expects, security problems can occur. For more details, see the MIME Sniffing specification. [MIMESNIFF]

2.6.5. Extracting character encodings from meta elements

The algorithm for extracting a character encoding from a meta element, given a string s, is as follows. It either returns a character encoding or nothing.

  1. Let position be a pointer into s, initially pointing at the start of the string.

  2. Loop: Find the first seven characters in s after position that are an ASCII case-insensitive match for the word "charset". If no such match is found, return nothing and abort these steps.

  3. Skip any space characters that immediately follow the word "charset" (there might not be any).

  4. If the next character is not a U+003D EQUALS SIGN (=), then move position to point just before that next character, and jump back to the step labeled loop.

  5. Skip any space characters that immediately follow the equals sign (there might not be any).

  6. Process the next character as follows:

    If it is a U+0022 QUOTATION MARK character (") and there is a later U+0022 QUOTATION MARK character (") in s
    If it is a U+0027 APOSTROPHE character (') and there is a later U+0027 APOSTROPHE character (') in s
    Return the result of getting an encoding from the substring that is between this character and the next earliest occurrence of this character.
    If it is an unmatched U+0022 QUOTATION MARK character (")
    If it is an unmatched U+0027 APOSTROPHE character (')
    If there is no next character
    Return nothing.
    Otherwise
    Return the result of getting an encoding from the substring that consists of this character up to but not including the first space character or U+003B SEMICOLON character (;), or the end of s, whichever comes first.

This algorithm is distinct from those in the HTTP specification (for example, HTTP doesn’t allow the use of single quotes and requires supporting a backslash-escape mechanism that is not supported by this algorithm). While the algorithm is used in contexts that, historically, were related to HTTP, the syntax as supported by implementations diverged some time ago. [HTTP]

2.6.6. CORS settings attributes

A CORS settings attribute is an enumerated attribute. The following table lists the keywords and states for the attribute — the keywords in the left column map to the states in the cell in the second column on the same row as the keyword.

Keyword State Brief description
anonymous Anonymous Requests for the element will have their mode set to "cors" and their credentials mode set to "same-origin".
use-credentials Use Credentials Requests for the element will have their mode set to "cors" and their credentials mode set to "include".

The empty string is also a valid keyword, and maps to the Anonymous state. The attribute’s invalid value default is the Anonymous state. For the purposes of reflection, the canonical case for the Anonymous state is the anonymous keyword. The missing value default, used when the attribute is omitted, is the No CORS state.

2.6.7. Referrer policy attributes

A referrer policy attribute is an enumerated attribute. Each referrer policy, including the empty string, is a keyword for this attribute, mapping to a state of the same name.

The attribute’s invalid value default and missing value default are both the empty string state.

The impact of these states on the processing model of various fetches is defined in more detail throughout this specification, in the WHATWG Fetch standard, and in Referrer Policy. [FETCH] [REFERRERPOLICY]

Several signals can contribute to which processing model is used for a given fetch; a referrer policy attribute is only one of them. In general, the order in which these signals are processed are:
  1. First, the presence of a noreferrer link type;
  2. Then, the value of a referrer policy attribute;
  3. Then, the presence of any meta element with name attribute set to referrer.
  4. Finally, the Referrer-Policy HTTP header.

2.6.8. Nonce attributes

A nonce content attribute represents a cryptographic nonce ("number used once") which can be used by Content Security Policy to determine whether or not a given fetch will be allowed to proceed. The value is text. [CSP3]

Elements that have a nonce content attribute ensure that the crytographic nonce is only exposed to script (and not to side-channels like CSS attribute selectors) by extracting the value from the content attribute, moving it into an internal slot named [[CryptographicNonce]], and exposing it to script via the HTMLOrSVGElement interface defined below. Unless otherwise specified, the slot’s value is the empty string.

element . nonce

Returns the value of the element’s [[CryptographicNonce]] internal slot.

Can be set, to update that slot’s value.

The nonce IDL attribute must, on getting, return the value of this element’s [[CryptographicNonce]]; and on setting, set this element’s [[CryptographicNonce]] to the given value.

Whenever an element including HTMLOrSVGElement's nonce attribute is set or changed, set this element’s [[CryptographicNonce]] to the given value.

Whenever an element including HTMLOrSVGElement becomes connected to a browsing context, the user agent must execute the following steps on the element:

  1. Let CSP list be element’s shadow-including root’s CSP list.

  2. If CSP list contains a header-delivered Content Security Policy, and element has a nonce content attribute attr whose value is not the empty string, then:

    1. Set an attribute value for element using "nonce" and the empty string.

As each Document's CSP list is append-only, user agents can optimize away the contains a header-delivered Content Security Policy check by, for example, holding a flag on the Document, set when initializing a new document object.

The cloning steps for elements that include HTMLOrSVGElement must set the [[CryptographicNonce]] slot on the copy to the value of the slot on the element being cloned.

2.7. Common DOM interfaces

2.7.1. Reflecting content attributes in IDL attributes

Some IDL attributes are defined to reflect a particular content attribute. This means that on getting, the IDL attribute returns the current value of the content attribute, and on setting, the IDL attribute changes the value of the content attribute to the given value.

In general, on getting, if the content attribute is not present, the IDL attribute must act as if the content attribute’s value is the empty string; and on setting, if the content attribute is not present, it must first be added.

If a reflecting IDL attribute is a USVString attribute whose content attribute is defined to contain a URLs, then on getting, if the content attribute is absent, the IDL attribute must return the empty string. Otherwise, the IDL attribute must parse the value of the content attribute relative to the element’s node document and if that is successful, return the resulting URL string. If parsing fails, then the value of the content attribute must be returned instead, converted to a USVString. On setting, the content attribute must be set to the specified new value.

If a reflecting IDL attribute is a DOMString attribute whose content attribute is an enumerated attribute, and the IDL attribute is limited to only known values, then, on getting, the IDL attribute must return the conforming value associated with the state the attribute is in (in its canonical case), if any, or the empty string if the attribute is in a state that has no associated keyword value or if the attribute is not in a defined state (e.g., the attribute is missing and there is no missing value default). On setting, the content attribute must be set to the specified new value.

If a reflecting IDL attribute is a nullable DOMString attribute whose content attribute is an enumerated attribute, then, on getting, if the corresponding content attribute is in its missing value default then the IDL attribute must return null, otherwise, the IDL attribute must return the conforming value associated with the state the attribute is in (in its canonical case). On setting, if the new value is null, the content attribute must be removed, and otherwise, the content attribute must be set to the specified new value.

If a reflecting IDL attribute is a DOMString or USVString attribute but doesn’t fall into any of the above categories, then the getting and setting must be done in a transparent, case-preserving manner.

If a reflecting IDL attribute is an enumeration attribute, then, on getting, if the corresponding content attribute’s value case-sensitively matches one of the enumerated values, then the IDL attribute must return the content attribute’s value; otherwise it must return the content attribute’s default value. On setting, the content attribute must be set to the specified new value.

If a reflecting IDL attribute is a boolean attribute, then on getting the IDL attribute must return true if the content attribute is set, and false if it is absent. On setting, the content attribute must be removed if the IDL attribute is set to false, and must be set to the empty string if the IDL attribute is set to true. (This corresponds to the rules for boolean content attributes.)

If a reflecting IDL attribute has a signed integer type (long) then, on getting, the content attribute must be parsed according to the rules for parsing signed integers, and if that is successful, and the value is in the range of the IDL attribute’s type, the resulting value must be returned. If, on the other hand, it fails or returns an out of range value, or if the attribute is absent, then the default value must be returned instead, or 0 if there is no default value. On setting, the given value must be converted to the shortest possible string representing the number as a valid integer and then that string must be used as the new content attribute value.

If a reflecting IDL attribute has a signed integer type (long) that is limited to only non-negative numbers then, on getting, the content attribute must be parsed according to the rules for parsing non-negative integers, and if that is successful, and the value is in the range of the IDL attribute’s type, the resulting value must be returned. If, on the other hand, it fails or returns an out of range value, or if the attribute is absent, the default value must be returned instead, or -1 if there is no default value. On setting, if the value is negative, the user agent must throw an IndexSizeError exception. Otherwise, the given value must be converted to the shortest possible string representing the number as a valid non-negative integer and then that string must be used as the new content attribute value.

If a reflecting IDL attribute has an unsigned integer type (unsigned long) then, on getting, the content attribute must be parsed according to the rules for parsing non-negative integers, and if that is successful, and the value is in the range 0 to 2147483647 inclusive, the resulting value must be returned. If, on the other hand, it fails or returns an out of range value, or if the attribute is absent, the default value must be returned instead, or 0 if there is no default value. On setting, first, if the new value is in the range 0 to 2147483647, then let n be the new value, otherwise let n be the default value, or 0 if there is no default value; then, n must be converted to the shortest possible string representing the number as a valid non-negative integer and that string must be used as the new content attribute value.

If a reflecting IDL attribute has an unsigned integer type (unsigned long) that is limited to only non-negative numbers greater than zero, then the behavior is similar to the previous case, but zero is not allowed. On getting, the content attribute must first be parsed according to the rules for parsing non-negative integers, and if that is successful, and the value is in the range 1 to 2147483647 inclusive, the resulting value must be returned. If, on the other hand, it fails or returns an out of range value, or if the attribute is absent, the default value must be returned instead, or 1 if there is no default value. On setting, if the value is zero, the user agent must throw an IndexSizeError exception. Otherwise, first, if the new value is in the range 1 to 2147483647, then let n be the new value, otherwise let n be the default value, or 1 if there is no default value; then, n must be converted to the shortest possible string representing the number as a valid non-negative integer and that string must be used as the new content attribute value.

If a reflecting IDL attribute has a floating-point number type (double or unrestricted double), then, on getting, the content attribute must be parsed according to the rules for parsing floating-point number values, and if that is successful, the resulting value must be returned. If, on the other hand, it fails, or if the attribute is absent, the default value must be returned instead, or 0.0 if there is no default value. On setting, the given value must be converted to the best representation of the number as a floating-point number and then that string must be used as the new content attribute value.

If a reflecting IDL attribute has a floating-point number type (double or unrestricted double) that is limited to numbers greater than zero, then the behavior is similar to the previous case, but zero and negative values are not allowed. On getting, the content attribute must be parsed according to the rules for parsing floating-point number values, and if that is successful and the value is greater than 0.0, the resulting value must be returned. If, on the other hand, it fails or returns an out of range value, or if the attribute is absent, the default value must be returned instead, or 0.0 if there is no default value. On setting, if the value is less than or equal to zero, then the value must be ignored. Otherwise, the given value must be converted to the best representation of the number as a floating-point number and then that string must be used as the new content attribute value.

The values Infinity and Not-a-Number (NaN) values throw an exception on setting, as defined in the Web IDL specification. [WEBIDL]

If a reflecting IDL attribute has the type DOMTokenList, then on getting it must return a DOMTokenList object whose associated element is the element in question and whose associated attribute’s local name is the name of the attribute in question.

If a reflecting IDL attribute has the type HTMLElement, or an interface that descends from HTMLElement, then, on getting, it must run the following algorithm (stopping at the first point where a value is returned):

  1. If the corresponding content attribute is absent, then the IDL attribute must return null.

  2. Let candidate be the element that the document.getElementById() method would find when called on the content attribute’s element’s node document if it were passed as its argument the current value of the corresponding content attribute.

  3. If candidate is null, or if it is not type-compatible with the IDL attribute, then the IDL attribute must return null.

  4. Otherwise, it must return candidate.

On setting, if the given element has an id attribute, and has the same tree as the element of the attribute being set, and the given element is the first element in that tree whose ID is the value of that id attribute, then the content attribute must be set to the value of that id attribute. Otherwise, the content attribute must be set to the empty string.

2.7.2. Collections

The HTMLFormControlsCollection and HTMLOptionsCollection interfaces are collections derived from the HTMLCollection interface. The HTMLAllCollection however, is independent as it has a variety of unique quirks that are not desirable to inherit from HTMLCollection.

2.7.2.1. The HTMLAllCollection interface

The HTMLAllCollection interface is used for the legacy document.all attribute. It operates similarly to HTMLCollection; it also supports a variety of other legacy features required for web compatibility such as the ability to be invoked like a function (legacycaller).

All HTMLAllCollection objects are rooted at a Document and have a filter that matches all elements, so the elements represented by the collection of an HTMLAllCollection object consist of all the descendant elements of the root Document.

[LegacyUnenumerableNamedProperties]
interface HTMLAllCollection {
  readonly attribute unsigned long length;
  getter Element (unsigned long index);
  getter (HTMLCollection or Element)? namedItem(DOMString name);
  legacycaller (HTMLCollection or Element)? item(optional DOMString nameOrItem);
};
collection . length
Returns the number of elements in the collection.
element = collection . item(index)
element = collection(index)
element = collection[index]
Returns the item with index index from the collection (determined by tree order.
element = collection . item(name)
collection = collection . item(name)
element = collection . namedItem(name)
collection = collection . namedItem(name)
element = collection(name)
collection = collection(name)
element = collection[name]
collection = collection[name]
Returns the item with ID or name name from the collection.

If there are multiple matching items, then an HTMLCollection object containing all those elements is returned.

The name attribute’s value provides a name for button, input, select, and textarea. Similarly, iframe's name, object's name, meta's name, map's name, and form's name attribute’s value provides a name for their respective elements. Only the elements mentioned have a name for the purpose of this method.

The object’s supported property indices are as defined for HTMLCollection objects.

The supported property names consist of the non-empty values of all the id and name attributes of all the elements represented by the collection, in tree order, ignoring later duplicates, with the id of an element preceding its name if it contributes both, they differ from each other, and neither is the duplicate of an earlier entry.

On getting, the length attribute must return the number of nodes represented by the collection.

The indexed property getter must return the result of getting the "all"-indexed element from this HTMLAllCollection given the passed index.

The namedItem(name) method must return the result of getting the "all"-named element or elements from this HTMLAllCollection given name.

The item(nameOrIndex) method must perform the following steps:

  1. If nameOrIndex was not provided, return null.

  2. Return the result of getting the "all"-named element or elements from this HTMLAllCollection given nameOrIndex.

The following elements are considered "all"-named elements: a, button, embed, form, frame, frameset, iframe, img, input, map, meta, object, select, and textarea.

To get the "all"-indexed element from an HTMLAllCollection collection given an index index, return the element with index index in collection, or null if there is no such element at index.

To get the "all"-named element or elements from an HTMLAllCollection collection given a name name, run the following algorithm:

  1. If name is the empty string, return null.

  2. Let subCollection be an HTMLCollection object rooted at the same Document as collection, whose filter matches only elements that are either:

  3. If there is exactly one element in subCollection, then return that element.

  4. Otherwise, if subCollection is empty, return null.

  5. Otherwise, return subCollection.

2.7.2.2. The HTMLFormControlsCollection interface

The HTMLFormControlsCollection interface is used for collections of listed elements in form elements.

interface HTMLFormControlsCollection : HTMLCollection {
  // inherits length and item()
  getter (RadioNodeList or Element)? namedItem(DOMString name); // shadows inherited namedItem()
};
interface RadioNodeList : NodeList {
  attribute DOMString value;
};
collection . length
Returns the number of elements in the collection.
element = collection . item(index)
element = collection[index]
Returns the item with index index from the collection. The items are sorted in tree order.
element = collection . namedItem(name)
radioNodeList = collection . namedItem(name)
element = collection[name]
radioNodeList = collection[name]
Returns the item with ID or name name from the collection.

If there are multiple matching items, then a RadioNodeList object containing all those elements is returned.

radioNodeList . value [ = value ]
Returns the value of the first checked radio button represented by the object.

Can be set, to check the first radio button with the given value represented by the object.

The object’s supported property indices are as defined for HTMLCollection objects.

The supported property names consist of the non-empty values of all the id and name attributes of all the elements represented by the collection, in tree order, ignoring later duplicates, with the id of an element preceding its name if it contributes both, they differ from each other, and neither is the duplicate of an earlier entry.

The properties exposed in this way must be unenumerable.

The namedItem(name) method must act according to the following algorithm:

  1. If name is the empty string, return null and stop the algorithm.

  2. If, at the time the method is called, there is exactly one node in the collection that has either an id attribute or a name attribute equal to name, then return that node and stop the algorithm.

  3. Otherwise, if there are no nodes in the collection that have either an id attribute or a name attribute equal to name, then return null and stop the algorithm.

  4. Otherwise, create a new RadioNodeList object representing a live view of the HTMLFormControlsCollection object, further filtered so that the only nodes in the RadioNodeList object are those that have either an id attribute or a name attribute equal to name. The nodes in the RadioNodeList object must be sorted in tree order.

  5. Return that RadioNodeList object.


Members of the RadioNodeList interface inherited from the NodeList interface must behave as they would on a NodeList object.

The value IDL attribute on the RadioNodeList object, on getting, must return the value returned by running the following steps:

  1. Let element be the first element in tree order represented by the RadioNodeList object that is an input element whose type attribute is in the Radio Button state and whose checkedness is true. Otherwise, let it be null.

  2. If element is null, return the empty string.

  3. If element is an element with no value attribute, return the string "on".

  4. Otherwise, return the value of element’s value attribute.

On setting, the value IDL attribute must run the following steps:

  1. If the new value is the string "on": let element be the first element in tree order represented by the RadioNodeList object that is an input element whose type attribute is in the Radio Button state and whose value content attribute is either absent, or present and equal to the new value, if any. If no such element exists, then instead let element be null.

    Otherwise: let element be the first element in tree order represented by the RadioNodeList object that is an input element whose type attribute is in the Radio Button state and whose value content attribute is present and equal to the new value, if any. If no such element exists, then instead let element be null.

  2. If element is not null, then set its checkedness to true.

2.7.2.3. The HTMLOptionsCollection interface

The HTMLOptionsCollection interface is used for collections of option elements. It is always rooted on a select element and has attributes and methods that manipulate that element’s descendants.

interface HTMLOptionsCollection : HTMLCollection {
  // inherits item(), namedItem()
  attribute unsigned long length; // shadows inherited length
  setter void (unsigned long index, HTMLOptionElement? option);
  void add((HTMLOptionElement or HTMLOptGroupElement) element, optional (HTMLElement or long)? before = null);
  void remove(long index);
  attribute long selectedIndex;
};
collection . length [ = value ]
Returns the number of elements in the collection.

When set to a smaller number, truncates the number of option elements in the corresponding container.

When set to a greater number, adds new blank option elements to that container.

element = collection . item(index)
element = collection[index]
Returns the item with index index from the collection. The items are sorted in tree order.
collection[index] = element
When index is a greater number than the number of items in the collection, adds new blank option elements in the corresponding container.

When set to null, removes the item at index index from the collection.

When set to an option element, adds or replaces it at index index from the collection.

element = collection . namedItem(name)
element = collection[name]
Returns the item with ID or name name from the collection.

If there are multiple matching items, then the first is returned.

collection . add(element [, before ] )
Inserts element before the node given by before.

The before argument can be a number, in which case element is inserted before the item with that number, or an element from the collection, in which case element is inserted before that element.

If before is omitted, null, or a number out of range, then element will be added at the end of the list.

This method will throw a HierarchyRequestError exception if element is an ancestor of the element into which it is to be inserted.

collection . remove(index)
Removes the item with index index from the collection.
collection . selectedIndex [ = value ]
Returns the index of the first selected item, if any, or -1 if there is no selected item.

Can be set, to change the selection.

The object’s supported property indices are as defined for HTMLCollection objects.

On getting, the length attribute must return the number of nodes represented by the collection.

On setting, the behavior depends on whether the new value is equal to, greater than, or less than the number of nodes represented by the collection at that time. If the number is the same, then setting the attribute must do nothing. If the new value is greater, then n new option elements with no attributes and no child nodes must be appended to the select element on which the HTMLOptionsCollection is rooted, where n is the difference between the two numbers (new value minus old value). Mutation events must be fired as if a DocumentFragment containing the new option elements had been inserted. If the new value is lower, then the last n nodes in the collection must be removed from their parent nodes, where n is the difference between the two numbers (old value minus new value).

Setting length never removes or adds any optgroup elements, and never adds new children to existing optgroup elements (though it can remove children from them).

The supported property names consist of the non-empty values of all the id and name attributes of all the elements represented by the collection, in tree order, ignoring later duplicates, with the id of an element preceding its name if it contributes both, they differ from each other, and neither is the duplicate of an earlier entry.

The properties exposed in this way must be unenumerable.

When the user agent is to set the value of a new indexed property or set the value of an existing indexed property for a given property index index to a new value value, it must run the following algorithm:

  1. If value is null, invoke the steps for the remove method with index as the argument, and abort these steps.

  2. Let length be the number of nodes represented by the collection.

  3. Let n be index minus length.

  4. If n is greater than zero, then append a DocumentFragment consisting of n-1 new option elements with no attributes and no child nodes to the select element on which the HTMLOptionsCollection is rooted.

  5. If n is greater than or equal to zero, append value to the select element. Otherwise, replace the indexth element in the collection by value.

The add(element, before) method must act according to the following algorithm:

  1. If element is an ancestor of the select element on which the HTMLOptionsCollection is rooted, then throw a HierarchyRequestError exception and abort these steps.

  2. If before is an element, but that element isn’t a descendant of the select element on which the HTMLOptionsCollection is rooted, then throw a NotFoundError exception and abort these steps.

  3. If element and before are the same element, then return and abort these steps.

  4. If before is a node, then let reference be that node. Otherwise, if before is an integer, and there is a beforeth node in the collection, let reference be that node. Otherwise, let reference be null.

  5. If reference is not null, let parent be the parent node of reference. Otherwise, let parent be the select element on which the HTMLOptionsCollection is rooted.

  6. Pre-insert element into parent node before reference.

The remove(index) method must act according to the following algorithm:

  1. If the number of nodes represented by the collection is zero, abort these steps.

  2. If index is not a number greater than or equal to 0 and less than the number of nodes represented by the collection, abort these steps.

  3. Let element be the indexth element in the collection.

  4. Remove element from its parent node.

The selectedIndex IDL attribute must act like the identically named attribute on the select element on which the HTMLOptionsCollection is rooted

2.7.3. The DOMStringList interface

The DOMStringList interface is a non-fashionable retro way of representing a list of strings.

interface DOMStringList {
  readonly attribute unsigned long length;
  getter DOMString? item(unsigned long index);
  boolean contains(DOMString string);
};

New APIs must use sequence<DOMString> or equivalent rather than DOMStringList.

strings . length

Returns the number of strings in strings.

strings[index]
strings . item()(index)

Returns the string with index index from strings.

strings . contains()(string)

Returns true if strings contains string, and false otherwise.

Each DOMStringList object has an associated list.

The supported property indices for a DOMStringList object are the numbers zero to the associated list’s size minus one. If its associated list is empty, it has no supported property indices.

The length attribute’s getter must this DOMStringList object’s associated list’s size.

The item(index) method, when invoked, must return the indexth item in this DOMStringList object’s associated list, or null if index plus one is less than this DOMStringList object’s associated list’s size.

The contains(string) method, when invoked, must return true if this DOMStringList object’s associated list contains string, and false otherwise.

2.7.4. Garbage collection

There is an implied strong reference from any IDL attribute that returns a pre-existing object to that object.

For example, the window.document attribute on the Window object means that there is a strong reference from a Window object to its Document object. Similarly, there is always a strong reference from a Document to any descendant nodes, and from any node to its owner node document.

2.8. Namespaces

The HTML namespace is: http://www.w3.org/1999/xhtml

The MathML namespace is: http://www.w3.org/1998/Math/MathML

The SVG namespace is: http://www.w3.org/2000/svg

The XLink namespace is: http://www.w3.org/1999/xlink

The XML namespace is: http://www.w3.org/XML/1998/namespace

The XMLNS namespace is: http://www.w3.org/2000/xmlns/


Data mining tools and other user agents that perform operations on content without running scripts, evaluating CSS or XPath expressions, or otherwise exposing the resulting DOM to arbitrary content, may "support namespaces" by just asserting that their DOM node analogs are in certain namespaces, without actually exposing the above strings.


In the HTML syntax, namespace prefixes and namespace declarations do not have the same effect as in XML. For instance, the colon has no special meaning in HTML element names.

2.9. Safe passing of structured data

This section uses the terminology and typographic conventions from the JavaScript specification. [ECMA-262]

2.9.1. Serializable objects

Serializable objects support being serialized, and later deserialized, in a way that is independent of any given JavaScript Realm. This allows them to be stored on disk and later restored, or cloned across Document and Worker boundaries (including across documents of different origins or in different event loops).

Not all objects are serializable objects, and not all aspects of objects that are serializable objects are necessarily preserved when they are serialized.

Platform objects can be serializable objects if they implement only interfaces decorated with the [Serializable] IDL extended attribute. Such interfaces must also define the following algorithms:

serialization steps, taking a platform object value, a Record serialized, and a boolean forStorage

A set of steps that serializes the data in value into fields of serialized. The resulting data serialized into serialized must be independent of any JavaScript Realm.

These steps may throw an exception if serialization is not possible.

These steps may perform a sub-serialization to serialize nested data structures. They should not call StructuredSerialize directly, as doing so will omit the important memory argument.

The introduction of these steps should omit mention of the forStorage argument if it is not relevant to the algorithm.

deserialization steps, taking a Record serialized and a platform object value

A set of steps that deserializes the data in serialized, using it to set up value as appropriate. value will be a newly-created instance of the platform object type in question, with none of its internal data set up; setting that up is the job of these steps.

These steps may throw an exception if deserialization is not possible.

These steps may perform a sub-deserialization to deserialize nested data structures. They should not call StructuredDeserialize directly, as doing so will omit the important targetRealm and memory arguments.

It is up to the definition of individual platform objects to determine what data is serialized and deserialized by these steps. Typically the steps are very symmetric.

The [Serializable] extended attribute must take no arguments, and must not appear on anything other than an interface. It must appear only once on an interface. It must not be used on a callback interface. If it appears on a partial interface or an interface that is really a mixin, then it must also appear on the original or mixed-in-to interface, and any supplied serialization steps and deserialization steps for the partial interface or mixin should be understood as being appended to those of the original or mixed-in-to interface.

Let’s say we were defining a platform object Person, which had associated with it two pieces of associated data:
  • a name value, which is a string;

  • and a best friend value, which is either another Person instance or null

We could then define Person instances to be serializable objects by annotating the Person interface with the [Serializable] extended attribute, and defining the following accompanying algorithms:

serialization steps
  1. Set serialized.[[Name]] to value’s associated name value.

  2. Let serializedBestFriend be the sub-serialization of value’s associated best friend value.

  3. Set serialized.[[BestFriend]] to serializedBestFriend.

deserialization steps
  1. Set value’s associated name value to serialized.[[Name]].

  2. Let deserializedBestFriend be the sub-deserialization of serialized.[[BestFriend]].

  3. Set value’s associated best friend value to deserializedBestFriend.

Objects defined in the JavaScript specification are handled by the StructuredSerialize abstract operation directly.

Originally, this specification defined the concept of "cloneable objects", which could be cloned from one JavaScript Realm to another. However, to better specify the behavior of certain more complex situations, the model was updated to make the serialization and deserialization explicit.

2.9.2. Transferable objects

Transferable objects support being transferred across event loops. Transferring is effectively recreating the object while sharing a reference to the underlying data and then detaching the object being transferred. This is useful to transfer ownership of expensive resources. Not all objects are transferable objects and not all aspects of objects that are transferable objects are necessarily preserved when transferred.

Transferring is an irreversible and non-idempotent operation. Once an object has been transferred, it cannot be transferred, or indeed used, again.

Platform objects can be transferable objects if they implement only interfaces decorated with the [Transferable] IDL extended attribute. Such interfaces must also define the following algorithms:

transfer steps, taking a platform object value and a Record dataHolder

A set of steps that transfers the data in value into fields of dataHolder. The resulting data held in dataHolder must be independent of any JavaScript Realm.

These steps may throw an exception if transferral is not possible.

transfer-receiving steps, taking a Record dataHolder and a platform object value

A set of steps that receives the data in dataHolder, using it to set up value as appropriate. value will be a newly-created instance of the platform object type in question, with none of its internal data set up; setting that up is the job of these steps.

These steps may throw an exception if it is not possible to receive the transfer.

It is up to the definition of individual platform objects to determine what data is transferred by these steps. Typically the steps are very symmetric.

The [Transferable] extended attribute must take no arguments, and must not appear on anything other than an interface. It must appear only once on an interface. It must not be used on a callback interface. If it appears on a partial interface or an interface that is really a mixin, then it must also appear on the original or mixed-in-to interface, and any supplied serialization steps and deserialization steps for the partial interface or mixin should be understood as being appended to those of the original or mixed-in-to interface.

Platform objects that are transferable objects have a [[Detached]] internal slot. This is used to ensure that once a platform object has been transferred, it cannot be transferred again.

Objects defined in the JavaScript specification are handled by the StructuredSerializeWithTransfer abstract operation directly.

2.9.3. StructuredSerializeInternal ( value, forStorage [ , memory ] )

The StructuredSerializeInternal abstract operation takes as input a JavaScript value value and serializes it to a Realm-independent form, represented here as a Record. This serialized form has all the information necessary to later deserialize into a new JavaScript value in a different Realm.

This process can throw an exception, for example when trying to serialize un-serializable objects.

  1. If memory was not supplied, let memory be an empty map.

    The purpose of the memory map is to avoid serializing objects twice. This ends up preserving cycles and the identity of duplicate objects in graphs.

  2. If memory[value] exists, then return memory[value].

  3. Let deep be false.

  4. If Type(value) is Undefined, Null, Boolean, String, or Number, then return { [[Type]]: "primitive", [[Value]]: value }.

  5. If Type(value) is Symbol, then throw a "DataCloneError" DOMException.

  6. Let serialized be an uninitialized value.

  7. If value has a [[BooleanData]] internal slot, then set serialized to { [[Type]]: "Boolean", [[BooleanData]]: value.[[BooleanData]] }.

  8. Otherwise, if value has a [[NumberData]] internal slot, then set serialized to { [[Type]]: "Number", [[NumberData]]: value.[[NumberData]] }.

  9. Otherwise, if value has a [[StringData]] internal slot, then set serialized to { [[Type]]: "String", [[StringData]]: value.[[StringData]] }.

  10. Otherwise, if value has a [[DateValue]] internal slot, then set serialized to { [[Type]]: "Date", [[DateValue]]: value.[[DateValue]] }.

  11. Otherwise, if value has a [[RegExpMatcher]] internal slot, then set serialized to { [[Type]]: "RegExp", [[RegExpMatcher]]: value.[[RegExpMatcher]], [[OriginalSource]]: value.[[OriginalSource]], [[OriginalFlags]]: value.[[OriginalFlags]] }.

  12. Otherwise, if value has an [[ArrayBufferData]] internal slot, then:

    1. Let size be value.[[ArrayBufferByteLength]].

    2. If ! IsSharedArrayBuffer(value) is true, then:

      1. If forStorage is true, then throw a "DataCloneError" DOMException.

      2. Set serialized to { [[Type]]: "SharedArrayBuffer", [[ArrayBufferData]]: value.[[ArrayBufferData]], [[ArrayBufferByteLength]]: size, [[AgentCluster]]: the current Realm Record’s corresponding agent cluster }.

    3. Otherwise:

      1. If ! IsDetachedBuffer(value) is true, then throw a "DataCloneError" DOMException.

      2. Let dataCopy be ? CreateByteDataBlock(size).

        This can throw a RangeError exception upon allocation failure.

      3. Perform ! CopyDataBlockBytes(dataCopy, 0, value.[[ArrayBufferData]], 0, size).

      4. Set serialized to { [[Type]]: "ArrayBuffer", [[ArrayBufferData]]: dataCopy, [[ArrayBufferByteLength]]: size }.

  13. Otherwise, if value has a [[ViewedArrayBuffer]] internal slot, then:

    1. Let buffer be the value of value’s [[ViewedArrayBuffer]] internal slot.

    2. Let bufferSerialized be ? StructuredSerializeInternal(buffer, forStorage, memory).

    3. Assert: bufferSerialized.[[Type]] is "ArrayBuffer".

    4. If value has a [[DataView]] internal slot, then set serialized to { [[Type]]: "ArrayBufferView", [[Constructor]]: "DataView", [[ArrayBufferSerialized]]: bufferSerialized, [[ByteLength]]: value.[[ByteLength]], [[ByteOffset]]: value.[[ByteOffset]] }.

    5. Otherwise:

      1. Assert: value has a [[TypedArrayName]] internal slot.

      2. Set serialized to { [[Type]]: "ArrayBufferView", [[Constructor]]: value.[[TypedArrayName]], [[ArrayBufferSerialized]]: bufferSerialized, [[ByteLength]]: value.[[ByteLength]], [[ByteOffset]]: value.[[ByteOffset]], [[ArrayLength]]: value.[[ArrayLength]] }.

  14. Otherwise, if value has [[MapData]] internal slot, then:

    1. Set serialized to { [[Type]]: "Map", [[MapData]]: a new empty List }.

    2. Set deep to true.

  15. Otherwise, if value has [[SetData]] internal slot, then:

    1. Set serialized to { [[Type]]: "Set", [[SetData]]: a new empty List }.

    2. Set deep to true.

  16. Otherwise, if value is an Array exotic object, then:

    1. Let valueLenDescriptor be ? OrdinaryGetOwnProperty(value, "length").

    2. Let valueLen be valueLenDescriptor.[[Value]].

    3. Set serialized to { [[Type]]: "Array", [[Length]]: valueLen, [[Properties]]: a new empty List }.

    4. Set deep to true.

  17. Otherwise, if value is a platform object that is a serializable object:

    1. If value has a [[Detached]] internal slot whose value is true, then throw a "DataCloneError" DOMException.

    2. Let typeString be the identifier of the primary interface of value.

    3. Set serialized to { [[Type]]: typeString }.

    4. Set deep to true.

  18. Otherwise, if value is a platform object, then throw a "DataCloneError" DOMException.

  19. Otherwise, if IsCallable(value) is true, then throw a "DataCloneError" DOMException.

  20. Otherwise, if value has any internal slot other than [[Prototype]] or [Extensible], then throw a "DataCloneError" DOMException.

    For instance, a [[PromiseState]] or [[WeakMapData]] internal slot.

  21. Otherwise, if value is an exotic object, then throw a "DataCloneError" DOMException.

    For instance, a proxy object.

  22. Otherwise:

    1. Set serialized to { [[Type]]: "Object", [[Properties]]: a new empty List }.

    2. Set deep to true.

  23. Set memory[value] to serialized.

  24. If deep is true, then:

    1. If value has a [[MapData]] internal slot, then:

      1. Let copiedList be a new empty List.

      2. For each Record { [[Key]], [[Value]] } entry of value.[[MapData]]:

        1. Let copiedEntry be a new Record { [[Key]]: entry.[[Key]], [[Value]]: entry.[[Value]] }.

        2. If copiedEntry.[[Key]] is not the special value empty, append copiedEntry to copiedList.

      3. For each Record { [[Key]], [[Value]] } entry of copiedList:

        1. Let serializedKey be ? StructuredSerializeInternal(entry.[[Key]], forStorage, memory).

        2. Let serializedValue be ? StructuredSerializeInternal(entry.[[Value]], forStorage, memory).

        3. Append { [[Key]]: serializedKey, [[Value]]: serializedValue } to serialized.[[MapData]].

    2. Otherwise, if value has a [[SetData]] internal slot, then:

      1. Let copiedList be a new empty List.

      2. For each entry of value.[[SetData]]:

        1. If entry is not the special value empty, append entry to copiedList.

      3. For each entry of copiedList:

        1. Let serializedEntry be ? StructuredSerializeInternal(entry, forStorage, memory).

        2. Append serializedEntry to serialized.[[SetData]].

    3. Otherwise, if value is a platform object that is a serializable object, then perform the appropriate serialization steps given value, serialized, and forStorage.

      The serialization steps may need to perform a sub-serialization. This is an operation which takes as input a value subValue, and returns StructuredSerializeInternal(subValue, forStorage, memory). (In other words, a sub-serialization is a specialization of StructuredSerializeInternal to be consistent within this invocation.)

    4. Otherwise:

      1. Let enumerableKeys be a new empty List.

      2. For each key in ! value.[[OwnPropertyKeys]]():

        1. If Type(key) is String, then:

          1. Let valueDesc be ! value.[[GetOwnProperty]](key).

          2. If valueDesc.[[Enumerable]] is true, then append key to enumerableKeys.

      3. For each key in enumerableKeys:

        1. If ! HasOwnProperty(value, key) is true, then:

          1. Let inputValue be ? value.[[Get]](key, value).

          2. Let outputValue be ? StructuredSerializeInternal(inputValue, forStorage, memory).

          3. Append { [[Key]]: key, [[Value]]: outputValue } to serialized.[[Properties]].

    The key collection performed above is very similar to the JavaScript specification’s EnumerableOwnProperties operation, but crucially it uses the deterministic ordering provided by the [[OwnPropertyKeys]] internal method, instead of reordering the keys in an unspecified manner as EnumerableOwnProperties does. [ECMA-262]

  25. Return serialized.

It’s important to realize that the Records produced by StructuredSerializeInternal might contain "pointers" to other records that create circular references. For example, when we pass the following JavaScript object into StructuredSerializeInternal:
const o = {};o.myself = o;

it produces the following result:

{  \[[Type]]: "Object",
  \[[Properties]]: «
    {
      \[[Key]]: "myself",
      \[[Value]]: <a pointer to this whole structure>
    }
  »
}

2.9.4. StructuredSerialize ( value )

  1. Return ? StructuredSerializeInternal(value, false).

2.9.5. StructuredSerializeForStorage ( value )

  1. Return ? StructuredSerializeInternal(value, true).

2.9.6. StructuredDeserialize ( serialized, targetRealm [ , memory ] )

The StructuredDeserialize abstract operation takes as input a Record serialized, which was previously produced by StructuredSerialize or StructuredSerializeForStorage, and deserializes it into a new JavaScript value, created in targetRealm.

This process can throw an exception, for example when trying to allocate memory for the new objects (especially ArrayBuffer objects).

  1. If memory was not supplied, let memory be an empty map.

    The purpose of the memory map is to avoid deserializing objects twice. This ends up preserving cycles and the identity of duplicate objects in graphs.

  2. If memory[serialized] exists, then return memory[serialized].

  3. Let deep be false.

  4. Let value be an uninitialized value.

  5. If serialized contains a [[TransferConsumed]] field, then:

    1. Assert: serialized.[[TransferConsumed]] is false. (It must be impossible to get in a situation where StructuredDeserialize is being called multiple times on the same serialization, if that serialization contains transfer data holders.)

    2. Set serialized.[[TransferConsumed]] to true.

    3. If serialized.[[Type]] is "ArrayBuffer", then set value to a new ArrayBuffer object in targetRealm whose [[ArrayBufferData]] internal slot value is serialized.[[ArrayBufferData]], and whose [[ArrayBufferByteLength]] internal slot value is serialized.[[ArrayBufferByteLength]].

      In cases where the original memory occupied by [[ArrayBufferData]] is accessible during the deserialization, this step is unlikely to throw an exception, as no new memory needs to be allocated: the memory occupied by [[ArrayBufferData]] is instead just getting transferred into the new ArrayBuffer. This could be true, for example, when both the source and target Realms are in the same process.

    4. Otherwise:

      1. Let interfaceName be serialized.[[Type]].

      2. If the interface identified by interfaceName is not exposed in targetRealm, then throw a "DataCloneError" DOMException.

      3. Set value to a new instance of the interface identified by interfaceName, created in targetRealm.

      4. Perform the appropriate transfer-receiving steps for the interface identified by interfaceName given serialized and value.

  6. Otherwise, if serialized.[[Type]] is "primitive", then set value to serialized.[[Value]].

  7. Otherwise, if serialized.[[Type]] is "Boolean", then set value to a new Boolean object in targetRealm whose [[BooleanData]] internal slot value is serialized.[[BooleanData]].

  8. Otherwise, if serialized.[[Type]] is "Number", then set value to a new Number object in targetRealm whose [[NumberData]] internal slot value is serialized.[[NumberData]].

  9. Otherwise, if serialized.[[Type]] is "String", then set value to a new String object in targetRealm whose [[StringData]] internal slot value is serialized.[[StringData]].

  10. Otherwise, if serialized.[[Type]] is "Date", then set value to a new Date object in targetRealm whose [[DateValue]] internal slot value is serialized.[[DateValue]].

  11. Otherwise, if serialized.[[Type]] is "RegExp", then set value to a new RegExp object in targetRealm whose [[RegExpMatcher]] internal slot value is serialized.[[RegExpMatcher]], whose [[OriginalSource]] internal slot value is serialized.[[OriginalSource]], and whose [[OriginalFlags]] internal slot value is serialized.[[OriginalFlags]].

  12. Otherwise, if serialized.[[Type]] is "SharedArrayBuffer", then:

    1. If targetRealm’s corresponding agent cluster is not serialized.[[AgentCluster]], then throw a "DataCloneError" DOMException.

    2. Otherwise, set value to a new SharedArrayBuffer object in targetRealm whose [[ArrayBufferData]] internal slot value is serialized.[[ArrayBufferData]] and whose [[ArrayBufferByteLength]] internal slot value is serialized.[[ArrayBufferByteLength]].

  13. Otherwise, if serialized.[[Type]] is "ArrayBuffer", then set value to a new ArrayBuffer object in targetRealm whose [[ArrayBufferData]] internal slot value is serialized.[[ArrayBufferData]], and whose [[ArrayBufferByteLength]] internal slot value is serialized.[[ArrayBufferByteLength]].

    If this throws an exception, then throw a "DataCloneError" DOMException.

    This step might throw an exception if there is not enough memory available to create such an ArrayBuffer object.

  14. Otherwise, if serialized.[[Type]] is "ArrayBufferView", then:

    1. Let deserializedArrayBuffer be ? StructuredDeserialize(serialized.[[ArrayBufferSerialized]], targetRealm, memory).

    2. If serialized.[[Constructor]] is "DataView", then set value to a new DataView object in targetRealm whose [[ViewedArrayBuffer]] internal slot value is deserializedArrayBuffer, whose [[ByteLength]] internal slot value is serialized.[[ByteLength]], and whose [[ByteOffset]] internal slot value is serialized.[[ByteOffset]].

    3. Otherwise, set value to a new typed array object in targetRealm, using the constructor given by serialized.[[Constructor]], whose [[ViewedArrayBuffer]] internal slot value is deserializedArrayBuffer, whose \[TypedArrayName]] internal slot value is serialized.[[Constructor]], whose [[ByteLength]] internal slot value is serialized.[[ByteLength]], whose [[ByteOffset]] internal slot value is serialized.[[ByteOffset]], and whose [[ArrayLength]] internal slot value is serialized.[[ArrayLength]].

  15. Otherwise, if serialized.[[Type]] is "Map", then:

    1. Set value to a new Map object in targetRealm whose [[MapData]] internal slot value is a new empty List.

    2. Set deep to true.

  16. Otherwise, if serialized.[[Type]] is "Set", then:

    1. Set value to a new Set object in targetRealm whose [[SetData]] internal slot value is a new empty List.

    2. Set deep to true.

  17. Otherwise, if serialized.[[Type]] is "Array", then:

    1. Let outputProto be the %ArrayPrototype% intrinsic object in targetRealm.

    2. Set value to ! ArrayCreate(serialized.[[Length]], outputProto).

    3. Set deep to true.

  18. Otherwise, if serialized.[[Type]] is "Object", then:

    1. Set value to a new Object in targetRealm.

    2. Set deep to true.

  19. Otherwise:

    1. Let interfaceName be serialized.[[Type]].

    2. If the interface identified by interfaceName is not exposed in targetRealm, then throw a "DataCloneError" DOMException.

    3. Set value to a new instance of the interface identified by interfaceName, created in targetRealm.

    4. Set deep to true.

  20. Set memory[serialized] to value.

  21. If deep is true, then:

    1. If serialized.[[Type]] is "Map", then:

      1. For each Record { [[Key]], [[Value]] } entry of serialized.[[MapData]]:

        1. Let deserializedKey be ? StructuredDeserialize(entry.[[Key]], targetRealm, memory).

        2. Let deserializedValue be ? StructuredDeserialize(entry.[[Value]], targetRealm, memory).

        3. Append { [[Key]]: deserializedKey, [Value]]: deserializedValue } to value.[[MapData]].

    2. Otherwise, if serialized.[[Type]] is "Set", then:

      1. For each entry of serialized.[[SetData]]:

        1. Let deserializedEntry be ? StructuredDeserialize(entry, targetRealm, memory).

        2. Append deserializedEntry to value.[[SetData]].

    3. Otherwise, if serialized.[[Type]] is "Array" or "Object", then:

      1. For each Record { [[Key]], [[Value]] } entry of serialized.[[Properties]]:

        1. Let deserializedValue be ? StructuredDeserialize(entry.[[Value]], targetRealm, memory).

        2. Let result be ! CreateDataProperty(value, entry.[[Key]], deserializedValue).

        3. Assert: result is true.

    4. Otherwise:

      1. Perform the appropriate deserialization steps for the interface identified by serialized.[[Type]], given serialized and value.

        The deserialization steps may need to perform a sub-deserialization. This is an operation which takes as input a previously-serialized Record subSerialized, and returns StructuredDeserialize(subSerialized, targetRealm, memory). (In other words, a sub-deserialization is a specialization of StructuredDeserialize to be consistent within this invocation.)

  22. Return value.

2.9.7. StructuredSerializeWithTransfer ( value, transferList )

  1. Let memory be an empty map.

    In addition to how it is used normally by StructuredSerializeInternal, in this algorithm memory is also used to ensure that StructuredSerializeInternal ignores items in transferList, and let us do our own handling instead.

  2. For each transferable of transferList:

    1. If transferable has neither an [[ArrayBufferData]] internal slot nor a [[Detached]] internal slot, then throw a "DataCloneError" DOMException.

    2. If transferable has an [[ArrayBufferData]] internal slot and either ! IsSharedArrayBuffer(transferable) is true or ! IsDetachedBuffer(transferable) is true, then throw a "DataCloneError" DOMException.

    3. If transferable has a [[Detached]] internal slot and transferable.[[Detached]] is true, then throw a "DataCloneError" DOMException.

    4. Let placeholder be a user-agent-defined placeholder object.

    5. Set memory[transferable] to placeholder.

  3. Let serialized be ? StructuredSerializeInternal(value, false, memory).

  4. Let transferDataHolders be a new empty List.

  5. For each transferable of transferList:

    1. Let placeholder be memory[transferable].

    2. Let dataHolder be an uninitialized value.

    3. If transferable has an [[ArrayBufferData]] internal slot, then:

      1. Set dataHolder to { [[TransferConsumed]]: false, [[Type]]: "ArrayBuffer", [[ArrayBufferData]]: transferable.[[ArrayBufferData]], [[ArrayBufferByteLength]]: transferable.[[ArrayBufferByteLength]] }.

      2. Perform ! DetachArrayBuffer(transferable).

    4. Otherwise:

      1. Assert: transferable is a platform object that is a transferable object.

      2. Let interfaceName be the identifier of the primary interface of transferable.

      3. Set dataHolder to { [[TransferConsumed]]: false, [[Type]]: interfaceName }.

      4. Perform the appropriate transfer steps for the interface identified by interfaceName, given transferable and dataHolder.

      5. Set transferable.[[Detached]] to true.

    5. Within serialized, replace all instances of placeholder with dataHolder.

    6. Append dataHolder to transferDataHolders.

  6. Return { [[Serialized]]: serialized, [[TransferDataHolders]]: transferDataHolders }.

2.9.8. StructuredDeserializeWithTransfer ( serializeWithTransferResult, targetRealm )

  1. Let memory be an empty map.

    In addition to how it is used normally by StructuredDeserialize, in this algorithm memory is also used to help us determine the list of transferred values.

  2. Let deserialized be ? StructuredDeserialize(serializeWithTransferResult.[[Serialized]], targetRealm, memory).

  3. Let transferredValues be a new empty List.

  4. For each transferDataHolder of serializeWithTransferResult.[[TransferDataHolders]]:

    1. Append memory[transferDataHolder] to transferredValues.

  5. Return { [[Deserialized]]: deserialized, [[TransferredValues]]: transferredValues }.

2.9.9. Performing serialization and transferring from other specifications

Other specifications may use the abstract operations defined here. The following provides some guidance on when each abstract operation is typically useful, with examples.

StructuredSerializeWithTransfer
StructuredDeserializeWithTransfer

Cloning a value to another JavaScript Realm, with a transfer list, but where the target Realm is not known ahead of time. In this case the serialization step can be performed immediately, with the deserialization step delayed until the target Realm becomes known.

messagePort.postMessage() uses this pair of abstract operations, as the destination Realm is not known until the MessagePort has been shipped.

StructuredSerialize
StructuredSerializeForStorage
StructuredDeserialize

Creating a JavaScript Realm-independent snapshot of a given value which can be saved for an indefinite amount of time, and then reified back into a JavaScript value later, possibly multiple times.

StructuredSerializeForStorage can be used for situations where the serialization is anticipated to be stored in a persistent manner, instead of passed between Realms. It throws when attempting to serialize SharedArrayBuffer objects, since storing shared memory does not make sense. Similarly, it can throw or possibly have different behavior when given a platform object with custom serialization steps when the forStorage argument is true.

history.pushState() and history.replaceState() use StructuredSerializeForStorage on author-supplied state objects, storing them as serialized state in the appropriate session history entry. Then, StructuredDeserialize is used so that the history.state property can return a clone of the originally-supplied state object.

broadcastChannel.postMessage() uses StructuredSerialize on its input, then uses StructuredDeserialize multiple times on the result to produce a fresh clone for each destination being broadcast to. Note that transferring does not make sense in multi-destination situations.

Any API for persisting JavaScript values to the filesystem would also use StructuredSerializeForStorage on its input and StructuredDeserialize on its output.

In general, call sites may pass in Web IDL values instead of JavaScript values; this is to be understood to perform an implicit conversion to the JavaScript value before invoking these algorithms.

This specification used to define a "structured clone" algorithm, and more recently a StructuredClone abstract operation. However, in practice all known uses of it were better served by separate serialization and deserialization steps, so it was removed.


Call sites that are not invoked as a result of author code synchronously calling into a user agent method must take care to properly prepare to run script and prepare to run a callback before invoking StructuredSerialize, StructuredSerializeForStorage, or StructuredSerializeWithTransfer abstract operations, if they are being performed on arbitrary objects. This is necessary because the serialization process can invoke author-defined accessors as part of its final deep-serialization steps, and these accessors could call into operations that rely on the entry and incumbent concepts being properly set up.

window.postMessage() performs StructuredSerializeWithTransfer on its arguments, but is careful to do so immediately, inside the synchronous portion of its algorithm. Thus it is able to use the algorithms without needing to prepare to run script and prepare to run a callback.

In contrast, a hypothetical API that used StructuredSerialize to serialize some author-supplied object periodically, directly from a task on the event loop, would need to ensure it performs the appropriate preparations beforehand. As of this time, we know of no such APIs on the platform; usually it is simpler to perform the serialization ahead of time, as a synchronous consequence of author code.

3. Semantics, structure, and APIs of HTML documents

3.1. Documents

Every XML and HTML document in an HTML UA is represented by a Document object. [DOM]

The Document object’s URL is defined in the DOM specification. It is initially set when the Document object is created, but that can change during the lifetime of the Document object; for example, it changes when the user navigates to a fragment on the page and when the pushState() method is called with a new URL. [DOM]

Interactive user agents typically expose the Document object’s URL in their user interface. This is the primary mechanism by which a user can tell if a site is attempting to impersonate another.

When a Document is created by a script using the createDocument() or createHTMLDocument() APIs, the Document is both ready for post-load tasks and completely loaded immediately.

The document’s referrer is a string (representing a URL) that can be set when the Document is created. If it is not explicitly set, then its value is the empty string.

Each Document object has a reload override flag that is originally unset. The flag is set by the document.open() and document.write() methods in certain situations. When the flag is set, the Document also has a reload override buffer which is a Unicode string that is used as the source of the document when it is reloaded.

When the user agent is to perform an overridden reload, given a source browsing context, it must act as follows:

  1. Let source be the value of the browsing context's active document's reload override buffer.

  2. Let address be the browsing context's active document's URL.

  3. Let HTTPS state be the HTTPS state of the browsing context's active document.

  4. Let referrer policy be the referrer policy of the browsing context's active document.

  5. Let CSP list be the CSP list of the browsing context's active document.

  6. Navigate the browsing context to a new response whose body is source, header list is Referrer-Policy/referrer policy, CSP list is CSP list and HTTPS state is HTTPS state, with the exceptions enabled flag set and replacement enabled. The source browsing context is that given to the overridden reload algorithm. When the navigate algorithm creates a Document object for this purpose, set that Document's reload override flag and set its reload override buffer to source. Rethrow any exceptions.

    When it comes time to set the document’s address in the navigation algorithm, use address as the override URL.

3.1.1. The Document object

The DOM specification defines a Document interface, which this specification extends significantly:

enum DocumentReadyState { "loading", "interactive", "complete" };

typedef (HTMLScriptElement or SVGScriptElement) HTMLOrSVGScriptElement;

[OverrideBuiltins]
partial interface Document {
  // resource metadata management
  [PutForwards=href, Unforgeable] readonly attribute Location? location;
  attribute USVString domain;
  readonly attribute USVString referrer;
  attribute USVString cookie;
  readonly attribute DOMString lastModified;
  readonly attribute DocumentReadyState readyState;

  // DOM tree accessors
  getter object (DOMString name);
  [CEReactions] attribute DOMString title;
  attribute DOMString dir;
  attribute HTMLElement? body;
  readonly attribute HTMLHeadElement? head;
  [SameObject] readonly attribute HTMLCollection images;
  [SameObject] readonly attribute HTMLCollection embeds;
  [SameObject] readonly attribute HTMLCollection plugins;
  [SameObject] readonly attribute HTMLCollection links;
  [SameObject] readonly attribute HTMLCollection forms;
  [SameObject] readonly attribute HTMLCollection scripts;
  NodeList getElementsByName(DOMString elementName);
  readonly attribute HTMLOrSVGScriptElement? currentScript; // classic scripts in a document tree only

  // dynamic markup insertion
  Document open(optional DOMString type = "text/html", optional DOMString replace = "");
  WindowProxy open(DOMString url, DOMString name, DOMString features, optional boolean replace = false);
  [CEReactions] void close();
  [CEReactions] void write(DOMString... text);
  [CEReactions] void writeln(DOMString... text);

  // user interaction
  readonly attribute WindowProxy? defaultView;
  readonly attribute Element? activeElement;
  boolean hasFocus();
  [CEReactions] attribute DOMString designMode;
  [CEReactions] boolean execCommand(DOMString commandId, optional boolean showUI = false, optional DOMString value = "");
  boolean queryCommandEnabled(DOMString commandId);
  boolean queryCommandIndeterm(DOMString commandId);
  boolean queryCommandState(DOMString commandId);
  boolean queryCommandSupported(DOMString commandId);
  DOMString queryCommandValue(DOMString commandId);

  // special event handler IDL attributes that only apply to Document objects
  [LenientThis] attribute EventHandler onreadystatechange;
};
Document implements GlobalEventHandlers;
Document implements DocumentAndElementEventHandlers;

The Document has an HTTPS state (an HTTPS state value), initially "none", which represents the security properties of the network channel used to deliver the Document's data.

The Document has a referrer policy (a referrer policy), initially the empty string, which represents the default referrer policy used by fetches initiated by the Document.

The Document has a CSP list, which is a CSP list containing all of the Content Security Policy objects active for the document. The list is empty unless otherwise specified.

The Document has a module map, which is a module map, initially empty.

3.1.2. Resource metadata management

document . referrer

Returns the URL of the Document from which the user navigated to this one, unless it was blocked or there was no such document, in which case it returns the empty string.

The noreferrer link type can be used to block the referrer.

The referrer attribute must return the document’s referrer.


document . cookie [ = value ]

Returns the HTTP cookies that apply to the Document. If there are no cookies or cookies can’t be applied to this resource, the empty string will be returned.

Can be set, to add a new cookie to the element’s set of HTTP cookies.

If the contents are sandboxed into a unique origin (e.g., in an iframe with the sandbox attribute), a "SecurityError" DOMException will be thrown on getting and setting.

The cookie attribute represents the cookies of the resource identified by the document’s URL.

A Document object that falls into one of the following conditions is a cookie-averse Document object:

On getting, if the document is a cookie-averse Document object, then the user agent must return the empty string. Otherwise, if the Document's origin is an opaque origin, the user agent must throw a "SecurityError" DOMException. Otherwise, the user agent must return the cookie-string for the document’s URL for a "non-HTTP" API, decoded using UTF-8 decode without BOM. [COOKIES] (This is a fingerprinting vector.)

On setting, if the document is a cookie-averse Document object, then the user agent must do nothing. Otherwise, if the Document's origin is an opaque origin, the user agent must throw a "SecurityError" DOMException. Otherwise, the user agent must act as it would when receiving a set-cookie-string for the document’s URL via a "non-HTTP" API, consisting of the new value encoded as UTF-8. [COOKIES] [ENCODING]

Since the cookie attribute is accessible across frames, the path restrictions on cookies are only a tool to help manage which cookies are sent to which parts of the site, and are not in any way a security feature.

The cookie attribute’s getter and setter synchronously access shared state. Since there is no locking mechanism, other browsing contexts in a multiprocess user agent can modify cookies while scripts are running. A site could, for instance, try to read a cookie, increment its value, then write it back out, using the new value of the cookie as a unique identifier for the session; if the site does this twice in two different browser windows at the same time, it might end up using the same "unique" identifier for both sessions, with potentially disastrous effects.


document . lastModified

Returns the date of the last modification to the document, as reported by the server, in the form "MM/DD/YYYY hh:mm:ss", in the user’s local time zone.

If the last modification date is not known, the current time is returned instead.

The lastModified attribute, on getting, must return the date and time of the Document's source file’s last modification, in the user’s local time zone, in the following format:

  1. The month component of the date.

  2. A U+002F SOLIDUS character (/).

  3. The day component of the date.

  4. A U+002F SOLIDUS character (/).

  5. The year component of the date.

  6. A U+0020 SPACE character.

  7. The hours component of the time.

  8. A U+003A COLON character (:).

  9. The minutes component of the time.

  10. A U+003A COLON character (:).

  11. The seconds component of the time.

All the numeric components above, other than the year, must be given as two ASCII digits representing the number in base ten, zero-padded if necessary. The year must be given as the shortest possible string of four ASCII digits representing the number in base ten, zero-padded if necessary.

The Document's source file’s last modification date and time must be derived from relevant features of the networking protocols used, e.g., from the value of the HTTP Last-Modified header of the document, or from metadata in the file system for local files. If the last modification date and time are not known, the attribute must return the current date and time in the above format.


document . readyState

Returns "loading" while the Document is loading, "interactive" once it is finished parsing but still loading sub-resources, and "complete" once it has loaded.

The readystatechange event fires on the Document object when this value changes.

Each document has a current document readiness. When a Document object is created, it must have its current document readiness set to the string "loading" if the document is associated with an HTML parser, an XML parser, or an XSLT processor, and to the string "complete" otherwise. Various algorithms during page loading affect this value. When the value is set, the user agent must fire an event named readystatechange at the Document object.

A Document is said to have an active parser if it is associated with an HTML parser or an XML parser that has not yet been stopped or aborted.

The readyState IDL attribute must, on getting, return the current document readiness.

3.1.3. DOM tree accessors

The html element of a document is its document element, if it’s an html element, and null otherwise.


document . head

Returns the head element.

The head element of a document is the first head element that is a child of the html element, if there is one, or null otherwise.

The head attribute, on getting, must return the head element of the document (a head element or null).


document . title [ = value ]

Returns the document’s title, as given by the title element for HTML and as given by the SVG title element for SVG.

Can be set, to update the document’s title. If there is no appropriate element to update, the new value is ignored.

The title element of a document is the first title element in the document (in tree order), if there is one, or null otherwise.

The title attribute must, on getting, run the following algorithm:

  1. If the document element is an SVG svg element, then let value be the child text content of the first SVG title element that is a child of the document element. [SVG11]

  2. Otherwise, let value be the child text content of the title element, or the empty string if the title element is null.

  3. Strip and collapse white space in value.

  4. Return value.

On setting, the steps corresponding to the first matching condition in the following list must be run:

If the document element is an SVG svg element
  1. If there is an SVG title element that is a child of the document element, let element be the first such element.

  2. Otherwise:

    1. Let element be the result of creating an element given the document element's node document, SVG title, and the SVG namespace.

    2. Insert element as the first child of the document element.

  3. Act as if the textContent IDL attribute of element was set to the new value being assigned.

If the document element is in the HTML namespace
  1. If the title element is null and the head element is null, then abort these steps.

  2. If the title element is non-null, let element be the title element.

  3. Otherwise:

    1. Let element be the result of creating an element given the document element's node document, title, and the HTML namespace.

    2. Append element to the head element.

  4. Act as if the textContent IDL attribute of element was set to the new value being assigned.

Otherwise

Do nothing.


document . body [ = value ]

Returns the body element.

Can be set, to replace the body element.

If the new value is not a body or frameset element, this will throw a "HierarchyRequestError" DOMException.

The body element of a document is the first child of the html element that is either a body element or a frameset element. If there is no such element, it is null.

The body attribute, on getting, must return the body element of the document (either a body element, a frameset element, or null). On setting, the following algorithm must be run:

  1. If the new value is not a body or frameset element, then throw a "HierarchyRequestError" DOMException and abort these steps.

  2. Otherwise, if the new value is the same as the body element, do nothing. Abort these steps.

  3. Otherwise, if the body element is not null, then replace the body element with the new value within the body element’s parent and abort these steps.

  4. Otherwise, if there is no document element, throw a "HierarchyRequestError" DOMException and abort these steps.

  5. Otherwise, the body element is null, but there’s a document element. Append the new value to the document element.


document . images

Returns an HTMLCollection of the img elements in the Document.

document . embeds
document . plugins

Return an HTMLCollection of the embed elements in the Document.

document . links

Returns an HTMLCollection of the a and area elements in the Document that have href attributes.

document . forms

Return an HTMLCollection of the form elements in the Document.

document . scripts

Return an HTMLCollection of the script elements in the Document.

The images attribute must return an HTMLCollection rooted at the Document node, whose filter matches only img elements.

The embeds attribute must return an HTMLCollection rooted at the Document node, whose filter matches only embed elements.

The plugins attribute must return the same object as that returned by the embeds attribute.

The links attribute must return an HTMLCollection rooted at the Document node, whose filter matches only a elements with href attributes and area elements with href attributes.

The forms attribute must return an HTMLCollection rooted at the Document node, whose filter matches only form elements.

The scripts attribute must return an HTMLCollection rooted at the Document node, whose filter matches only script elements.


collection = document . getElementsByName(name)

Returns a NodeList of elements in the Document that have a name attribute with the value name.

The getElementsByName(name) method takes a string name, and must return a live NodeList containing all the HTML elements in that document that have a name attribute whose value is equal to the name argument (in a case-sensitive manner), in tree order. When the method is invoked on a Document object again with the same argument, the user agent may return the same as the object returned by the earlier call. In other cases, a new NodeList object must be returned.


document . currentScript

Returns the script element, or the SVG script element, that is currently executing, as long as the element represents a classic script. In the case of reentrant script execution, returns the one that most recently started executing amongst those that have not yet finished executing.

Returns null if the Document is not currently executing a script element or SVG script element (e.g., because the running script is an event handler, or a timeout), or if the currently executing script or SVG script element represents a module script.

The currentScript attribute, on getting, must return the value to which it was most recently initialized. When the Document is created, the currentScript must be initialized to null.

This API has fallen out of favor in the implementor and standards community, as it globally exposes script or SVG script elements. As such, it is not available in newer contexts, such as when running module scripts or when running scripts in a shadow tree.


The Document interface supports named properties. The supported property names of a Document object document at any moment consist of the following, in tree order according to the element that contributed them, ignoring later duplicates, and with values from id attributes coming before values from name attributes when the same element contributes both:

To determine the value of a named property name for a Document, the user agent must return the value obtained using the following steps:

  1. Let elements be the list of named elements with the name name that are in a document tree with the Document as their root.

    There will be at least one such element, by definition.

  2. If elements has only one element, and that element is an iframe element, and that iframe element’s nested browsing context is not null, then return the WindowProxy object of the element’s nested browsing context.

  3. Otherwise, if elements has only one element, return that element.

  4. Otherwise return an HTMLCollection rooted at the Document node, whose filter matches only named elements with the name name.

Named elements with the name name, for the purposes of the above algorithm, are those that are either:

An embed or object element is said to be exposed if it has no exposed object ancestor, and, for object elements, is additionally either not showing its fallback content or has no object or embed descendants.


The dir attribute on the Document interface is defined along with the dir content attribute.

3.2. Elements

3.2.1. Semantics

Elements, attributes, and attribute values in HTML are defined (by this specification) to have certain meanings (semantics). For example, the ol element represents an ordered list, and the lang attribute represents the language of the content.

These definitions allow HTML processors, like web browsers and search engines, to present documents and applications consistently in different contexts.

In this example the HTML headings may be presented as large text in a desktop browser, or standard size text in bold in a mobile browser. In both cases the semantic information remains the same - that the h1 and h2 elements represent headings.
<!DOCTYPE html>
<html lang="en">
  <head>
    <title>Favorite books</title>
  </head>
  <body>
    <header>
      <img src="logo.png" alt="Favorite books logo">
    </header>
    <main>
      <h1>Favorite books</h1>
      <p>These are a few of my favorite books.</p>
      <h2>White Teeth</h2>
      <p>The award winning debut novel by Zadie Smith</p>
      <h2>The Hitchhiker’s Guide to the Galaxy</h2>
      <p>A trilogy of five books by Douglas Adams.</p>
    </main>
  </body>
</html>

This semantic information is critical to assistive technologies. For example, a screen reader will query the browser for semantic information and use that information to present the document or application in synthetic speech.

In some cases assistive technologies use semantic information to provide additional functionality. A speech recognition tool might provide a voice command for moving focus to the start of the main element for example.

When the appropriate HTML element or attribute is not used, it deprives HTML processors of valuable semantic information.

In this example styling may be used to create a visual representation of headings and other components, but because the appropriate HTML elements have not been used there is little semantic information available to web browsers, search engines and assistive technologies.
<!DOCTYPE html>
<html lang="en">
  <head>
    <title>Favorite books</title>
  </head>
  <body>
    <header>
      <img src="logo.png" alt="Favorite books logo">
    </header>
    <main>
      <span class="largeHeading">Favorite books</span>
      <p>These are a few of my favorite books.</p>
      <span class="smallHeading">White Teeth</span>
      <p>The award winning debut novel by Zadie Smith</p>
      <span class="smallHeading">The Hitchhiker’s Guide to the Galaxy</span>
      <p>A trilogy of five books by Douglas Adams.</p>
    </main>
  </body>
</html>

A document can change dynamically while it is being processed. Scripting and other mechanisms can be used to change attribute values, text, or the entire document structure. The semantics of a document are therefore based on the document’s state at a particular instance in time, but may also change in response to external events. User agents must update their presentation of the document to reflect these changes.

DOM nodes whose node document does not have a browsing context are exempt from all document conformance requirements other than the HTML syntax requirements and the XML syntax requirements.

In this example the audio element is used to play a music track. The controls attribute is used to show the user agent player, and as the music plays the controls are updated to indicate progress. The available semantic information is updated in response to these changes.
<audio src="comfortablynumb.mp3" controls>

3.2.2. Elements in the DOM

The nodes representing HTML elements in the DOM must implement, and expose to scripts, the interfaces listed for them in the relevant sections of this specification. This includes HTML elements in XML documents, even when those documents are in another context (e.g., inside an XSLT transform).

Elements in the DOM represent things; that is, they have intrinsic meaning, also known as semantics.

For example, an ol element represents an ordered list.

The basic interface, from which all the HTML elements' interfaces inherit, and which must be used by elements that have no additional requirements, is the HTMLElement interface.

[HTMLConstructor]
interface HTMLElement : Element {
  // metadata attributes
  [CEReactions] attribute DOMString title;
  [CEReactions] attribute DOMString lang;
  [CEReactions] attribute boolean translate;
  [CEReactions] attribute DOMString dir;

  // user interaction
  [CEReactions] attribute boolean hidden;
  void click();
  [CEReactions] attribute DOMString accessKey;
  [CEReactions] attribute boolean draggable;
  [CEReactions] attribute boolean spellcheck;
  void forceSpellCheck();
  [CEReactions, TreatNullAs=EmptyString] attribute DOMString innerText;
};
HTMLElement implements GlobalEventHandlers;
HTMLElement implements DocumentAndElementEventHandlers;
HTMLElement implements ElementContentEditable;

// Note: intentionally not [HTMLConstructor]
interface HTMLUnknownElement : HTMLElement { };

The HTMLElement interface holds methods and attributes related to a number of disparate features, and the members of this interface are therefore described in various different sections of this specification.

The HTMLUnknownElement interface must be used for HTML elements that are not defined by this specification (or other applicable specifications).

The element interface for an element with name name in the HTML namespace is determined as follows:

  1. If name is applet, bgsound, blink, isindex, multicol, nextid, or spacer, then return HTMLUnknownElement.

  2. If name is acronym, basefont, big, center, nobr, noembed, noframes, plaintext, rb, rtc, strike, or tt, then return HTMLElement.

  3. If name is listing or xmp, then return HTMLPreElement.

  4. Otherwise, if this specification defines an interface appropriate for the element type corresponding to the local name name, then return that interface.

  5. If other applicable specifications define an appropriate interface for name, then return the interface they define.

  6. If name is a valid custom element name, then return HTMLElement.

  7. Return HTMLUnknownElement.

The use of HTMLElement instead of HTMLUnknownElement in the case of valid custom element names is done to ensure that any potential future upgrades only cause a linear transition of the element’s prototype chain, from HTMLElement to a subclass, instead of a lateral one, from HTMLUnknownElement to an unrelated subclass.

Features shared between HTML and SVG elements are defined in the HTMLOrSVGElement interface mixin:

interface mixin HTMLOrSVGElement {
  [SameObject] readonly attribute DOMStringMap dataset;
  attribute DOMString nonce;

  [CEReactions] attribute long tabIndex;
  void focus(optional FocusOptions options);
  void blur();
};

HTMLElement includes HTMLOrSVGElement;
SVGElement includes HTMLOrSVGElement;

3.2.3. The CustomElementRegistry interface

Each Window object is associated with a unique instance of a CustomElementRegistry object, allocated when the Window object is created.

Custom element registries are associated with Window objects, instead of Document objects, since each custom element constructor inherits from the HTMLElement interface, and there is exactly one HTMLElement interface per Window object.

The customElements attribute of the Window interface returns the CustomElementRegistry object for that Window object.

interface CustomElementRegistry {
[CEReactions] void define(DOMString name, Function constructor, optional ElementDefinitionOptions options);
any get(DOMString name);
Promise<void> whenDefined(DOMString name);
};

dictionary ElementDefinitionOptions {
DOMString extends;
};

Every CustomElementRegistry has:

window . customElements . {{define()|define(name}}, constructor)
Defines a new autonomous custom element, mapping the given name to the given constructor.
window . customElements . {{define()|define(name, constructor), { extends: baseLocalName })}}
Defines a new customized built-in element, mapping the given name to the given constructor for the element type identified by the supplied baseLocalName. A "NotSupportedError" DOMException will be thrown upon trying to extend a custom element or an unknown element.
window . customElements . get(name)
Retrieves the custom element constructor defined for the given name. Returns undefined if there is no custom element definition with the given name.
window . customElements . {{whenDefined()|whenDefined(name)}}
Returns a promise that will be fulfilled when a custom element becomes defined with the name name. (If such a custom element is already defined, the returned promise will be immediately fulfilled.) Returns a promise rejected with a "SyntaxError" DOMException if name is not a valid custom element name.

Element definition is the process of adding a custom element definition to the CustomElementRegistry. This is accomplished by the define() method. When invoked, the {{define()|define(name, constructor, options)}} method must run these steps:

  1. If IsConstructor(constructor) is false, then throw a TypeError and abort these steps.

  2. If name is not a valid custom element name, then throw a "SyntaxError" DOMException and abort these steps.

  3. If this CustomElementRegistry contains an entry with name name, then throw a "NotSupportedError DOMException and abort these steps.

  4. If this CustomElementRegistry contains an entry with constructor constructor, then throw a "NotSupportedError" DOMException and abort these steps.

  5. Let localName be name.

  6. Let extends be the value of the extends member of options, or null if no such member exists.

  7. If extends is not null, then:

    1. If extends is a valid custom element name, then throw a "NotSupportedError" DOMException.

    2. If the element interface for extends and the HTML namespace is HTMLUnknownElement (e.g., if extends does not indicate an element definition in this specification), then throw a "NotSupportedError" DOMException.

    3. Set localName to extends.

  8. If this CustomElementRegistry's element definition is running flag is set, then throw a "NotSupportedError" DOMException and abort these steps.

  9. Set this CustomElementRegistry's element definition is running flag.

  10. Run the following substeps while catching any exceptions:

    1. Let prototype be Get(constructor, "prototype"). Rethrow any exceptions.

    2. If Type(prototype) is not Object, then throw a TypeError exception.

    3. Let lifecycleCallbacks be a map with the four keys "connectedCallback", "disconnectedCallback", "adoptedCallback", and "attributeChangedCallback", each of which belongs to an entry whose value is null.

    4. For each of the four keys callbackName in lifecycleCallbacks, in the order listed in the previous step:

      1. Let callbackValue be get(prototype, callbackName). Rethrow any exceptions.

      2. If callbackValue is not undefined, then set the value of the entry in lifecycleCallbacks with key callbackName to the result of converting callbackValue to the Web IDL Function callback type. Rethrow any exceptions from the conversion.

    5. Let observedAttributes be an empty sequence<DOMString>.

    6. If the value of the entry in lifecycleCallbacks with key "attributeChangedCallback" is not null, then:

      1. Let observedAttributesIterable be Get(constructor, "observedAttributes"). Rethrow any exceptions.

      2. If observedAttributesIterable is not undefined, then set observedAttributes to the result of converting observedAttributesIterable to a sequence<DOMString>. Rethrow any exceptions from the conversion.

    Then, perform the following substep, regardless of whether the above steps threw an exception or not:

    1. Unset this CustomElementRegistry's element definition is running flag.

    Finally, if the first set of substeps threw an exception, then rethrow that exception, and terminate this algorithm. Otherwise, continue onward.

  11. Let definition be a new custom element definition with name name, local name localName, constructor constructor, prototype prototype, observed attributes observedAttributes, and lifecycle callbacks lifecycleCallbacks.

  12. Add definition to this CustomElementRegistry.

  13. Let document be this CustomElementRegistry's relevant global object’s associated Document.

  14. Let upgrade candidates be all elements that are shadow-including descendants of document, whose namespace is the HTML namespace and whose local name is localName, in shadow-including tree order. Additionally, if extends is non-null, only include elements whose is value is equal to name.

  15. For each element element in upgrade candidates, enqueue a custom element upgrade reaction given element and definition.

  16. If this CustomElementRegistry's when-defined promise map contains an entry with key name:

    1. Let promise be the value of that entry.

    2. Resolve promise with undefined.

    3. Delete the entry with key name from this CustomElementRegistry's when-defined promise map.

When invoked, the get(name) method must run these steps:

  1. If this CustomElementRegistry contains an entry with name name, then return that entry’s constructor.

  2. Otherwise, return undefined.

When invoked, the {{whenDefined()|whenDefined(name)}} method must run these steps:

  1. If name is not a valid custom element name, then return a new promise rejected with a "SyntaxError" DOMException and abort these steps.

  2. If this CustomElementRegistry contains an entry with name name, then return a new promise resolved with undefined and abort these steps.

  3. Let map be this CustomElementRegistry's when-defined promise map.

  4. If map does not contain an entry with key name, create an entry in map with key name and whose value is a new promise.

  5. Let promise be the value of the entry in map with key name.

  6. Return promise.

The whenDefined() method can be used to avoid performing an action until all appropriate custom elements are defined. In this example, we combine it with the :defined pseudo-class to hide a dynamically-loaded article’s contents until all the autonomous custom elements it uses are defined.

articleContainer.hidden = true;

fetch(articleURL)
.then(response => response.text())
.then(text => {
 articleContainer.innerHTML = text;

 return Promise.all(
   [...articleContainer.querySelectorAll(":not(:defined)")]
     .map(el => customElements.whenDefined(el.localName))
 );
})
.then(() => {
 articleContainer.hidden = false;
});

3.2.4. Upgrades

To upgrade an element, given a custom element definition definition and an element element, run the following steps:

  1. If element is custom, abort these steps.

    This can occur due to reentrant invocation of this algorithm, as in the following example:

    <!DOCTYPE html>
    <x-foo id="a"></x-foo>
    <x-foo id="b"></x-foo>
    
    <script>
    // Defining enqueues upgrade reactions for both "a" and "b"
    customElements.define("x-foo", class extends HTMLElement {
    constructor() {
     super();
    
     const b = document.querySelector("#b");
     b.remove();
    
     // While this constructor is running for "a", "b" is still
     // undefined, and so inserting it into the document will enqueue a
     // second upgrade reaction for "b" in addition to the one enqueued
     // by defining x-foo.
     document.body.appendChild(b);
    }
    })
    </script>
    

    This step will thus bail out the algorithm early when upgrade an element is invoked with "b" a second time.

  2. If element’s custom element state is "failed", then abort these steps.

  3. For each attribute in element’s attribute list, in order, enqueue a custom element callback reaction with element, callback name "attributeChangedCallback", and an argument list containing attribute’s local name, null, attribute’s value, and attribute’s namespace.

  4. If element is connected, then enqueue a custom element callback reaction with element, callback name "connectedCallback", and an empty argument list.

  5. Add element to the end of definition’s construction stack.

  6. Let C be definition’s constructor.

  7. Let constructResult be Construct(C).

    If C non-conformantly uses an API decorated with the [CEReactions] extended attribute, then the reactions enqueued at the beginning of this algorithm will execute during this step, before C finishes and control returns to this algorithm. Otherwise, they will execute after C and the rest of the upgrade process finishes.

  8. Remove the last entry from the end of definition’s construction stack.

    Assuming C calls super() and that the call succeeds, this will be the already constructed marker that replaced the element pushed at the beginning of this algorithm. (The HTML element constructor carries out this replacement.)

    If C does not call super() (i.e. it is not conformant), or if any step in the HTML element constructor throws, then this entry will still be element.

  9. If constructResult is an abrupt completion, then:

    1. Set element’s custom element state to "failed".

    2. Return constructResult (i.e., rethrow the exception), and terminate these steps.

  10. If SameValue(constructResult.[[\value]], element) is false, then throw an "InvalidStateError" DOMException and terminate these steps.

    This can occur if C constructs another instance of the same custom element before calling super(), or if C uses JavaScript’s return-override feature to return an arbitrary object from the constructor.

  11. Set element’s custom element state to "custom".

  12. Set element’s custom element definition to definition.

To try to upgrade an element, given as input an element element, run the following steps:

  1. Let definition be the result of looking up a custom element definition given element’s node document, element’s namespace, element’s local name, and element’s is value.

  2. If definition is not null, then enqueue a custom element upgrade reaction given element and definition.

3.2.5. Custom element reactions

A custom element can react to certain occurrences by running author code. These reactions are called custom element reactions.

The way in which custom element reactions are invoked is done with special care, to avoid running author code during the middle of delicate operations. Effectively, they are delayed until "just before returning to user script". This means that for most purposes they appear to execute synchronously, but in the case of complicated composite operations (like cloning, or <>range manipulation), they will instead be delayed until after all the relevant user agent processing steps have completed, and then run together as a batch.

Additionally, the precise ordering of these reactions is managed by a stack-of-queues system, described below. The intention behind this system is to guarantee that custom element reactions are always invoked in the same order as their triggering actions, at least within the local context of a single custom element. (Because custom element reaction code can perform its own mutations, it is not possible to give a global ordering guarantee across multiple elements.)


Each unit of related similar-origin browsing contexts has a custom element reactions stack, which is initially empty. The current element queue is the element queue at the top of the custom element reactions stack. Each item in the stack is an element queue, which is initially empty as well. Each item in an element queue is an element. (The elements are not necessarily custom yet, since this queue is used for upgrades as well.)

Each custom element reactions stack has an associated backup element queue, which is an initially empty element queue. Elements are pushed onto the backup element queue during operations that affect the DOM without going through an API decorated with [CEReactions], or through the parser’s create an element for the token algorithm. An example of this is a user-initiated editing operation which modifies the descendants or attributes of an editable element. To prevent reentrancy when processing the backup element queue, each custom element reactions stack also has a processing the backup element queue flag, initially unset.

All elements have an associated custom element reaction queue, initially empty. Each item in the custom element reaction queue is of one of two types:

This is all summarised in the following schematic diagram:

@@A custom element reactions stack consists of a stack of element queues. Zooming in on a particular queue, we see that it contains a number of elements (in our example, <x-a>, then <x-b>, then <x-c>). Any particular element in the queue then has a custom element reaction queue. Zooming in on the custom element reaction queue, we see that it contains a variety of queued-up reactions (in our example, upgrade, then attribute changed, then another attribute changed, then connected).

To enqueue an element on the appropriate element queue, given an element element, run the following steps:

  1. If the custom element reactions stack is empty, then:

    1. Add element to the backup element queue.

    2. If the processing the backup element queue flag is set, abort this algorithm.

    3. Set the processing the backup element queue flag.

    4. Queue a microtask to perform the following steps:

      1. Invoke custom element reactions in the backup element queue.

      2. Unset the processing the backup element queue flag.

  2. Otherwise, add element to the current element queue.

To enqueue a custom element callback reaction, given a custom element element, a callback name callbackName, and a list of arguments args, run the following steps:

  1. Let definition be element’s custom element definition.

  2. Let callback be the value of the entry in definition’s lifecycle callbacks with key callbackName.

  3. If callback is null, then abort these steps.

  4. If callbackName is "attributeChangedCallback", then:

    1. Let attributeName be the first element of args.

    2. If definition’s observed attributes does not contain attributeName, then abort these steps.

  5. Add a new callback reaction to element’s custom element reaction queue, with callback function callback and arguments args.

  6. Enqueue an element on the appropriate element queue given element.

To enqueue a custom element upgrade reaction, given an element element and custom element definition definition, run the following steps:

  1. Add a new upgrade reaction to element’s custom element reaction queue, with custom element definition definition.

  2. Enqueue an element on the appropriate element queue given element.

To invoke custom element reactions in an element queue queue, run the following steps:

  1. For each custom element element in queue:

    1. Let reactions be element’s custom element reaction queue.

    2. Repeat until reactions is empty:

      1. Remove the first element of reactions, and let reaction be that element. Switch on reaction’s type:

        upgrade reaction
        Upgrade element using reaction’s custom element definition.
        callback reaction
        Invoke reaction’s callback function with reaction’s arguments, and with element as the callback this value.

        If this throws any exception, then report the exception.


The [CEReactions] IDL extended attribute is to ensure custom element reactions are triggered appropriately. It indicates that the relevant algorithm is to be supplemented with additional steps to appropriately track and invoke custom element reactions.

The [CEReactions] extended attribute must take no arguments, and must not appear on anything other than an operation, attribute, setter, or deleter. Additionally, it must not appear on readonly attributes, unless the readonly attribute is also annotated with [PutForwards].

Operations, attributes, setters, or deleters annotated with the [CEReactions] extended attribute must run the following steps surrounding the main algorithm specified for the operation, setter, deleter, or for the attribute’s setter:

Before executing the algorithm’s steps
Push a new element queue onto the custom element reactions stack.
After executing the algorithm’s steps
Pop the element queue from the custom element reactions stack, and invoke custom element reactions in that queue.
The intent behind this extended attribute is somewhat subtle. One way of accomplishing its goals would be to say that every operation, attribute, setter, and deleter on the platform should have these steps inserted, and to allow implementers to optimize away unnecessary cases (where no DOM mutation is possible that could cause custom element reactions to occur).

However, this could lead to non-interoperable implementations if implementations do not invoke these steps in some cases. Explicitly annotating all relevant IDL constructs helps ensure interoperable behavior, by identifying all cases where these steps are necessary.

Any nonstandard APIs introduced by the user agent that could modify the DOM in such a way as to enqueue a custom element callback reaction or enqueue a custom element upgrade reaction, for example by modifying any attributes or child elements, must also be decorated with the [CEReactions] attribute.

As of the time of this writing, the following nonstandard or not-yet-standardized APIs are known to fall into this category:

3.2.6. HTML: HTML element constructors

To support custom elements, all HTML elements have special constructor behavior. The [HTMLConstructor] IDL extended attribute indicates that the interface object for the given interface will have a specific behavior, as defined in detail below.

The [HTMLConstructor] extended attribute must take no arguments, and must not appear on anything other than an interface. It must appear only once on an interface, and the interface must not be annotated with the [Constructor] or [NoInterfaceObject] extended attributes. (However, the interface may be annotated with [NamedConstructor]; there is no conflict there.) It must not be used on a callback interface.

Interface objects for interfaces annotated with the [HTMLConstructor] extended attribute must run the following steps as the function body behavior for both [[Call]] and [[Construct]] invocations of the corresponding JavaScript function object.

When invoked with [[Call]], newTarget is undefined, and the algorithm will throw. When invoked with [[Construct]], the [[Construct]] newTarget parameter provides the NewTarget value.

  1. Let registry be the current global object’s CustomElementRegistry object.

  2. If NewTarget is equal to the active function object, then throw a TypeError and abort these steps.

    This can occur when a custom element is defined using an element interface as its constructor:

    customElements.define("bad-1", HTMLButtonElement);
    new HTMLButtonElement();          // (1)
    document.createElement("bad-1");  // (2)
    

    In this case, during the execution of HTMLButtonElement (either explicitly, as in (1), or implicitly, as in (2)), both the active function object and NewTarget are HTMLButtonElement. If this check was not present, it would be possible to create an instance of HTMLButtonElement whose local name was bad-1.

  3. Let definition be the entry in registry with constructor equal to NewTarget. If there is no such definition, then throw a TypeError and abort these steps.

    Since there can be no entry in registry whose constructor is undefined, this step also prevents HTML element constructors from being called as functions (since in that case NewTarget will be undefined).

  4. If definition’s local name is equal to definition’s name (i.e., definition is for an autonomous custom element), then:

    1. If the active function object is not HTMLElement, then throw a TypeError and abort these steps.

      This can occur when a custom element is defined to not extend any local names, but inherits from a non-HTMLElement class:

      customElements.define("bad-2", class Bad2 extends HTMLParagraphElement {});
      

      In this case, during the (implicit) super() call that occurs when constructing an instance of Bad2, the active function object is HTMLParagraphElement, not HTMLElement.

  5. Otherwise (i.e., if definition is for a customized built-in element):

    1. Let valid local names be the list of local names for elements defined in this specification or in