This document focuses on providing information on how a language can be designed for forwards compatible versioning, often the hardest type of versioning to plan for. It also provides motivation for versioning and some discourse on incompatible and backwards compatible versioning. Separate documents contain the versioning terminology definitions and XML specific versioning material.
This document has been developed for discussion by the W3C Technical Architecture Group. It does not yet represent the consensus opinion of the TAG.
Publication of this finding 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.
Additional TAG findings, both approved and in draft state, may also be available. The TAG expects to incorporate this and other findings into a Web Architecture Document that will be published according to the process of the W3C Recommendation Track.
1.2 Why Do Languages Change?
1.3 Kinds of Languages
2 Versioning Strategies
2.1 Why Have a Strategy?
2.1.1 Identifying Languages
220.127.116.11 Version Numbers
18.104.22.168 XML Namespaces
4 Backwards compatible
5 Forwards Compatible
5.1 Must Accept Unknowns
5.1.1 Discard or Preserve
5.1.2 Process all or only unknown part
5.2 Fallback Provided
5.3 Forwards Compatibility and Version identifiers
5.3.1 Latest version identifiers
5.3.2 Lowest possible version identifiers
5.4 Supporting functionality
A Change Log (Non-Normative)
The evolution of languages by adding, deleting, or changing syntax or information is called versioning. Making versioning work in practice is one of the most difficult problems in computing. Arguably, the Web rose dramatically in popularity because support for evolution and versioning were built into HTML and HTTP. Both systems provide explicit extensibility points and rules for understanding extensions that enable their decentralized extension and versioning.
This finding describes general problems and techniques in evolving systems in compatible ways. A number of design patterns and rules are discussed with a focus towards enabling language changes such that newer version(s) of a language are processable by software that only understands the older version(s) of the language, aka forwards-compatibility. There are a few crucial good practices that enable forwards compatible versioning in a language:
the language should be extensible;
any such extensions in a text of the language should have a well-defined default meaning (which often is that the extension conveys no information and can be ignored);
if the texts of the language contain version identifiers, then a given language version should define a set of compatible future version identifiers.
[Definition: A Language consists of a set of texts and a mapping between texts and information.]
[Definition: A language is Extensible if the syntax of a language allows information that is not defined in the current version of the language and provides default handling of documents in any extended set.]
[Definition: Every language has a Defined Text set, which contains only Texts that contain the texts explicitly defined by the language syntax constraints. ]Typically, a language will define a mapping from each of the definitions to information conveyed by instances of those definitions. [Definition: Each language has an Accept Text set, which contains texts that are allowed by the language constraints]. Typically, the Accept Text set contains Texts that are not in the Defined Text set.
[Definition: Language L2 is backwards compatible with Language L1 if L2 Accept Text set > (superset) L1 Defined Text Set AND every text in L1 Defined Text set is compatible with L2.]
[Definition: Language L1 is forwards compatible with Language L2 if L1 Accept Text set > (superset) Language L2 Defined Text set AND every text in L2 Defined Text set is compatible with L1.]
When languages are evolved by extending the Defined Text set, forwards and backwards compatibility can be combined for compatibility: [Definition: Language L2 is compatible with Language L1 if L1 Accept Text set > (superset) Language L2 Accept Text set > (superset) L2 Defined Text set > (superset) L1 Defined Text Set AND every text in L1 Defined Text set is compatible with L2 AND every text in L2 Defined Text set is compatible with L1.]
There are many reasons why a language may evolve. A few of them include:
Bugs may need to be fixed. Production use may reveal defects or oversights that need to be fixed. This may involve changes to texts of the language or changes to the information of existing texts.
Changing requirements may motivate changes in the language. For example, a person name structure may be extended with a middle name, prefix, suffix, and/or common name.
Different variations of a language may be desirable. For example, the XHTML 1.0 Recommendation defines strict, transitional, and frameset languages. Although the three forms use markup from the same namespace; strict, transitional and frameset are different languages. Additional languages may be defined by other specifications, such as the XHTML Basic Recommendation.
The specification of the language may be modified to more closely match actual implementation. CSS 2.1 is a good example of this evolution reason because it specifies "But most of all CSS 2.1 represents a 'snapshot' of CSS usage: it consists of all CSS features that are implemented interoperably at the date of publication of the Recommendation."
Whether ten, a hundred, or a million applications have been deployed, if a language is changed in such a way that those applications will determine that texts of the new language are invalid, a versioning problem with real costs has been introduced. Whatever the cause, over time, different versions of the language exist and designing applications using it to deal with the changes in a predictable, useful way will require a versioning strategy.
There are different kinds of languages and the versioning approaches and strategies that are appropriate for one kind of language may not be appropriate for another. Some of the various kinds of languages are:
Just strings: some languages are just set(s) of strings. Strings that identify countries, states or provinces, airport 3 letter codes, and traffic light colors are examples of "just string" languages.
Non-markup Text: languages designed with a character-based text format. These may be programming languages such as Java or ECMAScript, or data formats like CSS or Comma Separated Values. Typically these are intended for humans to author and/or view.
Markup: SGML, HTML, and the non-SGML variants of HTML are all character based markup languages. Versioning XML languages is described in Versioning: XML
binary: languages that are not in a text format. These may be image formats like GIF, JPEG, or even binary encoded XML.
This is by no means an exhaustive or exclusive list. Many languages actually define multiple kinds or formats of languages. For example, XQuery defines a non-XML language and an XML language.
Languages may be composed of sub languages. If a language is composed of other languages, then languages are considered nested. There may be different versioning strategies for each nested language, and they all combine together into the overall versioning strategy.
In broad terms, strategies for versioning fall into a number of classes ranging from "none" to "compatible" to "incompatible":
None.Consumers may only be able to determine whether a text they receive is compatible or incompatible after processing the most or all of the text.
Compatible. Designers are expected to limit changes to those that are backwards and/or forwards compatible. HTML is an example of a language that ensures that all versions are at least somewhat compatible.
Backwards compatible. (Definition) A change in the definition of a language is backward compatible if consumers of the evolved language can correctly process text written for the previous version of the language.. (Discussed in 4 Backwards compatible)
Forwards compatible. (Definition) A change in the definition of a language is forward compatible if consumers of the original language can correctly process text written for the evolved version of the language.(Discussed in 5 Forwards Compatible)
Incompatible. Applications are expected to abort or fail if they see an unexpected version. XML is an example of a language that has an incompatible evolution strategy. (Discussed in 3 Incompatible)
For the incompatible strategy, there are a range of possible failure outcomes:
generate an error and fail or abort.
generate errors and warnings and proceed
generate errors and prompt the user to decide whether to continue or abort.
generate errors and proceed silently. In many cases, this is the equivalent of proceed without failure.
The decision as to which type of error handling is to be used is application dependent. XML 1.0 is an example of where an error results in failure. HTML and CSS specify that most errors are ignored silently resulting in the equivalent of compatible behaviour. Further, the behaviour itself can evolve. Java is an example where features were first deprecated resulting in warnings, and then later removed resulting in failures if those features were used.
There's no single approach to compatibility that's always correct. Different choices or decisions may be appropriate for different applications. But by the same token, the approaches that are available depend on other decisions or constraints. One decision is the whether the language can be evolved by distributed parties such that parallel evolutionary development can occur. The point in the lifecycle of the language may also affect the selection of the versioning strategy for the language. A language commonly goes through a lifecycle of iterative development followed by deployment followed by deployment of new versions. A decision for the development cycle of the language could be different from the decision at the deployment. For example, many W3C languages adopt a strategy of allowing incompatible changes in Working Drafts and up to the Candidate Recommendation publication, but then keeping Proposed Recommendation and Recommendation compatible.
If versioning is not planned from the start, then the possible versioning strategies will be constrained by decisions that have been made implicitly rather than explicitly. Typically, not planning for versioning results in a language where compatible versioning is not available.
Just as there are a number of strategies, there are a number of designs for implementing a strategy. The Internet - including MIME, markup languages, and XML languages has successfully used various strategies, either singly or in combination. Summaries of strategies and requirements were produced for earlier technologies [Web Architecture: Extensible Languages] and guided XML Namespaces and Schema. Additionally, the design can facilitate the use of a particular strategy. For example, XML Namespaces are designed for distributed extensibility and allow some language changes or extensions to be compatible changes rather than incompatible changes.
Different kinds of languages and different versioning strategies expose different problems. Attempting to deploy a system that provides no versioning mechanism may put the burden of version "discovery" on consumers and my be impractical in anything except a closed system.
At the other end of the spectrum is an incompatible versioning approach which is also problematic. "Incompatible" versioning is usually a very coarse-grained approach to versioning. It often requires that all of the text must be understood and known by the consumer or the consumer will fault. It often establishes a single version identifier, such as a version number or namespace name, for an entire text. In many cases, a single version identifier is not an accurate representation of the "version" of the text, particularly when a text is composed of many sub-languages.
The semantics of the incompatible approach are that applications decide on the basis of the version(s) of the language they understand whether they can process that text. If parts or all of the text isn't recognized, perhaps including a version identifier, then the entire text is rejected. Typically, when introducing a new version using the incompatible approach without any provisions for backwards or forwards compatibility, any of the software that produces or consumes the texts cannot successfully communicate until it is updated and deployed. This incompatible approach to versioning is practical only in circumstances where there is a single controlling authority, and even in that case, it carries with it many problems. The process can take a considerable amount of time, leaving parts of the system out of commission for hours if not days. This can result in significant losses if the system is a key component of a revenue generating business process and the cost of coordinating the system overhaul can also be quite costly as well.
The incompatible approach is appropriate when the benefits of introducing the change outweigh the costs, such as when the new version is radically different from its predecessor. But in many cases, changes are incremental and often a consumer could cope with the new version. For example, it might be that there are many texts that don't use any new features from the new version or perhaps the unrecognized parts of the text could be ignored.
Consider a producer and a consumer exchanging texts of a particular language. Imagine that some future version of the language includes new texts in the language. Because producers and consumers are distributed, it may happen that an old consumer, one unprepared for the new text, encounters a newer text sent by a newer producer.
If incompatible versioning is used, old consumers will reject the new text. However, if the versioning strategy allows the old consumer to process the text even with the unrecognized content, it's quite possible that other components of the system would adapt to the previous behavior. In effect, the old system would ignore the new component or provide some other default interpretation for it. For the producer, the result would be that the request is fulfilled, though perhaps not taking into account the new features. For example, a request that results in a response may return a text of the language without the new component. In many cases this may be better behavior than receiving an error. In particular, producers using the new language can be written to cope with the possibility that they will be communicating with older consumers.
If the producer needs to require that the new text is understood, then it can change the language in a way to indicate that the new text is not considered optional, aka incompatible.
Often, what is needed is some sort of middle ground solution where compatible and incompatible changes are supported, and these are discussed in the sections on compatible and incompatible.
As part of a strategy for language design, the language may or may not include version identifiers in the text. One design is that the texts do not contain any version identifiers and all texts are intended to be backwards and forwards-compatible. [CSS] is an example of this strategy as there are no version identifiers in the stylesheets and the language follows the subsequently described rules for ensuring backwards and forward-compatible evolution. There is a benefit of no version identifiers in the text because it forces the language designer(s) to design and plan for compatible evolution as they cannot introduce an incompatible change and "simply" update the version identifier. In fact, the CSS language designers have constrained themselves to only evolve CSS compatibly with previous version(s).
The design of putting version identifiers in the text is very common. It is may be beneficial and necessary for a receiver to determine which version(s) of a language specification can be used to correctly interpret a given text and know this early in the text. This is often done by providing an identifier of the version early in a text, such as a version number or structure such as an XML Namespace.
Regardless of any particular technologies chosen, the language should have a version identification strategy.
Language specifications should specify the meaning of the structure of any version identifiers in texts with respect to compatible or incompatible evolution of the language.
The difficult, important, and often overlooked part of a version identification strategy is specifying the meaning and interpretation when a consumer encounters a version identifier it does not know about. A typical problem with version identifiers is that it is unclear what is being identified with the version identifiers. The common misconception is that a version identifier, such as a number, indicates the version of the language used when producing the document. That may be true, but in many cases the version identifier is unusable for versioning because the language and/or software does not have sufficient provisions for versioning.
There are many possibilities for what a version identifier identifies. For a text that is potentially in many different versions of a language, and assuming versions are ordered, an identifier that is a number could be:
The latest version of the language the document is compatible with
The lowest version of the language the document is compatible with
The lowest version of the language that has all the features the language uses
The set of versions of the language the document is compatible with
Imagine that version 1 of a language for people's names defines a sequence of a first name, a last name, and other extensions; and version 2 defines a sequence of first, last, optional middle, and extensions. If a text contains a first and a last, should the identifier be version 1 or 2? The previous options yield answers of: 2, 1, 1, 1-2. If a name contains a first, a last, and a middle then the previous options yield answers of: 2, 1, 2, 1-2
The usual scenario is the document contains the version number of the latest version of the language that the producing application understands - scenario #1. Thus the "newest" version of the identifier is used, even if the document itself is valid under older versions of the language. This usually works fine if the producer and consumer are at the same version, or even if the consumer understands the older and the newer version. But forwards compatibility requires that a consumer that doesn't understand the newer version must somehow treat the document as if it was an older version. Approaches for using version identifiers to enable forwards compatibility is covered in 5 Forwards Compatible
As a side note, version identifiers are often used in protocols that exchange documents. One scenario is that the version number identifies the latest version of the language the document is compatible with and that the producer will understand when it becomes a consumer of a response document. In this case it is a protocol version identifier, not just a format identifier. HTTP is a good example. The HTTP specification says "The protocol versioning policy is intended to allow the sender to indicate the format of a message and its capacity for understanding further HTTP communication, rather than the features obtained via that communication." Because HTTP is a request-response protocol, the capacity for further HTTP communication is the crucial "version" information conveyed. Most documents that might have version #s do not fit that "future capacity" use case, they are just documents that do not belong to a protocol. There are cases where a document format is combined with a protocol, such as Atom. These combination protocol/format case are fairly rare and not generally applicable.
Version identification is often done with a decimal separating the major versions from the minor versions, i.e. "8.1", "1.0". Often the definition of a "major" change is that it is incompatible, and the definition of a "minor" change is that it is forwards- and/or backwards - compatible. Usually the first broadly available version with a high probability of compatible future versions starts at "1.0". A compatible version change from 1.0 might be identified as "1.1" and an incompatible change as "2.0".
The version numbers can be contained in the texts, in the protocol messages containing the text, or the address for the protocol messages. Some examples are shown below:
<name version="2.0"> <given>Dave</given> <family>Orchard</family> </name> <span class="fn20">Dave Orchard</span> urn:nameschemev2:given:Dave:family:Orchard <?XML version="1.1"?> GET /name/123456789 HTTP/1.1 GET /name/v2/123456789/ HTTP 1.1
As previously mentioned, the version number may be only associated with the language specification is not part of any texts of the language.
It should be noted that associating version number changes with compatibility changes may be idealistic as there abundant cases where this system does not hold. New major version identifiers are often aligned with product releases, or incompatible changes identified as a "minor" change. A good example of an incompatible change that used the traditional minor version identifier change is XML 1.1. XML 1.0 processors cannot process all XML 1.1 documents because XML 1.1 extended XML 1.0 where XML 1.0 does not allow such extension.
Unfortunately, the strategy of texts containing version numbers often results in approaches that are very similar to the incompatible approach. In many approaches, each language is given a version identifier, almost always a number, that's incremented each time the language changes. Although it's possible to design a system with version numbers that enables both backward and forward compatibility - for example XSLT - typically a version change is treated as if that the new language is not backwards compatible with the old language.
Some efforts, such as HTTP, try to have the best of both worlds by allowing for extensibility (in HTTP's case, via headers) as well as version numbers that explicitly identify when a new version is backwards compatible with an old version.
One argument in favor of version numbers is that they allow one to determine what is a 'new version' and what is an 'old version'. But in practice this is not necessarily true. For example, RSS had 0.9x, 1.x, and 2.x versions that were all developed in parallel. In effect the version numbers, even though they appear to be ordered, were simply opaque identifiers. Using version numbers does not guarantee that version 1+x has any particular relationship to version 1.
Version numbers can serve as a useful optimization for consumers. A version number in a text can provide a consumer with an early indication of the compatibility of changes so that the consumer will not process the text any further.
Version numbers typically work best when versioning and extending a language is done in a centralized and linear manner. The makeup of each version can then be consistent and well described.
There are many cases where decentralized and non-linear versioning is desired. The desire for decentralized and non-linear versioning and extensibility was a large motivator for XML and for XML Namespaces. The self-describing and extensible nature of XML markup, and the addition of XML Namespaces, provides a framework for developing languages that can evolve in a decentralized manner. XML Namespaces [ XML Namespaces 1.0] provide a mechanism for associating a URI with an XML element or attribute name, thus specifying the language of the name. This also serves to prevent name collisions.
There are a variety of strategies involving namespaces for version identification. They range from: using a new namespace for every version of the language, even minor or compatible changes; to using a new namespace only for new parts of a language that are compatible additions; to no new namespaces ever. There are costs and benefits to each strategy. For example, using a new namespace for every version of a language often involves breaking any software that is namespace dependent such as XPath expressions. These strategies, including XML and XML Schema samples is provided in [Versioning XML including XML Schema]
As desirable as compatible evolution often is, sometimes a language may not want to allow it. In this model, a consumer will generate a fault if it finds a component it doesn’t understand. An example might be a security specification where a consumer must understand each and every extension. This suffers from the significant drawback that it does not allow compatible changes to occur in the language, as any changes require both consumer and producer to change. Another example is where a producer wants to indicate that an extension must be understood. This nature could be indicated inline using a mustUnderstand model, such as SOAP or an application specific model.
As this finding focuses on compatible versioning, we provide no more focus on incompatible evolution.
Backwards compatible evolution of a language means that consumers of texts in a language should be able to consume texts that producers that are based on an older version of the language will produce. It generally means supporting previous versions of text in a newer consumer. Note that a newer producer may be informed by the consumer as to which versions of a language it prefers. Also, the producer may explicitly provide metadata about the languages it supports, allowing a consumer to deliberately choose which version of the language the producer will produce. This can give the consumer the opportunity to select a version of the language that avoids any new features. There are two significant ways that backward compatibility can be supported.
In the replacement design, a new version of software replaces the old and the new version of the software supports the old and the new version. The producer may or may not need to distinguish between the old and the newer consumer or the texts produced. For example, a web resource that supports additional Name Information as input may not need to change the URI of the resource.
As our definition of backwards compatibility specifies that the newer language's Defined Text Set must be a superset of the older language's Defined Text Set, the typical change is the addition of optional content into the newer version of the language. The older producer simply won't produce texts with the newer content. It is possible to reduce the Defined Text Set by removing items and achieve backwards compatibility, as long as the newer Language's Accept Text set contains all the texts originally in the Defined Text Set. One mechanism to do this is to replace the content with a construct that allows the removed construct. For example, an element can be replaced with a construct (such as an XML schema wildcard) that allows the element.
An example of a replacement strategy is where dereferencing a URI returns an XML document with a person name that contains family and given names and is specified by an XML Schema. Version 2 of the person name adds optional middle names to the name. This new person name language can be deployed as part of the same URI because the V1 and the V2 person name producers can use the same consumer at the previously defined URI.
In the side-by-side design, the new version of the software and the old version of the software are deployed "side-by-side". One variant of the approach is offering both versions of the system, for example by using different URIs for the old and new with a particular focus on enabling older versions of applications to operate on inputs that make use of newer language features. The request to one resource gets mapped to the other resource behind the scenes using a proxy or gateway. This "alternative" approach works when the intermediary can completely handle or generate the new information (for backwards compatibility) or accept the new information (for forwards compatibility). For example, adding an optional element to a data type and deploying to a new URI is supported because a Web server can typically handle mapping the newer URI to the older URI or mapping the older URI to the newer URI. If both URIs are maintained, then the addition is a compatible change. Another example is where new information is required, such as the priority, and the intermediary can apply a default value to provide the required priority. However, this too has its costs as multiple versions of the software must be supported and maintained over time and there is the added cost of developing the proxy or gateway between the two environments. Further, this does not work in scenarios where the intermediary cannot generate the new required content. For example, if a middle name is required in V2, a middle cannot be generated from just a family and a given name.
Forwards compatible evolution of a language means that producers of texts in a language should be able to produce texts in a newer version of the language without consumers having to change existing implementations that know of only the original language. The most common characteristic of a compatible change is the addition of syntax and/or features in a language, usually using the original language's extensibility mechanisms. However, languages may change in a forwards compatible way through the removal of features or syntax. This finding deals with the common case, in which features and/or syntax are added not removed, and in which the goal is to maximize compatibility when making such changes.
A good example of a language designed for extensibility is HTML. The first version of HTML was designed for extensibility; it specified that "unknown markup" may be encountered and is ignored. An example of an extension is the addition of the IMG tag. Note that HTML does not have perfect extensibility, for example, the addition of forms caused form unaware browsers to display text that was often difficult to understand. Ignoring elements but not their descendents worked well for some types of extensions such as the <cite> and <tt> markup tags, but it has not worked well for other extensions such as the <script> tag. The decision of how much to ignore is described in Ignores All or Part. Another example is CSS, which allows extra property values and property names, particularly through vendor prefixes.
An example of a language designed with very little extensibility is XML 1.0. As XML is a meta-language, it can be used to create an unbounded variety of languages made up of elements and attributes. But XML allows for very little extensibility in the XML language itself. It does not allow for new characters in element or attribute names, new punctuation such as different quote characters. This made it very difficult to add new characters to element names for internationalization and other purposes.
The first rule introduced in this Finding relating to extensibility is:
To facilitate independent evolution of producers and consumers, languages in distributed systems should be Extensible.
A language whose syntax allows additional unknown text also requires a specification of what happens when a text contains such additional and unknown text. By the definition of Extensibility, there must be a default rule for interpreting any additional unknown text. If the extensibility is used in a forwards-compatible way, then by definition the software consuming the extension does not know about the extension and we call such extension an unknown extension. The behavior of software when it encounters an unknown extension should be clear.
The simplest rule that enables forwards-compatible changes is to require that a language consumer must accept content that is unknown. This rule is:
Must Accept Unknowns Rule: Consumers MUST accept text portions that they do not recognize where the language has allowed extensibility.
HTML 4.01 follows this approach in "If a user agent encounters an element it does not recognize, it should try to render the element's content". The Must Accept Unknowns rule for XML was first standardized in the WebDAV specification [RFC 2518] section 14 and later separately published as the [Flexible XML Processing Profile]. CSS specifies that unknown or illegal property names and values are ignored and specifies many different scenarios where parts of a style sheet are ignored, through specification text such as "CSS 2.1 reserves for future updates of CSS all property:value combinations and @-keywords that do not contain an identifier beginning with dash or underscore. Implementations must ignore such combinations (other than those introduced by future updates of CSS)."
An extension that affects an existing component in an incompatible way is an incompatible change because a consumer that is unaware of the extension will produce Information from the text that is incompatible with the intended Information. An additional rule is required:
Preserve existing information Rule: An Extensible Language SHOULD require that any texts with extensions SHOULD be processable as a text without the extensions.
An example of an incompatible extension because of a violation of this rule is a purchase order with a payment amount that is in US dollars that is extended by an element that specifies the payment amount is in Euros. An older consumer that ignores the extension element will incorrectly assume that the payment amount is still in US dollars when in fact it is in Euros.
There are 2 further decisions, each with 2 common options. The processing for the unknown content, typically preserve or discard, must be specified. In tree-based languages another decisions is the scope of the processing. Two options being the subtree (ie. an unknown element and all it's descendents) or just the minimal possible portion (ie. an unknown element's start and end tags). These 2 decisions are shown in a simple Venn Diagram:
|Entire Subtree||Discard entire Subtree||Preserve entire Subtree|
|Minimal unknown||Discard minimal unknown||Preserve minimal unknown|
Most forwards-compatible systems refine the Must Accept rule to qualify the handling beyond accepting. One variation is to discard the unknowns:
Must Accept and Discard Unknowns option: Consumers SHOULD accept and discard any text portion that they do not recognize.
This is commonly shortened to "Must Ignore Unknowns Rule". HTML 1, 2 and 3.2 follow the Must Accept and Discard Unknowns option as they specify that any unknown start tags or end tags are mapped to nothing during tokenization.
Discarding content is a simple solution for the default processing rule for unknown content. In order to achieve a compatible evolution, the newer texts of a language must be able to be treated as older texts when the unknown content is discarded.
Another option is to preserve the unknown:
Must Accept and Preserve Unknowns option: Consumers SHOULD accept and preserve any text portion that they do not recognize.
[HTTP 1.1] specifies that a transparent proxy should accept and preserve any headers it doesn't understand: "Unrecognized header fields SHOULD be ignored by the recipient and MUST be forwarded by transparent proxies."
Many languages are consumed in multiple stages, each of which may have different variants of these models. An example is a web browser processing HTML. It will typically tokenize HTML first, and HTML specifies that unknown markup is tokenized to nothing. The results of tokenization are used as the basis for rendering. The same markup is still placed in the DOM and is available for CSS or other DOM related technologies. Another way of looking at this combination is that there are two languages. The browser tokenizer understands HTML which involves ignoring unknown elements or attributes. By our language definitions, the renderer's Defined Text set does not include unknown elements or attributes, though the Accept Text Set does. The browser DOM understands any HTML elements or attributes. The DOM Defined and Accept Text set includes any elements or attributes.
In tree based languages, which includes most markup languages, there are two further variants or options of the Must Accept rules for dealing with extensions. Given the decision of discarding or preserving unknowns, how much of the subtree of the text is to be discarded or ignored.
The rule for accepting the entire subtree is:
Must Accept Unknown and Subtree option: The Must Accept option - preserve or discard - applies to unrecognized texts and any tree descendents.
This variation on must accept requires the consumer to accept and apply the preserve or discard processing to the text and any children it does not understand. Most data applications, such as Web services that use SOAP header blocks or WSDL extensions, adopt this approach to dealing with unexpected markup. For example, if a SOAP message is received with a SOAP header block that contains unrecognized child elements, the child elements and their children must be ignored unless marked as "Must Understand". Note that this rule is not broken if the unrecognized elements are written to a log file. That is, "accepted" or "ignored" doesn’t mean that unrecognized extensions can’t be processed; only that they can’t be the grounds for failure to process.
Other applications may need a different rule as the application may want to preserve as much of the text as possible, perhaps for display purposes.
Must Accept Minimal Unknown option: The Must Accept option - preserve or discard - applies only to the smallest possible portion of the text.
This variation on must accept requires the consumer to accept and discard or preserve the smallest part of the text that is unknown. In many scenarios, the rest of the subtree will continue to be processed. For markup languages, this option could apply to just element or attribute tags that it does not understand. The consumer could then continue to process any element contents. HTML is an example of this option, where the start or end tag is ignored and any children are processed in place of the start or end tag. The Must Accept Minimal Unknown option was described in [HTML 2.0]. This preserves the element descendents in the processing model so that they can still affect interpretation of the text, such as for display purposes.
Each option has different costs and benefits. Choosing to only ignore the minimal unknown helped HTML considerably, but there are some elements whose children also should be ignored for rendering, particularly the Script element. It is possible to design a more complicated system that mixes the two together. For example, HTML could have provided inline syntax such as an ignore-children attribute that allowed an author to specify the element and it's children should be treated with the Must accept all rule. This has various problems too, such as a cumbersome syntax for such additions especially when the extension becomes part of the language. The languages designers could have made the first official version of the Script element have the ignore-children attribute optional but that absence of the attribute meant a default of ignore all children. That would enable the old and the new browsers to correctly process the Script element. Then authors or even the language designers could decide at some point to even remove the attribute from the Script element.
A language can provide mechanisms for explicit fallback if the text is not supported. [MIME] provides multipart/alternative for equivalent, and hence fallback, representations of content. [HTML 4.0] uses this approach in the IMG element where the alt attribute is used if images are not supported or disabled. In XML, the XML Inclusions specification [XInclude] provides a fallback element to handle the case where the putatively included resource cannot be retrieved. There are many variations on where the fallback content can be found. For example, a schema language could specify that fallback content is found in a text, in a schema, or even in the schema for the schema language.
There can be difficulties with fallbacks when the consumer chooses one alternative, modifies it and then a subsequent or different consumer chooses a different alternative. In the multipart/alternative example, an email could be offered as HTML and plain text. If the consumer chooses plain text, replies (becoming a producer) with in-line quoting, then a next consumer chooses to reply to the new message using HTML, then the alternatives are not synchronized and are not true alternatives any more.
Providing forwards compatibility often requires more than a processing model for texts with unknown content, it may also provide a processing model for any version identifiers in the texts of the language. The Good Practice on version identifiers from earlier is:
Language specifications should specify the meaning of the structure of any version identifiers in texts with respect to compatible or incompatible evolution of the language.
Two popular models follow.
As described earlier, one algorithm for version identifiers is the producer creates texts that contain the identifier of latest version of the language that the producer understands. A given text may be multiple languages, but the text will contain the version identifier associated with the most recent language understood by the producer.
A significant problem with most designs of a single version identifier in a text is that it usually doesn't provide for a text to be valid under more than one version. For forwards compatibility in this version identifier strategy ( the latest version identifier known is used by a producer) the version identifier must be within a "space of versions" that a given text is valid under, whether that's a list or regular expression or some other algorithm. In particular, the version identification strategy should specify how unknown versions are dealt with.
When producers use the latest possible version identifier, languages SHOULD provide a default processing model for unknown version identifiers for forwards-compatible evolution.
The default processing model could be an algorithmic approach. For version numbers, one could say that version numbers will only have a "major" change if there is an incompatible change. For example, version 1.1 of a language is by definition compatible with version 1.0 and version 2.0 is incompatible. Then, when the producer puts 1.0, 1.1, or 2.0, a consumer at any level will know whether it can process the content. This also means that there is a choice about which version number to put in, the lowest or the latest. A document that contains "1.1" means that any 1.X processor can process it. A "2.0" document means that a 1.X processor cannot process it, but any "2.X" processor can.
Then the language should have wording about processing unknown version numbers. Sample wording for a default processing model for version identifiers is, "A processor of this version MUST not fault if it receives a document that contains the same major version number." This rule would be in conjunction with forwards-compatible design for the texts, such as "Must Accept Unknowns".
This design has the benefit for the producer that it does not have to understand multiple versions of the language, but it places the burden on the consumer to process unknown version identifiers.
HTML handled unknown version numbers in a forwards compatible way because browsers ignored the version numbers they didn't understand. On the other hand, XML 1.0 did not specify what an XML 1.0 processor should do when an XML document identified as XML 1.1 or XML 2.0 was encountered. As such, many XML 1.0 processors faulted when encountering XML 1.1 documents, whether those documents contained XML 1.1 content or not. Note that XML 1.1 specified that documents should only be specified as XML 1.1 documents if they had XML 1.1 content. Perhaps if XML 1.0 had specified that any document marked XML 1.X should be processable as an XML 1.0 document, in conjunction with other forwards-compatible versioning techniques, then the migration to XML 1.1 would have been easier.
Another algorithm is that the producer uses the lowest possible version that contains all the features of the text. The previous examples showed a version 2 producer that produced a text containing no version 2 features specifying version 1 in the text as well as a version 2 producer that produced a text containing an optional version 2 feature specifying version 1 in the text. These languages do not need the Default Unknown Version Handling Rule because the lowest version identifier possible is used for forwards compatibility, with result that any new version identifier is guaranteed to be incompatible.
When producers use the lowest possible version identifiers, languagues should only change the version identifer for incompatible changes.
This design has the benefit that the consumer does not have to understand unknown identifiers, but the cost is that the producer must be able to generate a potentially large number of different version identifiers based upon the features used and their optionality.
XML 1.1 used this design in an attempt at forwards compatibility when it specified that XML documents that did not use any XML 1.1 features must be identified as XML 1.0 documents. Another example is an XML Language that specifies that any compatible changes will be in existing namespace(s) and any incompatible changes will be in new namespace(s). For XML, this is shown in more detailin Versioning Strategy #3: all new components in existing namespace(s) for each compatible version
Additional functionality can be provided in a language for determining the capabilities of the system that the text is being interpreted in. A language can provide a mechanism for explicit testing, and if so, the language also needs a mechanism for the text to contain a 5.2 Fallback Provided for when the test returns false. The XSLT Specification provides a conditional logic element and a function to test for the existence of extension functions. This allows designers of stylesheets to deal with different consumer capabilities in an explicit fashion.
Languages can choose a mixture of approaches. For example, XSLT provides both an explicit fallback mechanism for some conditions and explicit testing for others. The SOAP specification, another example, specifies Must Accept as the default strategy and the ability to dynamically mark components as Must Understand, aka incompatible.
This Finding is intended to motivate language designers to plan for versioning and extensibility in the languages from the very first version. It details the downsides of ignoring versioning. To help the language designer provide versioning in their language, the finding describes a number of good practices for using in language construction and extension. The main goal of the practices is to help language designers understand their options for language design and how to make backwards- and forwards-compatible changes to their languages should that desirable.
The author thanks Norm Walsh for many contributions as co-editor until 2005. Also thanks the many reviewers that have contributed to the document particularly David Bau, William Cox, Ed Dumbill, Chris Ferris, Yaron Goland, Rhys Lewis, Hal Lockhart, Mark Nottingham, Jeffrey Schlimmer, Cliff Schmidt, and Norman Walsh.
|DBO||20070518||Incorporated Rhys' comments, added version identifier story to forwards compatible evolution, split part 1 into terminology and strategies documents.|
|DBO||20070518||Incorporated WG comments from May f2f which involved many updates to 22.214.171.124, made 1 table for case studies.|
|DBO||20080328||Incorporated WG comments from Feb f2f and Noah's Feb comments which all involved many updates.|