Techniques for User Agent Accessibility Guidelines 1.0

Abstract

This document provides techniques for satisfying the checkpoints defined in "User Agent Accessibility Guidelines 1.0" [UAAG10]. These techniques address key aspects of the accessibility of user interfaces, content rendering, application programming interfaces (APIs), and languages such as the Hypertext Markup Language (HTML), Cascading Style Sheets (CSS) and the Synchronized Multimedia Integration Language (SMIL).

The techniques listed in this document are not required for conformance to the Guidelines. These techniques are not necessarily the only way of satisfying the checkpoint, nor are they a definitive set of requirements for satisfying a checkpoint.

Status of this document

This section describes the status of this document at the time of its publication. Other documents may supersede this document. The latest status of this document series is maintained at the W3C.

This is the 16 October 2002 Working Draft of "Techniques for User Agent Accessibility Guidelines 1.0". It is a draft document and may be updated, replaced or obsoleted by other documents at any time. It is inappropriate to use W3C Working Drafts as reference material or to cite them as other than "work in progress". This is work in progress and does not imply endorsement by, or the consensus of, either W3C or participants in the User Agent Accessibility Guidelines Working Group (UAWG).

While User Agent Accessibility Guidelines 1.0 strives to be a stable document (as a W3C Recommendation), the current document is expected to evolve as technologies change and content developers discover more effective techniques for designing accessible Web sites and pages.

A list of changes to this document is available.

The latest information regarding patent disclosures related to this document is available on the Web. As of this publication, there are no disclosures.

Please send comments about this document, including suggestions for additional techniques, to the public mailing list w3c-wai-ua@w3.org; public archives are available.

This document is part of a series of accessibility documents published by the Web Accessibility Initiative (WAI) of the World Wide Web Consortium (W3C). WAI Accessibility Guidelines are produced as part of the WAI Technical Activity. The goals of the User Agent Accessibility Guidelines Working Group are described in the charter.

A list of current W3C Recommendations and other technical documents can be found at the W3C Web site.

1 Introduction

This document discusses implementation details that should be helpful to understanding how to satisfy the requirements of "User Agent Accessibility Guidelines 1.0" [UAAG10]. This document includes:

• This section provides some context for using this document.
• Section 2 lists each checkpoint of "User Agent Accessibility Guidelines 1.0" along with some possible techniques for satisfying it.
• Section 3 discusses general topics related to the implementation of accessibility features in user agents.

Differences from User Agent Accessibility Guidelines 1.0

In an effort to improve the readability of this document, some information from User Agent Accessibility Guidelines 1.0 has been copied here:

• each checkpoint (including the Notes following the checkpoints);
• the priority definitions;
• the glossary;
• the acknowledgments.

In an effort to reduce the size of the current document, some information that is in User Agent Accessibility Guidelines 1.0 has not been copied here:

• the introduction;
• the descriptions of how the guidelines and checkpoints are structured and organized;
• the prose of each guideline (i.e., the text after the guideline title and before the list of checkpoints);
• the conformance section (since one does not conform to the current document, only to User Agent Accessibility Guidelines 1.0).

The current document includes a list of references resources that is not part of User Agent Accessibility Guidelines 1.0.

Related resources

"Techniques for User Agent Accessibility Guidelines 1.0" and the "User Agent Accessibility Guidelines 1.0" are part of a series of accessibility guidelines published by the Web Accessibility Initiative (WAI). These documents explain the responsibilities of user agent developers in making the Web more accessible to users with disabilities. The series also includes the "Web Content Accessibility Guidelines 1.0" [WCAG10] (and techniques [WCAG10-TECHS]), which explain the responsibilities of authors, and the "Authoring Tool Accessibility Guidelines 1.0" [ATAG10] (and techniques [ATAG10-TECHS]), which explain the responsibilities of authoring tool developers.

2 The user agent accessibility guidelines

This section lists each checkpoint of "User Agent Accessibility Guidelines 1.0" [UAAG10] along with some possible techniques for satisfying it. Each checkpoint definition includes a link to the checkpoint definition in "User Agent Accessibility Guidelines 1.0". Each checkpoint definition is followed by one or more of the following:

• Notes and rationale: Additional rationale and explanation of the checkpoint;
• Who benefits: Which users with disabilities are expected to benefit from user agents that satisfy the checkpoint;
• Example techniques: Some techniques to illustrate how a user agent might satisfy the requirements of the checkpoint. Screen shots and other information about deployed user agents have been included as sample techniques. References to products are not endorsements of those products by W3C;
• Doing more: Techniques to achieve more than what is required by the checkpoint;
• Related techniques: Links to other techniques in section 3. The accessibility topics of section 3 generally apply to more than one checkpoint.
• References: References to other guidelines, specifications, or resources.

Note: Most of the techniques in this document are designed for graphical browsers and multimedia players running on desktop computers. However, some of them also make sense for assistive technologies and other user agents. In particular, techniques about communication between user agents will benefit assistive technologies. Refer, for example, to the appendix on loading assistive technologies for access to the document object model.

Priorities

Each checkpoint in this document is assigned a priority that indicates its importance for users with disabilities.

Priority 1 (P1)

If the user agent does not satisfy this checkpoint, one or more groups of users with disabilities will find it impossible to access the Web. Satisfying this checkpoint is a basic requirement for enabling some people to access the Web.

Priority 2 (P2)

If the user agent does not satisfy this checkpoint, one or more groups of users with disabilities will find it difficult to access the Web. Satisfying this checkpoint will remove significant barriers to Web access for some people.

Priority 3 (P3)

If the user agent satisfies this checkpoint, one or more groups of users with disabilities will find it easier to access the Web.

Note: This information about checkpoint priorities is included for convenience only. For detailed information about conformance to "User Agent Accessibility Guidelines 1.0" [UAAG10], please refer to that document.

Checkpoint definitions

1.1 Full keyboard access. (P1) Checkpoint 1.1

1. Ensure that the user can operate through keyboard input alone any user agent functionality available through the user interface.

Note: For example, ensure that the user can interact with enabled elements, select content, navigate viewports, configure the user agent, access documentation, install the user agent, and operate user interface controls, all entirely through keyboard input.

User agents generally support at least three types of keyboard operation:

1. Direct (e.g., keyboard shortcuts such a "F1" to open the help menu; see checkpoint 11.4 for single-key access requirements),
2. Sequential (e.g., navigation through cascading menus), and
3. Spatial (e.g., when the keyboard is used to move the pointing device in two-dimensional visual space to manipulate a bitmap image).

User agents should support direct or sequential keyboard operation for all functionalities. Furthermore, the user agent should satisfy this checkpoint by offering a combination of keyboard-operable user interface controls (e.g., keyboard operable print menus and settings) and direct keyboard shortcuts (e.g., to print the current page).

It is also possible to claim conformance to User Agent Accessibility Guidelines 1.0 [UAAG10] for full support through pointing device input and/or voice input. See the section on Input modality labels in UAAG 1.0.

1.2 Activate event handlers. (P1) Checkpoint 1.2

1. Allow the user to activate, through keyboard input alone, all input device event handlers that are explicitly associated with the element designated by the content focus.
2. In order to satisfy provision one of this checkpoint, the user must be able to activate as a group all event handlers of the same input device event type.

Note: Refer to the checkpoints of guideline 9 for more information about focus requirements.

1.3 Provide text messages. (P1) Checkpoint 1.3

1. Ensure that every message (e.g., prompt, alert, or notification) that is a non-text element and is part of the user agent user interface has a text equivalent.

Note: For example, if the user is alerted of an event by an audio cue, a visually-rendered text equivalent in the status bar could satisfy this checkpoint. Per checkpoint 6.5, a text equivalent for each such message must be available through an API. See also checkpoint 6.6 for requirements for programmatic notification of changes to the user interface.

Checkpoint definitions

2.1 Render content according to specification. (P1) Checkpoint 2.1

1. Render content according to format specification (e.g., for a markup language or style sheet language).

Note: If a conforming user agent does not render a content type, it should allow the user to choose a way to handle that content (e.g., by launching another application or by saving it to disk).

2.2 Provide text view. (P1) Checkpoint 2.2

1. For content authored in text formats, provide a view of the text source.

2.3 Render conditional content. (P1) Checkpoint 2.3

1. Allow configuration to provide access to each piece of unrendered conditional content "C".
2. When a specification does not explain how to provide access to this content, do so as follows:
   • If C is a summary, title, alternative, description, or expansion of another piece of content D, provide access through at least one of the following mechanisms:
     • (1a) render C in place of D;
     • (2a) render C in addition to D;
     • (3a) provide access to C by allowing the user to query D. In this case, the user agent must also alert the user, on a per-element basis, to the existence of C (so that the user knows to query D);
     • (4a) allow the user to follow a link to C from the context of D.
   • Otherwise, provide access to C through at least one of the following mechanisms:
     • (1b) render a placeholder for C, and allow the user to view the original author-supplied content associated with each placeholder;
     • (2b) provide access to C by query (e.g., allow the user to query an element for its attributes). In this case, the user agent must also alert the user, on a per-element basis, to the existence of C;
     • (3b) allow the user to follow a link in context to C.

Note: For instance, an HTML user agent might allow users to query each element for access to conditional content supplied for the "alt", "title", and "longdesc" attributes. Or, the user agent might allow configuration so that the value of the "alt" attribute is rendered in place of all IMG elements (while other conditional content might be made available through another mechanism).

2.4 Allow time-independent interaction. (P1) Checkpoint 2.4

1. For rendered content where user input is only possible within a finite time interval controlled by the user agent, allow configuration to provide a view where user interaction is time-independent.

Note: If the user agent satisfies this checkpoint by pausing automatically, it may be necessary to pause more than once when there are multiple opportunities for time-sensitive user interaction. When pausing, pause synchronized content as well (whether rendered in the same or different viewports) per checkpoint 2.6. In SMIL 1.0 [SMIL], for example, the "begin", "end", and "dur" attributes synchronize presentation components. See also checkpoint 3.5, which involves client-driven content retrieval.

2.5 Make captions, transcripts, audio descriptions available. (P1) Checkpoint 2.5

1. Allow configuration or control to render text transcripts, collated text transcripts, captions, and audio descriptions in content at the same time as the associated audio tracks and visual tracks.

2.6 Respect synchronization cues. (P1) Checkpoint 2.6

1. Respect synchronization cues (e.g., in markup) during rendering.

2.7 Repair missing content. (P2) Checkpoint 2.7

1. Allow configuration to generate repair text when the user agent recognizes that the author has failed to provide conditional content that was required by the format specification.

Note: Some markup languages (such as HTML 4 [HTML4] and SMIL 1.0 [SMIL] require the author to provide conditional content for some elements (e.g., the "alt" attribute on the IMG element).

2.8 No repair text. (P3) Checkpoint 2.8

1. Allow at least two configurations for when the user agent recognizes that conditional content required by the format specification is present but empty content:
   • generate no repair text, or
   • generate repair as described in checkpoint 2.7.

Note: In some authoring scenarios, empty content (e.g., alt="" in HTML) may make an appropriate text equivalent, such as when non-text content has no other function than pure decoration, or when an image is part of a "mosaic" of several images and does not make sense out of the mosaic. Refer to the Web Content Accessibility Guidelines 1.0 [WCAG10] for more information about text equivalents.

2.9 Render conditional content automatically. (P3) Checkpoint 2.9

1. Allow configuration to render all conditional content automatically.
2. As part of satisfying provision one of this checkpoint, provide access according to specification, or where unspecified, by applying one of the techniques 1a, 2a, or 1b defined in checkpoint 2.3.

Note: For instance, an HTML user agent might allow configuration so that the value of the "alt" attribute is rendered in place of all IMG elements (while other conditional content might be made available through another mechanism). The user agent may offer multiple configurations (e.g., a first configuration to render one type of conditional content automatically and a second to render another type).

2.10 Don't render text in unsupported writing systems. (P3) Checkpoint 2.10

1. For graphical user agents, allow configuration not to render text in unsupported scripts (i.e., writing systems) when that text would otherwise be rendered.
2. When configured per provision one of this checkpoint, indicate to the user in context that author-supplied content has not been rendered due to lack of support for a writing system.

Note: This checkpoint is designed primarily to benefit users with serial access to content or who navigate sequentially, allowing them to skip portions of content that would be unusable if rendered graphically as "garbage".

In addition to the techniques below, refer also to the section on user control of style.

Checkpoint definitions

3.1 Toggle background images. (P1) Checkpoint 3.1

1. Allow configuration not to render background image content.

Note: When background images are not rendered, they are considered conditional content. See checkpoint 2.3 for information about providing access to conditional content.

3.2 Toggle audio, video, animated images. (P1) Checkpoint 3.2

1. Allow configuration not to render audio, video, or animated image content, except on explicit user request.

Note: See guideline 4 for additional requirements related to the control of rendered audio, video, and animated images. When these content types are not rendered, they are considered conditional content. See checkpoint 2.3 for information about providing access to conditional content.

3.3 Toggle animated or blinking text. (P1) Checkpoint 3.3

1. Allow configuration to render animated or blinking text content as motionless, unblinking text. Blinking text is text whose visual rendering alternates between visible and invisible, at any rate of change.

Note: Animation (a rendering effect) differs from streaming (a delivery mechanism). Streaming content might be rendered as an animation (e.g., an animated stock ticker or vertically scrolling text) or as static text (e.g., movie subtitles, which are rendered for a limited time, but do not give the impression of movement).

3.4 Toggle scripts. (P1) Checkpoint 3.4

1. Allow configuration not to execute any executable content (e.g., scripts and applets).

Note: Scripts and applets may provide very useful functionality, not all of which causes accessibility problems. Developers should not consider that the user's ability to turn off scripts is an effective way to improve content accessibility; turning off scripts means losing the benefits they offer. Instead, developers should provide users with finer control over user agent or content behavior known to raise accessibility barriers. The user should only have to turn off scripts as a last resort.

3.5 Toggle automatic content retrieval. (P1) Checkpoint 3.5

1. Allow configuration so that the user agent only retrieves content on explicit user request.

Note: For example, if the user agent supports automatic content retrieval, to ensure that the user does not become disoriented by sudden automatic changes, allow configurations such as "Never retrieve content automatically" and "Require confirmation before content retrieval."

3.6 Toggle images. (P2) Checkpoint 3.6

1. Allow configuration not to render image content.

Note: When images are not rendered, they are considered conditional content. See checkpoint 2.3 for information about providing access to conditional content.

In addition to the techniques below, refer also to the section on user control of style.

Checkpoint definitions for visually rendered text

4.1 Configure text scale. (P1) Checkpoint 4.1

1. Allow global configuration of the scale of visually rendered text content. Preserve distinctions in the size of rendered text as the user increases or decreases the scale.
2. As part of satisfying provision one of this checkpoint, provide a configuration option to override rendered text sizes specified by the author or user agent defaults.
3. As part of satisfying provision one of this checkpoint, offer a range of text sizes to the user that includes at least:
   • the range offered by the conventional utility available in the operating environment that allows users to choose the text size (e.g., the font size), or
   • if no such utility is available, the range of text sizes supported by the conventional APIs of the operating environment for drawing text.

4.2 Configure font family. (P1) Checkpoint 4.2

1. Allow global configuration of the font family of all visually rendered text content.
2. As part of satisfying provision one of this checkpoint, provide a configuration option to override font families specified by the author or by user agent defaults.
3. As part of satisfying provision one of this checkpoint, offer a range of font families to the user that includes at least:
   • the range offered by the conventional utility available in the operating environment that allows users to choose the font family, or
   • if no such utility is available, the range of font families supported by the conventional APIs of the operating environment for drawing text.

Note: For example, allow the user to specify that all text is to be rendered in a particular sans-serif font family.

4.3 Configure text colors. (P1) Checkpoint 4.3

1. Allow global configuration of the foreground and background color of all visually rendered text content.
2. As part of satisfying provision one of this checkpoint, provide a configuration option to override foreground and background colors specified by the author or user agent defaults.
3. As part of satisfying provision one of this checkpoint, offer a range of colors to the user that includes at least:
   • the range offered by the conventional utility available in the operating environment that allows users to choose colors, or
   • if no such utility is available, the range of colors supported by the conventional APIs of the operating environment for specifying colors.

Note: User configuration of foreground and background colors may inadvertently lead to the inability to distinguish ordinary text from selected text or focused text. See checkpoint 10.2 for more information about highlight styles.

Checkpoint definitions for multimedia presentations and other presentations that change continuously over time

4.4 Slow multimedia. (P1) Checkpoint 4.4

1. Allow the user to slow the presentation rate of rendered audio and animation content (including video and animated images).
2. As part of satisfying provision one of this checkpoint, for a visual track, provide at least one setting between 40% and 60% of the original speed.
3. As part of satisfying provision one of this checkpoint, for a prerecorded audio track including audio-only presentations, provide at least one setting between 75% and 80% of the original speed.
4. When the user agent allows the user to slow the visual track of a synchronized multimedia presentation to between 100% and 80% of its original speed, synchronize the visual and audio tracks (per checkpoint 2.6). Below 80%, the user agent is not required to render the audio track.

Note: The style exception of this checkpoint is based on the assumption that authors have satisfied the requirements of the "Web Content Accessibility Guidelines 1.0" [WCAG10] not to convey information through style alone (e.g., through color alone or style sheets alone).

4.5 Start, stop, pause, and navigate multimedia. (P1) Checkpoint 4.5

1. Allow the user to stop, pause, and resume rendered audio and animation content (including video and animated images) that last three or more seconds at their default playback rate.
2. Allow the user to navigate efficiently within audio and animations (including video and animated images) that last three or more seconds at their default playback rate.

Note: The lower bound of three seconds is part of this checkpoint since control is not required for brief audio and animation content, such as short clips or beeps. Respect synchronization cues per checkpoint 2.6.

4.6 Do not obscure captions. (P1) Checkpoint 4.6

1. For graphical viewports, allow configuration so that captions synchronized with a visual track in content are not obscured by it.

Checkpoint definitions for audio volume control

4.7 Global volume control. (P1) Checkpoint 4.7

1. Allow global configuration of the volume of all rendered audio, with an option to override audio volumes specified by the author or user agent defaults.
2. As part of satisfying provision one of this checkpoint, allow the user to choose zero volume (i.e., silent).

Note: User agents should allow configuration of volume through available operating environment mechanisms.

4.8 Independent volume control. (P1) Checkpoint 4.8

1. Allow independent control of the volumes of rendered audio content synchronized to play simultaneously.

Note: The user agent should satisfy this checkpoint by allowing the user to control independently the volumes of all audio sources (e.g., by implementing a general audio mixer type of functionality). See checkpoint 4.10 for information about controlling the volume of synthesized speech.

Checkpoint definitions for synthesized speech rendering

See also techniques for synthesized speech rendering.

4.9 Configure synthesized speech rate. (P1) Checkpoint 4.9

1. Allow configuration of the synthesized speech rate, according to the full range offered by the speech synthesizer.

Note: The range of synthesized speech rates offered by the speech synthesizer may depend on natural language.

4.10 Configure synthesized speech volume. (P1) Checkpoint 4.10

1. Allow control of the synthesized speech volume, independent of other sources of audio.

Note: See checkpoint 4.8 for information about independent volume control of different sources of audio.

4.11 Configure synthesized speech characteristics. (P1) Checkpoint 4.11

1. Allow configuration of synthesized speech characteristics according to the full range of values offered by the speech synthesizer.

Note: Some speech synthesizers allow users to choose values for synthesized speech characteristics at a higher abstraction layer, i.e., by choosing from present options that group several characteristics. Some typical options one might encounter include: "voice (e.g., "adult male voice", "female child voice", "robot voice"), "pitch", and "stress". Ranges for values may vary among speech synthesizers.

4.12 Specific synthesized speech characteristics. (P2) Checkpoint 4.12

1. Allow configuration of synthesized speech pitch. Pitch refers to the average frequency of the speaking voice.
2. Allow configuration of synthesized speech pitch range. Pitch range specifies a variation in average frequency.
3. Allow configuration of synthesized speech stress. Stress refers to the height of "local peaks" in the intonation contour of the voice.
4. Allow configuration of synthesized speech richness. Richness refers to the richness or brightness of the voice.

Note: This checkpoint is more specific than checkpoint 4.11. It requires support for the voice characteristics listed in the provisions of this checkpoint. Definitions for these characteristics are based on descriptions in section 19 of the Cascading Style Sheets Level 2 Recommendation [CSS2]; refer to that specification for additional informative descriptions. Some speech synthesizers allow users to choose values for synthesized speech characteristics at a higher abstraction layer, for example, by choosing from present options distinguished by "gender", "age", or "accent." Ranges of values may vary among speech synthesizers.

4.13 Configure synthesized speech features. (P2) Checkpoint 4.13

1. Provide support for user-defined extensions to the synthesized speech dictionary.
2. Provide support for spell-out: where text is spelled one character at a time, or according to language-dependent pronunciation rules.
3. Allow at least two configurations for speaking numerals: one where numerals are spoken as individual digits, and one where full numbers are spoken.
4. Allow at least two configurations for speaking punctuation: one where punctuation is spoken literally, and one where punctuation is rendered as natural pauses.

Note: Definitions for the functionalities listed in the provisions of this checkpoint are based on descriptions in section 19 of the Cascading Style Sheets Level 2 Recommendation [CSS2]; refer to that specification for additional informative descriptions.

Checkpoint definitions related to style sheets

4.14 Choose style sheets. (P1) Checkpoint 4.14

1. Allow the user to choose from and apply alternative author style sheets (such as linked style sheets).
2. Allow the user to choose from and apply at least one user style sheet.
3. Allow the user to turn off (i.e., ignore) author and user style sheets.

Note: By definition, the user agent's default style sheet is always present, but may be overridden by author or user styles. Developers should not consider that the user's ability to turn off author and user style sheets is an effective way to improve content accessibility; turning off style sheet support means losing the many benefits they offer. Instead, developers should provide users with finer control over user agent or content behavior known to raise accessibility barriers. The user should only have to turn off author and user style sheets as a last resort.

Checkpoint definitions

5.1 No automatic content focus change. (P2) Checkpoint 5.1

1. Allow configuration so that if a viewport opens without explicit user request, neither its content focus nor its user interface focus automatically becomes the current focus.

5.2 Keep viewport on top. (P2) Checkpoint 5.2

1. For graphical user interfaces, allow configuration so that the viewport with the current focus remains "on top" of all other viewports with which it overlaps.

5.3 Manual viewport open only. (P2) Checkpoint 5.3

1. Allow configuration so that viewports only open on explicit user request.
2. When configured per provision one of this checkpoint, instead of opening a viewport automatically, alert the user and allow the user to open it with an explicit request (e.g., by confirming a prompt or following a link generated by the user agent).
3. Allow the user to close viewports.

Note: Generally, viewports open automatically as the result of instructions in content. See also checkpoint 5.1 (for control over changes of focus when a viewport opens) and checkpoint 6.6 (for programmatic notification of changes to the user interface).

5.4 Selection and focus in viewport. (P2) Checkpoint 5.4

1. Ensure that when a viewport's selection or content focus changes, it is at least partially in the viewport after the change.

Note: For example, if users navigating links move to a portion of the document outside a graphical viewport, the viewport should scroll to include the new location of the focus. Or, for users of audio viewports, allow configuration to render the selection or focus immediately after the change.

5.5 Confirm form submission. (P2) Checkpoint 5.5

1. Allow configuration to prompt the user to confirm (or cancel) any form submission.

Note: Examples of automatic form submission include: script-driven submission when the user changes the state of a particular form control associated with the form (e.g., via the pointing device), submission when all fields of a form have been filled out, and submission when a "mouseover" or "change" event occurs.

Checkpoint definitions

6.1 Programmatic access to HTML/XML infoset. (P1) Checkpoint 6.1

1. Provide programmatic read access to XML content by making available all of the information items defined by the W3C XML Infoset [INFOSET].
2. Provide programmatic read access to HTML content by making available all of the following information items defined by the W3C XML Infoset [INFOSET]:
   • Document Information item: children, document element, base URI, charset
   • Element Information items: element-type name, children, attributes, parent
   • Attribute Information items: attribute-type name, normalized value, specified, attribute type, references, owner element
   • Character Information items: character code, parent element
   • Comment Information items: content, parent
3. If the user can modify the state or value of a piece of HTML or XML content through the user interface (e.g., by checking a box or editing a text area), allow programmatic read access to the current state or value, and allow the same degree of write access programmatically as is available through the user interface.

6.2 DOM access to HTML/XML content. (P1) Checkpoint 6.2

1. Provide access to the content required in checkpoint 6.1 by conforming to the following modules of the W3C Document Object Model (DOM) Level 2 Core Specification [DOM2CORE] and exporting bindings for the interfaces they define:
   • for HTML: the Core module.
   • for XML: the Core and XML modules.
2. As part of satisfying provision one of this checkpoint,
   • Export the normative bindings specified in the DOM Level 2 Core Specification [DOM2CORE] (namely, for Java [JAVA] and ECMAScript [ECMASCRIPT] operating environments).
   • For other environments, the bindings exported to satisfy provision one of this checkpoint (e.g., C++ bindings) must be publicly documented.

Note: Refer to the "Document Object Model (DOM) Level 2 Core Specification" [DOM2CORE] for information about HTML and XML versions covered. This checkpoint stands apart from checkpoint 6.1 to emphasize the distinction between what information is required and how to provide access to that information. Furthermore, the DOM Level 2 Core Specification does not provide access to current states and values referred to in provision three of checkpoint checkpoint 6.1. For HTML content, the DOM HTML interfaces do provide access to current states and values.

6.3 Programmatic access to non-HTML/XML content. (P1) Checkpoint 6.3

1. For content other than HTML and XML, provide structured programmatic read access to content.
2. If the user can modify the state or value of a piece of non-HTML/XML content through the user interface (e.g., by checking a box or editing a text area), allow programmatic read access to the current state or value, and allow the same degree of write access programmatically as is available through the user interface.
3. As part of satisfying provision one of this checkpoint, implement at least one API according to this API cascade:
   • The API is defined by a W3C Recommendation, or the API is publicly documented and designed to enable interoperability with assistive technologies.
   • If no such API is available, or if available APIs do not enable the user agent to satisfy the requirements,
     • implement at least one publicly documented API to satisfy the requirements, and
     • follow operating environment conventions for the use of input and output APIs.

Note: This checkpoint addresses content not covered by checkpoint 6.1 and checkpoint 6.2.

6.4 Programmatic access to information about rendered content. (P1) Checkpoint 6.4

1. For graphical user agents, make available bounding dimensions and coordinates of rendered graphical objects. Coordinates must be relative to the point of origin in the graphical environment (e.g., with respect to the desktop), not the viewport.
2. For graphical user agents, provide access to the following information about each piece of rendered text: font family, font size, and foreground and background colors.
3. As part of satisfying provisions one and two of this checkpoint, implement at least one API according to the API cascade described in provision two of checkpoint 6.3.

Note: User agents should provide programmatic access to additional useful information about rendered content that is not available through the APIs required by checkpoints 6.2 and 6.3, including the correspondence (in both directions) between graphical objects and their source in the document object, and information about the role of each graphical object.

6.5 Programmatic operation of user agent user interface. (P1) Checkpoint 6.5

1. Provide programmatic read access to user agent user interface controls, selection, content focus, and user interface focus.
2. If the user can modify the state or value of a user agent user interface control (e.g., by checking a box or editing a text area), allow programmatic read access to the current state or value, and allow the same degree of write access programmatically as is available through the user interface.
3. As part of satisfying provisions one and two of this checkpoint, implement at least one API according to the API cascade described in provision two of checkpoint 6.3.

Note: APIs used to satisfy the requirements of this checkpoint may vary. For instance, they may be independent of a particular operating environment (e.g., the W3C DOM), conventional APIs for a particular operating environment, conventional APIs for programming languages, plug-ins, or virtual machine environments. User agent developers are encouraged to implement APIs that allow assistive technologies to interoperate with multiple types of software in a given operating environment (e.g., user agents, word processors, and spreadsheet programs), as this reuse will benefit users and assistive technology developers. User agents should always follow operating environment conventions for the use of input and output APIs.

6.6 Programmatic notification of changes. (P1) Checkpoint 6.6

1. Provide programmatic notification of changes to content, states and values of content, user agent user interface controls, selection, content focus, and user interface focus.
2. As part of satisfying provision one of this checkpoint, implement at least one API according to the API cascade of provision two of checkpoint 6.3.

Note: For instance, provide programmatic notification when user interaction in one frame causes automatic changes to content in another.

6.7 Conventional keyboard APIs. (P1) Checkpoint 6.7

1. Implement APIs for the keyboard as follows:
   • Follow operating environment conventions.
   • If no conventions exist, implement publicly documented APIs.

Note: An operating environment may define more than one conventional API for the keyboard. For instance, for Japanese and Chinese, input may be processed in two stages, with an API for each.

6.8 API character encodings. (P1) Checkpoint 6.8

1. For an API implemented to satisfy requirements of this document, support the character encodings required for that API.

Note: Support for character encodings is an important part of ensuring that text is correctly communicated to assistive technologies. For example, the DOM Level 2 Core Specification [DOM2CORE], section 1.1.5 requires that the DOMString type be encoded using UTF-16.

6.9 DOM access to CSS style sheets. (P2) Checkpoint 6.9

1. For user agents that implement Cascading Style Sheets (CSS), provide programmatic access to style sheets by conforming to the CSS module of the W3C Document Object Model (DOM) Level 2 Style Specification [DOM2STYLE] and exporting bindings for the interfaces it defines.
2. As part of satisfying provision one of this checkpoint:
   • Export the normative bindings specified in the CSS module of the DOM Level 2 Style Specification [DOM2STYLE] (namely, for Java [JAVA] and ECMAScript [ECMASCRIPT] operating environments).
   • For other environments, the bindings exported to satisfy provision one of this checkpoint must be publicly documented.

Note: Refer to the "Document Object Model (DOM) Level 2 Style Specification" [DOM2STYLE] for information about CSS versions covered.

6.10 Timely exchanges through APIs. (P2) Checkpoint 6.10

1. For APIs implemented to satisfy the requirements of this document, ensure that programmatic exchanges proceed in a timely manner.

Note: For example, the programmatic exchange of information required by other checkpoints in this document should be efficient enough to prevent information loss, a risk when changes to content or user interface occur more quickly than the communication of those changes. Timely exchange is also important for the proper synchronization of alternative renderings. The techniques for this checkpoint explain how developers can reduce communication delays. This will help ensure that assistive technologies have timely access to the document object model and other information that is important for providing access.

Checkpoint definitions

7.1 Respect focus and selection conventions. (P1) Checkpoint 7.1

1. Follow operating environment conventions that benefit accessibility when implementing the selection, content focus, and user interface focus.

Note: See checkpoints 9.1 and 9.2 for more information about content focus and user interface focus.

7.2 Respect input configuration conventions. (P1) Checkpoint 7.2

1. Ensure that default input configurations of the user agent do not interfere with operating environment accessibility conventions (e.g., for keyboard accessibility).

Note: See checkpoint 11.5 for information about the user agent's default input configuration.

7.3 Respect operating environment conventions. (P2) Checkpoint 7.3

1. Follow operating environment conventions that benefit accessibility. In particular, follow conventions that benefit accessibility for user interface design, keyboard configuration, product installation, and documentation.

7.4 Provide input configuration indications. (P2) Checkpoint 7.4

1. Follow operating environment conventions to indicate the input configuration.

Note: For example, in some operating environments, when a functionality may be triggered through a menu and through the keyboard, the developer may design the menu entry so that the character of the activating key is also shown. See checkpoint 11.5 for information about the user agent's default input configuration.

Checkpoint definitions

8.1 Implement accessibility features. (P1) Checkpoint 8.1

1. Implement the accessibility features of specifications (e.g., markup languages, style sheet languages, metadata languages, and graphics formats).

8.2 Conform to specifications. (P2) Checkpoint 8.2

1. Use and conform to either
   • W3C Recommendations when they are available and appropriate for a task, or
   • non-W3C specifications that enable the creation of content that conforms at level A or better to the Web Content Accessibility Guidelines 1.0 [WCAG10].

Note: For instance, for markup, the user agent may conform to HTML 4 [HTML4], XHTML 1.0 [XHTML10], and/or XML 1.0 [XML]. For style sheets, the user agent may conform to CSS ([CSS1], [CSS2]). For mathematics, the user agent may conform to MathML 2.0 [MATHML20]. For synchronized multimedia, the user agent may conform to SMIL 1.0 [SMIL].

Checkpoint definitions

9.1 Provide content focus. (P1) Checkpoint 9.1

1. Provide at least one content focus for each viewport (including frames) where enabled elements are part of the rendered content.
2. Allow the user to make the content focus of each viewport the current focus.

Note: For example, when two frames of a frameset contain enabled elements, allow the user to make the content focus of either frame the current focus. Note that viewports "owned" by plug-ins that are part of a conformance claim are also covered by this checkpoint. See checkpoint 7.1 for information about implementing content focus according to operating environment conventions.

9.2 Provide user interface focus. (P1) Checkpoint 9.2

1. Provide a user interface focus.

Note: See checkpoint 7.1 for information about implementing user interface focus according to operating environment conventions.

9.3 Move content focus. (P1) Checkpoint 9.3

1. Allow the user to move the content focus to any enabled element in the viewport.
2. Allow configuration so that the content focus of a viewport only changes on explicit user request.
3. If the author has not specified a navigation order, allow at least forward sequential navigation, in document order, to each element in the set established by provision one of this checkpoint.

Note: In addition to forward sequential navigation, the user agent should also allow reverse sequential navigation. See checkpoint 9.9 for information about structured navigation. See checkpoints 5.1 and 6.6 for more information about focus changes.

9.4 Restore viewport state history. (P1) Checkpoint 9.4

1. For user agents that implement a viewport history mechanism, for each state in a viewport's browsing history, maintain information about the point of regard, content focus, and selection.
2. When the user returns to any state in the viewport history (e.g., via the "back button"), restore the saved values for the point of regard, content focus, and selection.

9.5 No events on focus change. (P2) Checkpoint 9.5

1. Allow configuration so that moving the content focus to or from an enabled element does not automatically activate any explicitly associated event handlers of any event type.

Note: For instance, in this configuration for an HTML document, do not activate any handlers for the 'onfocus', 'onblur', or 'onchange' attributes. In this configuration, user agents should still apply any stylistic changes (e.g., highlighting) that may occur when there is a change in content focus.

9.6 Show event handlers. (P2) Checkpoint 9.6

1. For the element with content focus, make available the list of input device event types for which there are event handlers explicitly associated with the element.

Note: For example, allow the user to query the element with content focus for the list of input device event types, or add them directly to the sequential navigation order described in checkpoint 9.3. See checkpoint 1.2 for information about activation of event handlers associated with the element with focus.

9.7 Move content focus in reverse. (P2) Checkpoint 9.7

1. Extend the functionality required in provision three of checkpoint 9.3 by allowing the same sequential navigation in reverse document order.
2. As part of satisfying provision one of this checkpoint, the user agent must not include disabled elements in the navigation order.

9.8 Provide text search. (P2) Checkpoint 9.8

1. Allow the user to search within rendered text content for a sequence of characters from the document character set.
2. Allow the user to start a forward search (in document order) from any selected or focused location in content.
3. When there is a match, do both of the following:
   • move the viewport so that the matched text content is within it, and
   • allow the user to search for the next instance of the text from the location of the match.
4. Alert the user when there is no match or after the last match in content (i.e., prior to starting the search over from the beginning of content).
5. Provide a case-insensitive search option for text in scripts (i.e., writing systems) where case is significant.

Note: If the user has not indicated a start position for the search, the search should start from the beginning of content. Per checkpoint 7.3, use operating environment conventions for indicating the result of a search (e.g., selection or content focus).

9.9 Allow structured navigation. (P2) Checkpoint 9.9

1. Allow the user to navigate efficiently to and among important structural elements in rendered content.
2. As part of satisfying provision one of this checkpoint, allow forward and backward sequential navigation.

Note: This specification intentionally does not identify which "important elements" must be navigable as this will vary by specification. What constitutes "efficient navigation" may depend on a number of factors as well, including the "shape" of content (e.g., sequential navigation of long lists is not efficient) and desired granularity (e.g., among tables, then among the cells of a given table).

9.10 Configure important elements. (P3) Checkpoint 9.10

1. Allow configuration of the set of important elements and attributes identified for checkpoints 9.9 and 10.4.
2. As part of satisfying provision one of this checkpoint, allow the user to include and exclude element types in the set.

Note: For example, allow the user to navigate only paragraphs, or only headings and paragraphs, or to suppress and restore navigation bars, or to navigate within and among tables and table cells.

Checkpoint definitions

10.1 Associate table cells and headers. (P1) Checkpoint 10.1

1. For graphical user agents that render tables, for each table cell, allow the user to view associated header information.

10.2 Highlight selection, content focus, enabled elements, visited links. (P1) Checkpoint 10.2

1. Allow global configuration to highlight the following four classes of information in each viewport: the selection, content focus, enabled elements, and recently visited links.
2. For graphical user interfaces, as part of satisfying provision one of this checkpoint, allow at least one configuration where the highlight mechanisms for the four classes of information:
   • differ from each other, and
   • do not rely on rendered text foreground and background colors alone.
3. For graphical user interfaces, as part of satisfying provision one of this checkpoint, if a highlight mechanism involves text size, font family, rendered text foreground and background colors, or text decorations, offer at least the following range of values:
   • for text size, the range required by provision three of checkpoint 4.1.
   • for font family, the range required by provision three of checkpoint 4.2.
   • for text foreground and background colors and decorations, the range offered by the conventional utility available in the operating environment for users to choose rendered text colors or decorations (e.g., the standard font and color dialog box resources supported by the operating system). If no such utility is available, the range supported by the conventional APIs of the operating environment for specifying text colors or drawing text.
4. Highlight enabled elements according to the granularity specified in the format. For example, an HTML user agent rendering a PNG image as part of a client-side image map is only required to highlight the image as a whole, not each enabled region. An SVG user agent rendering an SVG image with embedded graphical links is required to highlight each (enabled) link that may be rendered independently according to the SVG specification.

Note: Examples of highlight mechanisms for selection and content focus include foreground and background color variations, underlining, distinctive synthesized speech prosody, and border styling. Because the selection and focus change frequently, user agents should not highlight them using mechanisms (e.g., font size variations) that cause content to reflow, as this may disorient the user. Graphical highlight mechanisms that generally do not rely on rendered text foreground and background color alone include underlines or border styling. Per checkpoint 7.1, follow operating environment conventions that benefit accessibility when implementing the selection and content focus. For instance, if specified at the level of the operating environment, inherit the user's preferences for selection styles.

10.3 Single highlight configuration. (P2) Checkpoint 10.3

1. Extend the functionality required by provision two of checkpoint 10.2 by allowing configuration through a single setting.

10.4 Provide outline view. (P2) Checkpoint 10.4

1. Make available to the user an "outline" view of rendered content, composed of labels for important structural elements (e.g., heading text, table titles, form titles, and other labels that are part of the content).

Note: This outline view will provide the user with a simplified view of content (e.g, a table of contents). For information about what constitutes the set of important structural elements, see the Note following checkpoint 9.9. By making the outline view navigable, it is possible to satisfy this checkpoint and checkpoint 9.9 together: allow users to navigate among the important elements of the outline view, and to navigate from a position in the outline view to the corresponding position in a full view of content. See checkpoint 9.10 for additional configuration options.

10.5 Provide link information. (P3) Checkpoint 10.5

1. To help the user decide whether to traverse a link in content, make available the following information about it:
   • link element content,
   • link title,
   • whether the link is internal to the resource (e.g., the link is to a target in the same Web page),
   • whether the user has traversed the link recently, and
   • information about the type, size, and natural language of linked Web resources.

Checkpoint definitions for the user interface

10.6 Highlight current viewport. (P1) Checkpoint 10.6

1. Highlight the viewport with the current focus (including any frame that takes current focus).
2. For graphical viewports, as part of satisfying provision one of this checkpoint, provide at least one highlight mechanism that does not rely on rendered text foreground and background colors alone (e.g., use a thick outline).
3. If the techniques used to satisfy provision one of this checkpoint involve rendered text size, font family, rendered text foreground and background colors, or text decorations, allow global configuration and offer same ranges of values required by provision three of checkpoint 10.2.

Note: See checkpoint 7.1 for information about implementing highlight mechanisms according to operating environment conventions.

10.7 Indicate viewport position. (P3) Checkpoint 10.7

1. Indicate the viewport's position relative to rendered content (e.g., the proportion of an audio or video clip that has been played, and the proportion of a Web page that has been viewed).

Checkpoint definitions

11.1 Current user input configuration. (P1) Checkpoint 11.1

1. Provide information to the user about current user preferences for input configurations.

11.2 Current author input configuration. (P2) Checkpoint 11.2

1. Provide a centralized view of the current author-specified input configuration.

Note: For example, for HTML documents, provide a view of keyboard bindings specified by the author through the "accesskey" attribute. The intent of this checkpoint is to centralize information about author-specified bindings so that the user does not have to read an entire document to look for available bindings.

11.3 Allow override of bindings. (P2) Checkpoint 11.3

1. Allow the user to override any binding that is part of the user agent default input configuration.

Note: See checkpoint 11.5 for default input configuration requirements and checkpoint 12.3 for information about their documentation.

11.4 Single-key access. (P2) Checkpoint 11.4

1. Allow the user to override any binding in the user agent default keyboard configuration with a binding to either a key plus modifier keys or to a single key.
2. For each functionality in the set required by checkpoint 11.5, allow the user to configure a single-key binding. A single-key binding is one where a single key press performs the task, with zero modifier keys.

Note: Because single-key access is so important to some users with physical disabilities, user agents should ensure that: (1) most keys of the physical keyboard may be configured for single-key bindings, and (2) most functionalities of the user agent may be configured for single-key bindings. For information about access to user agent functionality through a keyboard API, see checkpoint 6.7.

11.5 Default input configuration. (P2) Checkpoint 11.5

1. Ensure that the user agent default input configuration includes bindings for the following functionalities required by other checkpoints in this document:
   • move content focus to the next enabled element in document order, and move content focus to the previous enabled element in document order (checkpoints 9.3 and 9.7);
   • activate the link designed by the content focus (checkpoints 1.1 and 9.1);
   • search for text, search again for same text (checkpoint 9.8);
   • increase the scale of rendered text, and decrease the scale of rendered text (checkpoint 4.1);
   • increase global volume, and decrease global volume (checkpoint 4.7);
   • stop, pause, resume, and navigate efficiently selected audio and animations, including video and animated images (checkpoint 4.5).
2. If the user agent supports the following functionalities, the default input configuration must also include bindings for them:
   • next history state (forward), and previous history state (back);
   • enter URI for a new resource;
   • add a URI to favorites (i.e., bookmarked resources);
   • view favorites;
   • reload a resource;
   • interrupt a request to load or reload a resource;
   • for graphical viewports: navigation forward and backward through rendered content by approximately the height of the viewport;
   • for user agents that render content in lines of (at least) text: move point of regard to next line, and previous line.

Note: This checkpoint does not make any requirements about the ease of use of default input configurations, though clearly the default configuration should include single-key bindings and allow easy operation. Ease of use is addressed by the configuration requirements of checkpoint 11.3.

11.6 User profiles. (P2) Checkpoint 11.6

1. For the configuration requirements of this document, allow the user to save user preferences in at least one user profile.
2. Allow the user to choose from among available user agent default profiles, profiles created by the same user, and no profile (i.e., the user agent default settings).

11.7 Tool bar configuration. (P3) Checkpoint 11.7

1. For graphical user agent user interfaces with tool bars, allow the user to configure the position of user agent user interface controls on those tool bars.
2. Offer a predefined set of controls that may be added to or removed from tool bars.
3. Allow the user to restore the default tool bar configuration.

Checkpoint definitions

12.1 Provide accessible documentation. (P1) Checkpoint 12.1

1. Ensure that at least one version of the user agent documentation conforms to at least level Double-A of the Web Content Accessibility Guidelines 1.0 [WCAG10].

12.2 Document accessibility features. (P1) Checkpoint 12.2

1. Document all user agent features that benefit accessibility.

Note: The help system should include discussion of user agent features that benefit accessibility. The user agent should satisfy this checkpoint by providing both centralized and integrated views of accessibility features in the documentation.

12.3 Document default bindings. (P1) Checkpoint 12.3

1. Document the default user agent input configuration (e.g., the default keyboard bindings).

Note: Documentation should warn the user whenever the default input configuration is inconsistent with conventions of the operating environment.

12.4 Document changes between versions. (P2) Checkpoint 12.4

1. Document changes from the previous version of the user agent to features that benefit accessibility, including features of the user interface.

12.5 Provide dedicated accessibility section. (P2) Checkpoint 12.5

1. Provide a centralized view of all features of the user agent that benefit accessibility, in a dedicated section of the documentation.

Note: Developers are encouraged to integrate descriptions of accessibility features into the documentation alongside other features, in addition to providing a centralized view.

3 Accessibility topics

This section presents general techniques that may be relevant to more than one checkpoint. This section is not written in terms of the requirements of the checkpoints. Instead, it is organized according to other topics that should also be familiar to user agent designers.

3.1 Access to content

User agents need to ensure that users have access to content, either rendered through the user interface or made available to assistive technologies through an API. While providing serial access to a stream of content would satisfy this requirement, this would be analogous to offering recorded music on a cassette: other technologies exist (e.g., CD-ROMs) that allow direct access to music. It is just as important for user agents to allow users to access Web content efficiently, whether the content is being rendered as a two-dimensional graphical layout, an audio stream, or series of lines of braille. Providing efficient access to content involves:

• Preserving structure when rendering;
• Allowing the user to select specific content and query its structure or context (what am I examining?);
• Using and generating metadata to provide context (where am I?).

These topics are addressed below.

Note: Throughout this document, the term context menu refers to a user interaction mechanism whereby:

1. The user designates a piece of content (e.g., with the selection or focus).
2. The user invokes a menu (e.g., through a keyboard shortcut) consisting of options determined by the context selected by the user. For instance, if the user has selected an HTML table cell, the context menu might offer information about associated headers.

Preserve and provide structure

When used properly, markup languages structure content in ways that allow user agents to communicate that structure across different renderings. A table describes relationships among cells and headers. Graphically, user agents generally render tables as a two-dimensional grid. However, relationships also have to be apparent for users with serial access to content or who navigate sequentially, otherwise users may not understand the purpose of the table and the relationships among its cells (see the section on table techniques). User agents need to render content in ways that allow users to understand the underlying document structure, which may consist of headings, lists, tables, synchronized multimedia, and link relationships. Providing alternative renderings (e.g., an outline view) will also help users understand document structure.

Note: Even though the structure of a language like HTML may be defined by a document type definition (DTD) or a schema, user agents may convey structure according to a "more intelligent" document model when that model is well-known. For instance, in the HTML 4 [HTML4] and XHTML 1.0 [XHTML10] DTDs, heading elements (H1 - H6) do not nest, but presenting the document as nested headings may convey the document's structure more effectively than as a flat list of headers.

Allow access to selected content

The guidelines emphasize the importance of navigation as a way to provide efficient access to content. Navigation allows users to access content more efficiently and, when used in conjunction with selection and focus mechanisms, allows users to query content for metadata. For instance, blind users often navigate a document by skipping from link to link, deciding whether to follow each link based on metadata about the link. User agents can help them decide whether to follow a link by allowing them to query each focused link for the link text, title information, and information about whether the link has been visited. While much of this information may be rendered, the information has to also be available to assistive technologies.

For example, the Amaya browser/editor [AMAYA] makes available all attributes and their values to the user through a context menu. The user selects an element and opens an attribute menu that shows which attributes are available for the element and which have been assigned values. The user may read or write values to attributes (since Amaya is an editor as well as a browser). Information about attributes is also available through Amaya's structured view, which renders the document tree as structured text.

The selection may be widened (moved to the nearest node one level up the document tree) by pressing the Escape key; this is a form of structured navigation based on the underlying document object model.

Users may want to select content based on structure alone (as offered by Amaya) but also based on how the content has been rendered. For instance, most user agents allow users to select ranges of rendered text that may cross "element boundaries" (e.g., to select part of a paragraph that includes a phrase that is emphasized).

Context

Authors and user agents provide context to users through content, structure, navigation mechanisms, and query mechanisms. Titles, dimensions, dates, relationships, the number of elements, and other metadata all help orient the user, particularly when available as text. For instance, user agents can help orient users by allowing them to request that document headings and lists be numbered. See also the section on table techniques, which explains how user agents can offer table navigation, and the ability to query a table cell for information about the cell's row and column position and associated headers.

• User agents can use style sheet languages such as CSS 2 [CSS2] and XSLT [XSLT] to generate context information (see techniques for generated content).
• For information about elements and attributes that convey metadata in HTML, refer to the index of elements and attributes in "Techniques for Web Content Accessibility Guidelines 1.0" [WCAG10-TECHS].
• For information about elements and attributes that convey metadata in SMIL, refer to the index of attributes in the W3C Note "Accessibility Features of SMIL" [SMIL-ACCESS].
• Describe a selected element's position within larger structures (e.g., numerical or relative position in a document, table, or list). For example: tenth link of fifty links; document heading 3.4; list one of two, item 4.5; third table, three rows and four columns; current cell in third row, fourth column. Allow users to get this information on demand (e.g., through a keyboard shortcut). Provide this information on the status line on demand from the user.

3.2 User control of rendering and style

To ensure accessibility, users need to be able to configure the style of rendered content and the user interface. Author-specified styles, while important, may make content inaccessible to some users. User agents need to allow users to increase the scale of rendered text, to change colors and color combinations, and to slow down multimedia presentations.

Cascading Style Sheets (CSS, defined in [CSS1] and [CSS2]) give authors design flexibility and allow users to control important aspects of content style; see checkpoint 4.14 for information about allowing users to choose from among available style sheets. CSS includes mechanisms for tailoring rendering for a particular output medium, including audio, braille, screen, and print.

• User agents should implement the cascade order of CSS 2 ([CSS2], section 6.4.1) not CSS 1. In CSS 2, user style sheets with "!important" declarations (section 6.4.2) take precedence over author styles. Refer also to Web Content Accessibility Guidelines 1.0 checkpoint 3.3 [WCAG10] for content requirements related to style sheets.
• CSS-enabled user agents should consider as part of the cascade the markup used for style, giving it a lower weight than actual style sheets. This allows authors to specify style through markup for older user agents and to use more powerful style sheets for CSS-enabled user agents. Refer to the section on the precedence of non-CSS presentational hints in CSS 2 ([CSS2], section 6.4.4).
• To hide the CSS syntax from the user, user agents may implement user style sheets through the user agent user interface. User agents can generate a user style sheet from user preferences or behave as though it did. Amaya [AMAYA] provides a GUI-based interface to create and apply internal style sheets. The same technique may be used to control a user style sheet.
• In JavaScript, the following may be used to change style information:
  document.all.myElement style.color = "red";

3.3 Link techniques

User agents make links accessible by providing navigation to links, helping users decide whether to follow them, and allowing interaction in a device-independent manner. Link techniques include the following:

• See sequential navigation techniques for information about navigating to links.
• When the user follows a link where the URI has a fragment identifier (e.g., "#chap1"), render the beginning of the target fragment at the top of the viewport (not at the bottom, even if closer). Note: This technique applies to formats like HTML where the fragment identifier semantics are to identify a piece of content.
• Provide a link view that lists all links in the document. Allow the user to configure how the links are sorted (e.g., by document order, sequential navigation order, alphabetical order, visited or unvisited or both, internal or external or both).
• Help the user remember links by including metadata in the link view. For example, identify a selected link as "Link X of Y", where "Y" is the total number of links. Lynx [LYNX] numbers each link and provides information about the relative position in the document. Position is relative to the current page and the number of the current page out of all pages. Each page usually has 24 lines.
• Allow the user to configure how much information about a link to present in the content view (when a link receives focus). For instance, allow the user to choose between "Display links using hyperlink text" or "Display links by title (if present)", with an option to toggle between the two views. For a link without a title, use the link text.
• Here is a sample algorithm for ensuring that an HTML link that has image content has associated text.
  1. If the author has specified conditional content (that is not empty content) for the image (e.g., "alt" in HTML), use that as the link text;
  2. Otherwise, use the link title if available;
  3. [Repair] Otherwise, use title information of the designated Web resource (e.g., the TITLE element of HTML for links to HTML documents).
  4. [Repair] Otherwise, render part of the filename or URI of the designated Web resource.
  5. [Repair] Otherwise, insert a generic text placeholder (e.g., [LINK]) in place of the image (if configured to do so).
• For an image in link content, ensure that the user has access to the link and any long description associated with the image.

3.4 List techniques

User agents can make lists accessible by ensuring that list structure – and in particular, embedded list structure – is available through navigation and rendering.

• Allow users to turn on "contextual" rendering of lists (even for unordered "bullet" lists). Use compound numbers (or other symbols) to introduce each list item (e.g., "1, 1.1, 1.2, 1.2.1, 1.3, 2, 2.1"). This provides more context and does not rely on the information conveyed by a graphical rendering, as in:

  1.
    1.
    2.
      1.
    3.
  2.
    1.
  

  which might be serialized for synthesized speech or braille as "1, 1, 2, 1, 2, 3, 2, 1".

• Specify list numbering styles in CSS. Refer to the section generated content, automatic numbering, and lists in CSS ([CSS2], section 12).

  Example.

  The following CSS 2 style sheet (taken from CSS 2, section 12.5) shows how to specify compound numbers for nested lists created with either UL or OL elements. Items are numbered as "1", "1.1", "1.1.1", etc.

  <STYLE type="text/css">
   UL, OL { counter-reset: item }
   LI { display: block }
   LI:before { content: counters(item, "."); 
               counter-increment: item }
  </STYLE>
  

  End example.

3.5 Table techniques

The HTML TABLE element was designed to represent relationships among data ("data" tables). Even when authored well and used according to format specification, tables may pose problems for users with disabilities for a number of reasons:

• Users who have serial access to content or who navigate sequentially to a a table may have difficulty grasping the relationships among cells, especially for large and complex tables.
• Users with cognitive disabilities may have trouble grasping or remembering relationships between cells and headers, especially for large and complex tables.
• Users of screen magnifiers or with physical disabilities may have difficulties navigating to the desired cells of a table.

For these situations, user agents may assist these users by providing table navigation mechanisms and supplying context that is present in a two-dimensional rendering (e.g., the cells surrounding a given cell).

To complicate matters, many authors use tables to lay out Web content ("layout" tables). Not only are table structures used to arrange objects horizontally and vertically on the screen, table elements such as TH (table header) in HTML are used to change the style of text (as table headers are often rendered in bold fonts) rather than to indicate a true table header. These practices make it difficult for assistive technologies to rely on markup to convey document structure. Consequently, assistive technologies often resort to interpreting the rendered content, even though the rendered content has "lost" information encoded in the markup. For instance, when an assistive technology "reads" a table from its graphical rendering, the contents of multiline cells may become intermingled. For example, consider the following table:

This is the top left cell    This is the top right cell 
of the table.                of the table.

This is the bottom left      This is the bottom right 
cell of the table.           cell of the table.

Screen readers that read rendered content line by line would read the table cells incorrectly as "This is the top left cell This is the top right cell". So that assistive technologies are not required to gather incomplete information from renderings, these guidelines require that user agents provide access to content through an API (see checkpoint 6.3).

The following sections discuss techniques for providing improved access to tables.

Table metadata

Users with serial access to content or who navigate sequentially cannot gather information "at a glance" about a two-dimensional table. User agents can make tables more accessible by providing the user with table metadata such as the following:

• The table caption (the CAPTION element in HTML) or summary information (the "summary" attribute in HTML).
• The number of column groups and columns. Note that the number of columns may change according to the row. Also, some parts of a table may have more than two "dimensions." Project dimensionality higher than two onto two when rendering information.
• The number of row groups and rows, in particular information about table headers and footers.
• Which rows contain header information (whether at the top or bottom of the table).
• Which columns contain header information (whether at the left or right of the table).
• Whether there are subheads.
• How many rows or columns a header spans.

When navigating, quick access to table metadata will allow users to decide whether to navigate within the table or skip over it. Other techniques:

• Allow users to query table summary information from inside a cell.
• Allow the user to choose different levels of detail for the summary (e.g., brief table summary and a more detailed summary).
• Allow the user to configure navigation so that table metadata is not (re-)rendered each time the user enters the table.

Linear rendering of tables

A linear rendering of tables – cells presented one at a time, row by row or column by column – may be useful, but generally only for simple tables. For more complex tables, user agents need to convey more information about relationships among cells and their headers. A linear rendering of a table may be useful as an equivalent for a multi-dimensional table.

Note: The following techniques apply to columns as well as rows. The elements listed in this section are HTML 4.01 table elements ([HTML4], section 11).

• Provide access to one row (TR) at a time, beginning with any column header (TH). If a header is associated with more than one row, offer that header for each row concerned.
• Render cells (TD) with their associated headers. Allow the user to configure how often headers are rendered (e.g., by implementing the 'speak-header' property in CSS 2 [CSS2], section 17.7.1). Note also that the "abbr" attribute in HTML 4 specifies abbreviated headers for synthesized speech and other rendering ([HTML4], section 11.2.6). See also information about cell headers later in this section.
• Provide access to cell content as marked up in the document source.
• Refer to techniques for authoring accessible tables in "Techniques for Web Content Accessibility Guidelines 1.0" [WCAG10-TECHS].

Cell rendering

The most important aspect of rendering a table cell is that the cell's contents be rendered faithfully and be identifiable as the contents of a single cell. However, user agents may provide additional information to help orient the user:

• Render the row and column position of the cell in the table.
• Indicate how many rows and columns a cell spans.
• Since the contents of a cell in a data table may only be comprehensible in context (e.g., with associated header information, row/column position, and neighboring cell information), allow users to navigate to cells and query them for this information.
• For HTML tables, refer to the section on associating header information with data cells of HTML 4 ([HTML4], section 11.4.1).
• In a table with a leading row and column of TH cells, the interpretation of the corner cell as an empty TD or TH should not contribute to the set of headings for cells in that row and column.
• For nested tables, render information about the level of nesting.
• Since a cell may belong to N different dimensions in a multi-dimensional table, provide information about headers from each dimension.

Cell header algorithm

Properly constructed data tables distinguish header cells from data cells. How headers are associated with table cells depends on the markup language. The following algorithm is based on the HTML 4.01 algorithm to calculate header information ([HTML4], section 11.4.3). For the sake of brevity, it assumes a left-to-right ordering, but will also work (with minor modifications) for right-to-left tables (refer to the "dir" attribute of HTML 4 [HTML4], section 8.2). For a given cell:

• Search left from the cell's position to find row header (TH) cells. Then search upwards from the cell's position to find column header cells. The search in a given direction stops when the edge of the table is reached or when a data cell is found after a header cell. If no headers are found in either direction (left or up), search in the other directions (right or down).
• Allow the user to configure where the header text comes from. For example, in HTML 4, either the header cell element's content or the value of the "abbr" attribute value ([HTML4], section 11.2.6).
• Insert row headers into the list in the (left-to-right) order they appear in the table. Include values implicitly resulting from header cells in prior rows with rowspan="R", sufficient to extend into the current row.
• Insert column headers after row headers, in the (top-to-bottom) order they appear in the table. Include values implicitly resulting from header cells in other columns with colspan="C", sufficient to extend into the current column containing the TD cell.
• If a header cell has a value for the "headers" attribute, then insert it into the list and stop the search for the current direction.
• Treat cells with a value for the "axis" attribute as header cells.
• Be sure to take into account header cells that span several rows or columns.

Cell header repair strategies

Not all data tables include proper header markup, which the user agent may be able to detect. Some repair strategies for finding header information include the following:

• Consider that the top or bottom row contains header information.
• Consider that the leftmost or rightmost column in a column group contains header information.
• If cells in an edge row or column span more than one row or column, consider the following row or column to contain header information as well.
• When trying to guess table structure, present several solutions to the user.

Other repair issues to consider:

• Consider TH cells on both the left and right of the table.
• For TH cells with "rowspan" set: consider the content of those TH cells for each of the N-1 rows below the one containing that TH content.
• An internal TH surrounded by TD elements makes it difficult to know whether the header applies to cells to its left or right in the same row (or in both directions) or cells above or below it in the same column (or in both directions).
• Finding column header cells assumes they are all above the TD element to which they apply.
• A TH element with "colspan" set needs to be included in the list of TH elements for the N-1 columns to its right.

Table navigation

To permit efficient access to tables, user agents should allow users to navigate to tables and within tables, to select individual cells, and to query them for information about the cell and the table as a whole.

• Allow users to navigate to a table, down to one of its cells, and back up to the table level. This should work recursively for nested tables.
• Allow users to navigate to a cell by its row and column position.
• Allow users to navigate to all cells under a given header.
• Allow users to navigate row by row or column by column.
• Allow users to navigate to the cells around the current cell.
• Allow users to navigate to the first or last cell of a row, column, or the table.
• Allow users to navigate from a cell directly to its related headers (if it's possible to navigate to the headers).
• Allow the user to search for text content within a table (i.e., without searching outside of the table). Allow the user to search for text within specific rows or columns, row groups or column groups, or limited by associated headers.
• Alert the user when the navigation reaches a table edge and when a cell contains another table.
• Allow relative and direct navigation. For example, entering "-3, 20" might mean "left three cells, up 20 cells").
• Allow navigation of table headers or footers only.
• Consider the issues raised by navigation to or from a cell that spans more than one row or column.
• For examples of table navigation, refer to the table navigation script from the Trace Research Center [TABLENAV].

3.6 Image map techniques

One way to make an image map accessible to some users (e.g., users with blindness) is to render the links it contains as text links. This allows assistive technologies to render the links as synthesized speech or braille, and benefits users with slow access to the Web and users of small Web devices that do not support images but can support hypertext. User agents may allow users to toggle back and forth between a graphical mode for image maps and a text mode.

To construct a text version of an image map in HTML:

• If the content of the MAP element includes links, use them.
• Otherwise, for each AREA in the map, if (not empty conditional text content is available (e.g., the "alt" attribute), use it as the content of a generated link.
• When the author has specified empty conditional text content ("alt=''"), do not render the link.
• When the author has specified no text equivalent (no "alt"), repair the missing content per checkpoint 2.7.

Furthermore, user agents that render a text image map instead of an image may preface the text image map with inline metadata such as:

• a string that announces the image map (e.g., "Start map")
• any conditional text content associated with the image (e.g., "alt" for IMG).
• the number of links in the map.

Allow users to suppress, shrink, and expand text versions of image maps so that they may quickly navigate to an image map (which may be, for example, a navigation tool bar) and decide whether to "expand" it and follow the links of the map. The metadata listed above will allow users to decide whether to expand the map. Ensure that the user can expand and shrink the map and navigate its links using the keyboard and other input devices.

3.7 Frame techniques

HTML frames (see, for example, the FRAME and IFRAME elements in HTML 4 [HTML4]) were originally designed so that authors could divide up graphic real estate and allow the pieces to change independently (e.g., selecting an entry in a table of contents in one frame changes the contents of a second frame). While frames are not inherently inaccessible, they raise some accessibility issues:

• Equivalents to frame content. Some users cannot make use of frames because they cannot grasp the (spatial or logical) relationships conveyed by frame layout. Others cannot use them because their user agents or assistive technology does not support them or makes access difficult (e.g., users with screen readers or screen magnifiers).
• Navigation. Users need to be able to navigate from frame to frame in a device independent manner.
• Orientation. Users need to know what frame they are in (so, for example, authors should provide a title for each frame), what other frames are available, and how the frames of a frameset are organized.
• Dynamic changes. Users need to know how the changes they cause in one frame affect other frames.

To name a frame in HTML, use the following algorithm:

1. Use the "title" attribute on FRAME; or if not present,
2. Use the "name" attribute on FRAME; or if not present,
3. Use title information of the referenced frame source (e.g., the TITLE element of the source HTML document); or
4. Use title information of the referenced long description (e.g., what "longdesc" refers to in HTML), or
5. Use frame context (e.g., "Frame 2.1.3" to indicate the path to this frame in nested framesets).

To make frames accessible, user agents should do the following:

• Make available conditional content related to frames (e.g., provided by the HTML 4 NOFRAMES element ([HTML4], section 16.4.1).
• Here is a technique for the case of a frameset that does not contain a NOFRAMES element but the individual frames have associated long descriptions ("longdesc"):
  1. For each frameset, render the frameset title as an  H1 heading.
  2. For each frame, render the frame title in an H2 heading, followed by the content of the associated long description.
  3. Create a navigable table of contents according to the (possibly nested) frameset structure. Each entry in the table of contents should link to a frameset or frame. The end of the content used for each frame should include a link back to this table of contents.
• Alert the user when the viewport contains a frameset.
• Render a frameset as a list of links to named frames so the user can identify the number of frames. The list of links may be nested if framesets are nested.
• Provide information about the number of frames in the frameset.
• Highlight the frameset with the current focus (e.g., by using a thick border, by displaying the name of the frameset in the status bar).
• Allow the user to query the frame with the current focus for metadata about the frame. Make available the frame title for speech synthesizers and braille displays. Users may also use information about the number of images and words in the frame to guess the purpose of the frame.
• Allow navigation between frames (forward and backward through the nested structure, return to a top-level list of links to frames).
• Alert the user when an action in one frame causes the content of another frame to change. Allow the user to navigate with little effort to the frame(s) that changed.
• Authors can suppress scrolling of HTML frames with scrolling="no".
• In order to ensure that content is accessible, allow the user to override some attributes of the FRAME element of HTML 4 ([HTML4], section 16.2.2): "noresize", "scrolling", and "frameborder".

The following screen shot illustrates how Home Page Reader [HPR] renders a frameset by allowing the user to navigate (on the left side) to each of five frames in a frameset.

The next screen shot illustrates how the user agent can provide information about the number and structure of frames in the user agent user interface.

3.8 Form techniques

To make a form accessible, the user agent needs to ensure that:

• the user can navigate to all of the form elements;
• information about the form and its elements is available on demand;
• the user can interact with all form elements through the keyboard alone (or voice alone or pointing device alone).

Form navigation techniques

• Allow users to navigate to forms and to all controls within a form (refer also to table navigation techniques). Opera [OPERA] and Navigator [NAVIGATOR] offer a number of "form navigation" keyboard commands. When invoked, these "form navigation" commands move the user agent's current focus to the first form element (if any) in the document.
• If there are no forms in a document and the user attempts to navigate to a form, alert the user.
• Provide a navigable, structured view of form elements (e.g., those grouped by LEGEND or OPTGROUP in HTML) along with their labels.
• Allow the user to navigate away from a menu without selecting any option (e.g., by pressing the Escape key).

Form orientation techniques

Provide the following information about forms on demand:

• The number of forms in the document.
• The percentage of a form that has already been filled out. This will help users who navigate sequentially to form controls know whether they have completed the form. Otherwise, users who encounter a submit button that is not the last control of the form might inadvertently submit the incomplete form.

Form element orientation techniques

In conjunction with navigation:

• As the user navigates to a form element, provide information about whether the control has to be activated before form submission. For instance, in section 6.1.3 of XForms 1.0 [XFORMS10], the required property describes whether a value is required before the form's instance data is submitted.
• For labels associated with form elements in markup (e.g., the "for" attribute on LABEL in HTML), make available label information when the user navigates among the form elements.
• As the user navigates to a form element, provide information (e.g., through context-sensitive help) about how the user can activate the element. Provide information about what is required for each form element. Lynx [LYNX] conveys this information by providing information about the currently selected form element via a status line message:
  • Radio Button: Use right-arrow or Return to toggle
  • Checkbox Field: Use right-arrow or Return to toggle
  • Option List: Press return and use arrow keys and return to select option
  • Text Entry Field: Enter Text. Use Up or Down arrows or Tab to move off
  • Textarea: Enter text. Up or Down arrows or Tab to move off (^Ve for editor) Note: The ^Ve (caret-V, e) command, included in the TEXTAREA status line message, enables the user to invoke an external editor defined in the local Lynx configuration file (lynx.cfg).

Provide the following information about the elements in a form on demand (e.g., for the element with focus):

• Indicate the number of elements in the form.
• Indicate the number of elements that have not yet been completed.
• Provide a list of elements that have to be activated before form submission.
• Provide information about the order of form elements (e.g., as specified by "tabindex" in HTML). This is important since:
  1. Most forms are visually oriented, employing changes in font size and color.
  2. Users who navigate sequentially to form controls need to know they have supplied all the necessary information before submitting the form.
• Provide information about which element has focus (e.g., "element X of Y for the form named MyForm"). The form name is very important for documents that contain more than one form. This will help users who navigate sequentially to form controls know whether they have completed the form.
• Allow the user to query a form element for information about title, value, grouping, type, status, and position.
• When a group of radio buttons receives content focus, identify the radio button with content focus as "Radio Button X of Y", where "Y" represents the total number of radio buttons in the group. HTML 4 specifies the FIELDSET element ([HTML4], section 17.10), which allows authors to group thematically related elements and labels. The LEGEND element ([HTML4], section 17.10) assigns a caption to a FIELDSET. For example, the LEGEND element might identify a FIELDSET of radio buttons as "Connection Rate". Each button could have a LABEL element ([HTML4], section 17.9.1) stating a rate. When it receives content focus, identify the radio button as "Connection Rate: Radio button X of Y: 28.8kpbs", where "Y" represents the total number of radio buttons in the grouping and "28.8kbps" is the information contained in the LABEL.
• Allow the user to invoke an external editor instead of editing directly in a TEXTAREA element. This allows users to use all the features of the external editor (e.g., macros, spell-checkers, validators, known input configurations, and autosave features).
• Provide an option for transforming menus into checkboxes or radio buttons. In the transformation, retain the accessibility information specified by the author for the original form elements. Preserve the labels provided for the OPTGROUP and each individual OPTION, and re-associate them with the generated checkboxes. The LABEL defined for the OPTGROUP should be converted into a LEGEND for the result FIELDSET, and each checkbox should retain the LABEL defined for the corresponding OPTION. Lynx [LYNX] does this for HTML SELECT elements that have the "multiple" attribute specified.

3.9 Generated content techniques

User agents may help orient users by generating additional content that "announces" a context change. This may be done through CSS 2 [CSS2] style sheets using a combination of selectors (including the ':before' and ':after' pseudo-elements described in section 12.1) and the 'content' property (section 12.2).

For instance, the user might choose to hear "language:German" when the natural language changes to German and "language:default" when it changes back. This may be implemented in CSS 2 with the ':before' and ':after' pseudo-elements ([CSS2], section 5.12.3)

[lang|=es]:before { content: "start Spanish "; }
[lang|=es]:after  { content: " end Spanish"; }

<p lang="es" class="Spanish">
 <a href="foo_esp.html" 
    hreflang="es">Esta pagina en español</a></p>

The following example uses style sheets to distinguish visited from unvisited links with color and a text prefix.

A:link           { color: red }   /* Unvisited links */
A:visited        { color: green } /* Visited links */
A:visited:before { content: "Visited link: " }

To hide content, use the CSS 'display' or 'visibility' properties ([CSS2], sections 9.2.5 and 11.2, respectively). The 'display' property suppresses rendering of an entire subtree. The 'visibility' property causes the user agent to generate a rendering structure, but the content is invisible.

The following XSLT style sheet (excerpted from the XSLT Recommendation [XSLT], Section 7.7) shows how one might number H4 elements in HTML with a three-part label.

<xsl:template match="H4">
 <fo:block>
  <xsl:number level="any" from="H1" count="H2"/>
  <xsl:text>.</xsl:text>
  <xsl:number level="any" from="H2" count="H3"/>
  <xsl:text>.</xsl:text>
  <xsl:number level="any" from="H3" count="H4"/>
  <xsl:text> </xsl:text>
  <xsl:apply-templates/>
 </fo:block>
</xsl:template>

3.10 Content repair techniques

When generating repair content, user agent developers should consider the following issues:

• Some assistive technologies rely on an accurate mapping between the document object and what the user agent renders. If repair content is not included in the document object, but is used to determine rendering, there is likely to be a mismatch between the two that may lead to confusion. User agent developers should therefore consider including repair content in the document object. On the other hand, some users may wish to leave the author's content unmodified, and others may wish to have repair content included in the document object, but to be informed about what was repaired.
• Repair content inserted in the document object should conform to the Web Content Accessibility Guidelines 1.0 [WCAG10]. For example, if the user agent inserts a graphical icon in the document object, that icon should have a text equivalent: since the icon is known to the user agent developer, the developer can provide a sensible text equivalent to accompany it (for the benefit of users of assistive technologies).

See also the section on table cell header repair strategies. Refer also to the W3C document "Techniques for Authoring Tool Accessibility Guidelines 1.0" [ATAG10-TECHS].

3.11 Script and applet techniques

User agents need to make dynamic content accessible to users who may be disoriented by changes in content, who may have a physical disability that prevents them from interacting with a document within a time interval specified by the author, or whose user agent does not support scripts or applets. Not only do user agents make available equivalents to dynamic content (e.g., audio, animations), they have to allow users to turn off scripts, to stop animations, and to adjust timing parameters. Some user agents also allow users to turn off scripts for security reasons.

Script techniques

Certain elements of a markup language may have associated event handlers that are activated when certain events occur. User agents need to be able to identify those elements with event handlers statically associated (i.e., associated in the content, not in a script). In HTML 4 ([HTML4], section 18.2.3), intrinsic events are specified by the attributes beginning with the prefix "on": "onblur", "onchange", "onclick", "ondblclick", "onkeydown", "onkeypress", "onkeyup", "onload", "onmousedown", "onmousemove", "onmouseout", "onmouseover", "onmouseup", "onreset", "onselect", "onsubmit", and "onunload".

Techniques for providing access to scripts include the following:

• Allow the user to configure the user agent so that mouseover/mouseout event handlers are activated by (and activate) focus/blur events. Similarly, allow the user to use a key command, such as enter and Shift-Enter to fire "onclick" and "ondblclick" events.
• Implement "Document Object Model (DOM) Level 2 Events Specification" [DOM2EVENTS] events with a single activation event and provide a method for firing that event with each supported input device or input API. These should be the same as the click events and mappings provided above (but note that a user agent which is also an editor may wish to use single click events for moving a system caret, and want to provide a different behavior to activate using the mouse). For example, Amaya [AMAYA] uses a "doAction" command for activating links and form elements, which can be activated either by the mouse (and it is possible to set it for single-click or double-click) or by the keyboard (it is possible to set it for any key using Amaya's keyboard configuration)
• For authors: Document the effects of known important scripts to give users an idea in advance of what they do. Do so by using the relevant elements or attributes of the (markup language) specification, or if there aren't any, make available a description of the script behavior.

Applet techniques

When a user agent loads an Java applet, it should support the Java system conventions for loading an assistive technology (see the appendix on loading assistive technologies for DOM access). If the user is accessing the applet through an assistive technology, the assistive technology should alert the user when the applet receives content focus as this will likely result in the launch of an associated plug-in or browser-specific Java Virtual Machine. The user agent then needs to turn control of the applet over to the assistive technology. User agents need to make conditional content available to the assistive technology. Applets generally include an application frame that provides title information.

3.12 Input configuration techniques

User agents that allow users to modify default input configurations need to account for configuration information from several sources: user agent defaults, user preferences, author-specified configurations, and operating environment conventions. Some examples include:

• Author: in HTML, the author may specify keyboard bindings with the "accesskey" attribute ([HTML4], section 17.11.2).
• User: Users generally specify their preferences through the user interface but may also do so programmatically or through a profile.
• Operating environment: Users may specify preferred color contrasts at the operating environment level.

To the user, the most important information is the final configuration once all sources have been cascaded (combined) and all conflicts resolved. Knowing the default configuration is also important; checkpoint 12.3 requires that the default configuration be documented. The user may also want to know how the current configuration differs from the default configuration and what configuration in the viewport with the current focus comes from the author. This information may also be useful to technical support personnel who may be assisting users.

• The user interfaces for viewing and editing the input configuration may be combined, but need not be. When a single interface is available to edit bindings from any source, allow the user to apply filters to the list of bindings (e.g., author-specified only, user agent default, user preferences only, or final configuration).
• The user interfaces for viewing and editing the input configuration need to be accessible; see the requirements of this document.
• In the user interface, associate with each binding a short text description of the function to be activated. For example, if "Control-P" maps to a print functionality, a short description might be "Print" or "Print setup". For author-specified configurations, use available information (e.g., "title") or use generic descriptions of what action will be taken(e.g., "Follow the link with this link text").
• Allow users to query user interface controls for pertinent input configuration information (e.g., what key will activate the functionality).

Resolution of input configuration conflicts

In general, user preferences should override other configurations, however this may not always be desirable. For example, users should be prevented from configuring the user agent in a way that would interfere with important functionalities such as quitting the user agent or reconfiguring it.

Some options for resolving conflicts:

• Allow author configurations to override other configurations and alert the user when this happens.
• Do not allow author configurations to override other configurations. Alert the user when an author-specified binding has been overridden and provide access to the author-specified control through other means (e.g., an unused binding, a menu, or in a list of all author-specified bindings)
• Author-specified keyboard bindings in combination with the user agent's native configuration may conflict with operating environment conventions. For example, Internet Explorer [IE-WIN] in Windows uses the Alt key as the Compose key for author-specified bindings. If the author has specified a configuration with the characters "h" or "f", this will interfere with the operating environment conventions for accessing help and the file menu, respectively. In addition to the previous two options for handling conflicts, the user agent may allow the user to choose another Compose key (either globally or on a per-document basis when conflicts are detected).

Refer to the section on restricted functionality and conformance in UAAG 1.0 for more information about limitations in functionality due to content.

3.13 Synthesized speech techniques

The following techniques apply to user agents that render content as synthesized speech.

• Since user agents that render content as synthesized speech do not always pronounce it correctly, they should provide additional context to facilitate understanding. Techniques include:
  • Spelling words;
  • Indicating punctuation and capitalization;
  • Allowing users to repeat words alone and in context;
  • Using auditory nuances – including pitch, articulation model, volume, and orientation – to convey meaning the way fonts, spacing, and borders do in graphical media;
  • Generating context. For example, a user agent might speak the word "link" before a link, "heading" before the text content of a heading or "item 1.4" before a list item;
  • Rendering text according in the appropriate natural language.
• User agents that synthesize speech should implement the CSS 2 aural style sheet properties ([CSS2], section 19) to allow users to configure synthesized speech rate, volume, and pitch.
• User agents that provide accessible solutions for images should, by default, provide no information about images where the author has provided empty conditional content associated with the image, otherwise information may clutter the user's view of the content and cause confusion. See checkpoints 2.7 and 2.8 for more information about repair content.
• User agents may recognize different natural languages and be able to render content according to language markup defined for a certain part of the document. For instance, a screen reader might change the pronunciation of spoken text according to the language definition. This is usually desired and done according to the capabilities of the tool. Some specialized tools might give some finer user control for the pronunciation as well.
• Switching natural languages for blocks of content may be more helpful than switching for short phrases. In some language combinations (e.g., Japanese being the primary and English being the secondary or quoted language), short foreign language phrases are often well-integrated in the primary language. Dynamic switching for these short phrases may make the content sound unnatural and possibly harder to understand. User agents might allow users to choose elements for which they want to be alerted.
• Announce different classes of links differently (see checkpoint 10.5). For instance, announce links internal to a resource as being different from links to another page in the same domain or from links to another domain. Announce visited links differently as well.
• The following techniques for speaking data tables are adapted from the "Tape Recording Manual" produced by the (USA) National Braille Association [NBA]:
  1. Read the title, source, captions and any explanatory keys.
  2. Describe the structure of the table. Include the number of columns, the headers of each column and any associated sub-columns, reading from left to right (for left-to-right tables). The subhead is not considered a column. If column heads have footnotes, read them following each header.
  3. Explain whether the table will be read by rows (horizontally) or by columns (vertically). The horizontal reading is usual but, in some cases, the vertical reading better conveys the content. On rare occasions it is necessary to read a table both ways.
  4. Repeat the column headers with the figures under them for the first two rows. If the table is long, repeat the headers every fifth row. Always repeat them during the reading of the last row.
  5. Indicate the last row by saying, "and finally . . . " or "last row ..."
  6. At the completion of the reading say "End table X." If the table appeared on a page other than the one you were recording, add "Returning to text on page Y."

References:

• For more information about voice browser technology developed at W3C, refer to "Voice Browsers: An introduction and glossary for the requirements drafts" [VOICEBROWSER].
• For information about voice recognition and accessibility, refer to "Speak to Write" [SPEAK2WRITE].

3.14 Techniques for reducing dependency on spatial interactions

Users with serial access to content or who navigate sequentially to content may have difficulty interacting with content in two-dimensional space (e.g., to move a pointing device). Using the keyboard to move the pointing device may help some users, but this technique usually requires a significant amount of visual feedback as well as physical dexterity, both of which may not be possible for users with some disabilities.

To illustrate ways to reduce dependency on spatial interactions, consider a Web site for travel through Europe. The author provides a map of Europe and allows users to select regions they wish to visit using a pointing device. This type of application is very convenient to some users, but may be inaccessible to others. Authors can design such content to place the emphasis on objects (in this case European countries) rather than screen coordinates. In HTML, for example, authors should use client-side image maps instead of server-side image maps.

When this is done, the user agent can present the list of countries in an alternative fashion (e.g., as a list of links or menu entries) for users who may find it difficult to select countries with the pointing device. This type of interface generally benefits all users (e.g., some users who may not recognize countries by shape or flag, some users may prefer the keyboard, some users may have images turned off, and the text may be searched).

3.15 Accessibility and internationalization techniques

The following techniques may be considered when integrating accessibility features and internationalization.

• Implement content negotiation so that users may specify language preferences. Or allow the user to choose manually from among related resources available in different languages.
• Consider operating environment level natural language preferences as the user's default language preference. However, take caution about sending HTTP Accept-Language request headers ([RFC2616], section 14.4) based on the operating environment preferences. First, there may be a privacy problem as indicated in RFC 2616, section 15.1.4 "Privacy Issues Connected to Accept Headers". Also, the operating environment may define only one language, while the Accept-Language request header may include many languages in different priorities. Setting Accept-Language to be the operating environment language may prevent a user from receiving content from a server that does not have a match for this particular language but does for other languages acceptable to the user.
• Render characters with the appropriate directionality. Refer to the "dir" attribute and the BDO element in HTML 4 ([HTML4], sections 8.2 and 8.2.4, respectively).

For more information about content internationalization, refer to W3C's "Character Model for the World Wide Web" [CHARMOD] and the Unicode Consortium's Unicode specification [UNICODE].

3.16 Appendix: Impact matrix

This matrix summarizes which checkpoints are expected to benefit users with certain types of disabilities. For more information about types of disabilities, assistive technologies, access strategies, and more, refer to W3C's "How People with Disabilities Use the Web" [PWD-USE-WEB].

learning

4.4

hard of hearing

1.3, 2.2, 2.5, 2.6, 4.1, 4.2, 4.3, 4.7, 4.12

memory

9.4, 9.10, 10.4, 10.5, 11.1, 11.2, 11.3, 11.4, 11.5, 11.6, 11.7

color deficiency

3.1, 4.3, 10.2, 10.3, 10.6

low vision

2.2, 2.5, 2.7, 2.8, 2.9, 3.2, 3.3, 3.5, 4.1, 4.2, 4.4, 4.6, 4.7, 4.8, 4.9, 4.10, 4.11, 4.12, 4.13, 5.1, 5.4, 9.1, 9.2, 9.3, 9.4, 9.7, 9.8, 9.10, 10.2, 10.3, 10.4, 10.5, 10.6

seizure disorder

3.3, 3.4

physical

1.1, 1.2, 2.4, 2.9, 2.10, 5.3, 5.5, 9.1, 9.2, 9.3, 9.4, 9.5, 9.6, 9.7, 9.8, 9.9, 9.10, 10.3, 10.4, 10.5, 11.1, 11.2, 11.3, 11.4, 11.5, 11.6, 11.7

all

2.1, 2.3, 4.14, 6.1, 6.2, 6.3, 6.4, 6.5, 6.6, 6.7, 6.8, 6.9, 6.10, 7.1, 7.2, 7.3, 7.4, 8.1, 8.2, 12.1, 12.2, 12.3, 12.4, 12.5

deafness

1.3, 2.2, 2.5, 2.6, 4.1, 4.2, 4.3, 4.6

cognitive

2.6, 3.1, 3.2, 3.3, 3.5, 3.6, 4.2, 4.3, 4.4, 4.5, 4.6, 5.1, 5.2, 5.3, 5.4, 5.5, 9.4, 9.8, 9.10, 10.4, 10.5, 10.7, 11.1, 11.2, 11.3, 11.4, 11.5, 11.6, 11.7

blindness

1.1, 1.2, 1.3, 2.2, 2.4, 2.5, 2.7, 2.8, 2.9, 2.10, 3.2, 3.3, 3.5, 4.4, 4.7, 4.8, 4.9, 4.10, 4.11, 4.12, 4.13, 5.1, 5.3, 5.4, 5.5, 9.1, 9.2, 9.3, 9.4, 9.5, 9.6, 9.7, 9.8, 9.9, 9.10, 10.2, 10.3, 10.4, 10.5, 10.6, 10.7, 11.1, 11.2, 11.3, 11.5, 11.6, 11.7

3.17 Appendix: Accessibility features of some operating systems

Several operating systems include built-in accessibility features designed to assist individuals with disabilities. Despite operating systems differences, the built-in accessibility features use a similar naming convention and offer similar functionalities, within the limits imposed by each operating system (or particular hardware platform). The following is a list of built-in accessibility features common to several operating environments:

StickyKeys

StickyKeys allows users who have difficulties with pressing several keys simultaneously to press and release in sequence each key of the configuration.

MouseKeys

These allow users to move the mouse cursor and activate the mouse button(s) from the keyboard.

RepeatKeys

RepeatKeys allows users to set how fast a key repeats ("repeat rate") when the key is held pressed. It also allows users to control how quickly the key starts to repeat after the key has been pressed ("delay until repeat"). Users can also turn off key repeating.

SlowKeys

SlowKeys instructs the computer not to accept a key as pressed until it has been pressed and held down for more than a user-configurable length of time.

BounceKeys

BounceKeys prevents extra characters from being typed if the user bounces (e.g., due to a tremor) on the same key when pressing or releasing it.

ToggleKeys

ToggleKeys provides an audible indication for the status of keys that have a toggled state (keys that maintain status after being released). The most common toggling keys include Caps Lock, Num Lock, and Scroll Lock.

SoundSentry

SoundSentry monitors the operating system and applications for sounds in order to provide a graphical indication when a sound is being played. Older versions of SoundSentry may have flashed the entire display screen for example, while newer versions of SoundSentry provide the user with a selection of options, such as flashing the viewport that has the current focus or flashing the active window caption bar.

The next three built-in accessibility features are not as commonly available as the above group of features, but are included here for completeness and future compatibility.

ShowSounds

ShowSounds are user settings or software switches that cause audio messages to be presented using both audio and graphics. Applications may use these switches as the basis of user preferences.

HighContrast

HighContrast sets fonts and colors designed to make the screen easier to read.

TimeOut

TimeOut turns off built-in accessibility features automatically if the computer remains idle for a user-configurable length of time. This is useful for computers in public settings such as a library. TimeOut might also be referred to as "reset" or "automatic reset".

The next accessibility feature listed here is not considered to be a built-in accessibility feature (since it only provides an alternative input channel) and is presented here only for completeness and future compatibility.

SerialKeys

SerialKeys allows a user to perform all keyboard and mouse functions from an external assistive device (such as communication aid) communicating with the computer via a serial character stream (e.g., through a serial port or infra-red port) rather than or in conjunction with, the keyboard, mouse, and other conventional input devices and methods.

Microsoft Windows 95, Windows 98, and Windows NT 4.0

The following accessibility features can be adjusted from the Accessibility Options Control Panel:

• StickyKeys: modifier keys include Shift, Control, and Alt.
• FilterKeys: grouping term for SlowKeys, RepeatKeys, and BounceKeys.
• MouseKeys
• ToggleKeys
• SoundSentry
• ShowSounds
• Automatic reset (term used for TimeOut)
• High Contrast
• SerialKeys

Additional accessibility features available in Windows 98:

Magnifier

Magnifier is a windowed, screen enlargement and enhancement program used by people with low vision to magnify an area of the graphical display (e.g., by tracking the text cursor or current focus). Magnifier can also invert colors within the magnification window.

Accessibility Wizard

The Accessibility Wizard is a setup tool to assist users with the configuration of operating environment accessibility features.

References:

• To find out about built-in accessibility features on Windows platforms, ask the operating system via the "SystemParametersInfo" function. Refer to "Software accessibility guidelines for Windows applications" [MS-ENABLE] for more information.
• For information about Microsoft keyboard configurations (Internet Explorer, Windows 95, Windows 98, and more), refer to documentation on keyboard assistance for Internet Explorer and MS Windows [MS-KEYBOARD].

Apple Macintosh operating system

The following accessibility features can be adjusted from the Easy Access Control panel. Note: The Apple naming convention for accessibility features is to put spaces between the terms (e.g., "Sticky Keys" instead of "StickyKeys").

• Sticky Keys: modifier keys include the Shift, Command (Open apple), Option (Alt), and Control keys.
• Slow Keys
• Mouse Keys

The following accessibility features can be adjusted from the Keyboard Control Panel.

• Key Repeat Rate (part of RepeatKeys)
• Delay Unit Repeat (part of RepeatKeys)

The following accessibility feature can be adjusted from the Sound or Monitors and Sound Control Panel (depending on operating system version).

• Adjusting the volume to off or mute causes the Macintosh to flash the title bar whenever the operating system detects a sound (e.g., SoundSentry)

Additional accessibility features available for the Macintosh OS:

CloseView

CloseView is a full screen, screen enlargement and enhancement program used by people with low vision to magnify the information on the graphical display, and it can also change the colors used by the operating system.

SerialKeys

SerialKeys is available as freeware from Apple and several other Web sites.

AccessX, X Keyboard Extension (XKB), and the X Window System

The following accessibility features can be adjusted from the AccessX graphical user interface X client on some DEC, SUN, and SGI operating systems. Other operating systems supporting XKB may require command line interaction.

• StickyKeys: modifier keys are platform-dependent, but usually include the Shift, Control, and Meta keys.
• RepeatKeys
• SlowKeys
• BounceKeys
• MouseKeys
• ToggleKeys

Note: AccessX became a supported part of the X Window System X Server with the release of the X Keyboard Extension in version X11R6.1.

DOS (Disk Operating System)

The following accessibility features are available from a freeware program called AccessDOS, which is available from several Internet Web sites including IBM, Microsoft, and the Trace Center, for either PC-DOS or MS-DOS versions 3.3 or higher.

• StickyKeys: modifier keys include the Shift, Control, and Alt keys.
• Keyboard Response Group: grouping term for SlowKeys, RepeatKeys, and BounceKeys
• MouseKeys
• ToggleKeys
• SoundSentry (incorrectly named ShowSounds)
• SerialKeys
• TimeOut

3.18 Appendix: Loading assistive technologies for access to the document object model

Many of the checkpoints in the guidelines require a "host" user agent to communicate information about content and the user interface to assistive technologies. This appendix explains how developers can ensure the timely exchange of this information (see checkpoint 6.10). The techniques described here include:

1. Loading the entire assistive technology in the address space of the host user agent;
2. Loading part of the assistive technology in the address space of the host user agent (e.g., piece of stub code, a dynamically linked library (DLL) or a browser helper object);
3. Out-of-process access to the document object model.

The first two techniques are similar, differing in the amount of, or capability of, the assistive technology loaded in the same process or address space as the host user agent. These techniques are likely to provide faster access to the document object model since they will not be subject to inter-process communication overhead.

Loading assistive technologies for direct navigation of the document object model

First, the host user agent needs to know which assistive technology to load. One technique for this is to store a reference to an assistive technology in a system registry file or, in the case of Java, a properties file. Registry files are common among many operating environments:

• Windows: use the system registry file
• IBM OS/2: use the system.ini
• On client/server systems: use a system registry server that an application running on the network client computer can query.
• In Sun Java 2, use the "accessibility.properties" file, which causes the system event queue to examine the file for assistive technologies required for loading. If the file contains a property called "assistive_technologies", it will load all registered assistive technologies and start them on their own thread in the Java Virtual Machine that is a single process.

assistive_technologies=com.ibm.sns.svk.AccessEngine

In Microsoft Windows, a similar technique could be followed by storing the name of a Dynamic Link Library (DLL) for an assistive technology in a designated assistive technology key name/assistive technology pair.

HKEY_LOCAL_MACHINE\Software\Accessibility\DOM 
    "ScreenReader, VoiceNavigation"

Attaching the assistive technologies to the document object model

Once the assistive technology has been registered, any other user agent can determine whether it needs to be loaded and then load it. Once loaded, the assistive technology can monitor the document object model as needed.

On a non-Java platform, a technique to do this would be to create a separate thread with a reference to the document object model using a DLL. This new thread will load the DLL and call a specified DLL entry name with a pointer to the document object model interface. The assistive technology process will then run as long as required.

The assistive technology has the option to either:

• communicate with a main assistive technology of its own and process the document object model as a caching mechanism for the main assistive technology, or
• act as a bridge to the document object model for the main assistive technology.

In the future, it will be necessary to provide a more comprehensive reference to the application that not only provides direct navigation to its client area document object model, but also multiple document object models that it is processing and an event model for monitoring them.

Java's direct access

Java can facilitate timely access to accessibility components. In this example, an assistive technology running on a separate thread monitors user interface events such as focus changes. When focus changes, the assistive technology is alerted of which component object has focus. The assistive technology can communicate directly with all components in the application by walking the parent/child hierarchy and connecting to each component's methods and monitor events directly. In this case, an assistive technology has direct access to component specific methods as well as those provided for by the Java Accessibility API. There is no reason that a document object model interface to user agent components could not be provided via Java.

In Java 1.1.x, Sun's Java access utilities load an assistive by monitoring the Java awt.properties file for the presence of assistive technologies and loads them as shown in the following code example:

import java.awt.*;
import java.util.*;
      
String atNames = 
  Toolkit.getProperty("AWT.assistive_technologies",null);
if (atNames != null) {
  StringTokenizer parser = new StringTokenizer(atNames," ,");
  String atName;
  while (parser.hasMoreTokens()) {
   atName = parser.nextToken();
   try {
     Class.forName(atName).newInstance();
   } 
   catch (ClassNotFoundException e) {
     throw new AWTError("Assistive Technology not found: " 
                     + atName);
   } 
   catch (InstantiationException e) {
     throw new AWTError("Could not instantiate Assistive"
                     + " Technology: " + atName);
   } 
   catch (IllegalAccessException e) {
     throw new AWTError("Could not access Assistive"
                     + " Technology: " + atName);
   } catch (Exception e) {
     throw new AWTError("Error trying to install Assistive"
                     + " Technology: " + atName + " " + e);
   }
  }
}

In the following code example, the constructor for the assistive technology, AccessEngine, adds a focus change listener using Java accessibility utilities. When the assistive technology is alerted that an object has received focus, it has direct access to that object. If the Object, o, has implemented a document object model interface, the assistive technology will have direct access to the document object model in the same process space as the application.

import java.awt.*;
import javax.accessibility.*;
import com.sun.java.accessibility.util.*;
import java.awt.event.FocusListener;

class AccessEngine implements FocusListener {
 public AccessEngine() {
   //Add the AccessEngine as a focus change listener
   SwingEventMonitor.addFocusListener((FocusListener)this);
 }

 public void focusGained(FocusEvent theEvent) {
   // get the component object source
   Object o = theEvent.getSource();
   // check to see if this is a dom component
   if (o instanceof DOM) {
     ...
   }
 }
 public void focusLost(FocusEvent theEvent) {
   // Do Nothing
 }
}

Loading part of the assistive technologies for direct access to the document object model

In order to attach to a running instance of Internet Explorer 4.0, you can use a Browser Helper Object ([BHO]), which is a DLL that will attach itself to every new instance of Internet Explorer 4.0 [IE-WIN] (only if you run iexplore.exe). You can use this feature to gain access to the object model of Internet Explorer and to monitor events. This can be tremendously helpful when many method calls need to be made to Internet Explorer, as each call will be executed much more quickly than the out of process case.

There are some requirements when creating a Browser Helper Object:

• The application that you create must be an in-process server (that is, DLL).
• This DLL must implement IObjectWithSite.
• The IObjectWithSite::SetSite() method must be implemented. It is through this method that your application receives a pointer to Internet Explorer's IUnknown. Internet Explorer actually passes a pointer to IWebBrowser2 but the implementation of SetSite() receives a pointer to IUnknown. You can use this IUnknown pointer to automate Internet Explorer or to sink events from Internet Explorer.
• It must be registered as a Browser Helper Object as described above.

Java access bridge

To provide native Microsoft Windows assistive technologies access to Java applications without creating a Java native solution, Sun Microsystems provides the "Java Access Bridge." This bridge is loaded as an assistive technology as described in the section on loading assistive technologies for direct navigation of the document object model. The bridge uses a Java Native Invocation (JNI) to Dynamic Link Library (DLL) communication and caching mechanism that allows a native assistive technology to gather and monitor accessibility information in the Java environment. In this environment, the assistive technology determines that a Java application or applet is running and communicates with the Java Access Bridge DLL to process accessibility information about the application/applet running in the Java Virtual Machine.

Loading assistive technologies for indirect access to the document object model

Access to application specific data across process boundaries or address spaces might be costly in terms of performance. However, there are other reasons to consider when accessing the document object model that might lead a developer to wish to access it from their own process or memory address space. One obvious protection this method provides is that, if the user agent fails, it does not disable the user's assistive technology as well. Another consideration would be legacy systems, where the user relies on their assistive technology for access to software other than the user agent, and thus would have their application loaded all the time.

There are several ways to gain access to the user agent's document object model. Most user agents support some kind of external interface, or act as a mini-server to other applications running on the desktop. Internet Explorer [IE-WIN] is a good example of this, as IE can behave as a component object model (COM) server to other applications. Mozilla [MOZILLA], the open source release of Navigator also supports cross-platform COM (XPCOM).

The following example illustrates the use of COM to access the IE object model. This is an example of how to use COM to get a pointer to the WebBrowser2 module, which in turn enables access to an interface/pointer to the document object, or IE document object model for the content.

/* first, get a pointer to the WebBrowser2 control */
if (m_pIE == NULL) {
  hr = CoCreateInstance(CLSID_InternetExplorer, 
       NULL, CLSCTX_LOCAL_SERVER, IID_IWebBrowser2, 
       (void**)&m_pIE);

  /* next, get a interface/pointer to the document in view, 
     this is an interface to the document object model (DOM)*/

  void CHelpdbDlg::Digest_Document() {
   HRESULT hr;
   if (m_pIE != NULL) {
    IDispatch* pDisp;
    hr = m_pIE->QueryInterface(IID_IDispatch, 
                                  (void**) &pDisp);
     if (SUCCEEDED(hr)) {
      IDispatch* lDisp;
      hr = m_pIE->get_Document(&lDisp);
      if (SUCCEEDED(hr)) {
        IHTMLDocument2* pHTMLDocument2;
    hr = lDisp->QueryInterface(IID_IHTMLDocument2,
         (void**) &pHTMLDocument2);
    if (SUCCEEDED(hr)) {
      /* with this interface/pointer, IHTMLDocument2*,
         you can then work on the document */
        IHTMLElementCollection* pColl;
        hr = pHTMLDocument2->get_all(&pColl);
        if (SUCCEEDED(hr)) {

                LONG c_elem;
                hr = pColl->get_length(&c_elem);
                if (SUCCEEDED(hr)) {
           FindElements(c_elem, pColl);
        }
           pColl->Release();
        }
        pHTMLDocument2->Release();
        }
        lDisp->Release();
    }
    pDisp->Release();
      }
    }
  }
}   

For a working example of this method, refer to HelpDB [HELPDB].

4 Glossary

This glossary is normative. However, some terms (or parts of explanations of terms) may not have an impact on conformance.

Note: In this document, glossary terms generally link to the corresponding entries in this section. These terms are also highlighted through style sheets and identified as glossary terms through markup.

Activate

In this document, the verb "to activate" means (depending on context) either:

• To execute or carry out one or more behaviors associated with an enabled element.
• To execute or carry out one or more behaviors associated with a component of the user agent user interface.

The effect of activation depends on the type of the user interface control. For instance, when a link is activated, the user agent generally retrieves the linked Web resource. When a form element is activated, it may change state (e.g., check boxes) or may take user input (e.g., a text entry field).

Alert

In this document, "to alert" means to make the user aware of some event, without requiring acknowledgement. For example, the user agent may alert the user that new content is available on the server by displaying a text message in the user agent's status bar. See checkpoint 1.3 for requirements about alerts.

Animation

In this document, an "animation" refers to content that, when rendered, creates a visual movement effect automatically (i.e., without manual user interaction). This definition of animation includes video and animated images. Animation techniques include:

• graphically displaying a sequence of snapshots within the same region (e.g., as is done for video and animated images). The series of snapshots may be provided by a single resource (e.g., an animated GIF image) or from distinct resources (e.g., a series of images downloaded continuously by the user agent).
• scrolling text (e.g., achieved through markup or style sheets).
• displacing graphical objects around the viewport (e.g., a picture of a ball that is moved around the viewport giving the impression that it is bouncing off of the viewport edges). For instance, the SMIL 2.0 [SMIL20] animation modules explain how to create such animation effects in a declarative manner (i.e., not by composition of successive snapshots).

Applet

An applet is a program (generally written in the Java programming language) that is part of content, and that the user agent executes.

Application Programming Interface (API), conventional input/output/device API

An application programming interface (API) defines how communication may take place between applications.

Implementing APIs that are independent of a particular operating environment (as are the W3C DOM Level 2 specifications) may reduce implementation costs for multi-platform user agents and promote the development of multi-platform assistive technologies. Implementing conventional APIs for a particular operating environment may reduce implementation costs for assistive technology developers who wish to interoperate with more than one piece of software running on that operating environment.

A "device API" defines how communication may take place with an input or output device such as a keyboard, mouse, or video card.

In this document, an "input/output API" defines how applications or devices communicate with a user agent. As used in this document, input and output APIs include, but are not limited to, device APIs. Input and output APIs also include more abstract communication interfaces than those specified by device APIs. A "conventional input/output API" is one that is expected to be implemented by software running on a particular operating environment. For example, on desktop computers today, the conventional input APIs are for the mouse and keyboard. For touch screen devices or mobile devices, conventional input APIs may include stylus, buttons, and voice. The graphical display and sound card are considered conventional output devices for a graphical desktop computer environment, and each has an associated API.

Assistive technology

In the context of this document, an assistive technology is a user agent that:

1. relies on services (such as retrieving Web resources and parsing markup) provided by one or more other "host" user agents. Assistive technologies communicate data and messages with host user agents by using and monitoring APIs.
2. provides services beyond those offered by the host user agents to meet the requirements of users with disabilities. Additional services include alternative renderings (e.g., as synthesized speech or magnified content), alternative input methods (e.g., voice), additional navigation or orientation mechanisms, and content transformations (e.g., to make tables more accessible).

For example, screen reader software is an assistive technology because it relies on browsers or other software to enable Web access, particularly for people with visual and learning disabilities.

Examples of assistive technologies that are important in the context of this document include the following:

• screen magnifiers, which are used by people with visual disabilities to enlarge and change colors on the screen to improve the visual readability of rendered text and images.
• screen readers, which are used by people who are blind or have reading disabilities to read textual information through synthesized speech or braille displays.
• voice recognition software, which may be used by people who have some physical disabilities.
• alternative keyboards, which are used by people with certain physical disabilities to simulate the keyboard.
• alternative pointing devices, which are used by people with certain physical disabilities to simulate mouse pointing and button activations.

Beyond this document, assistive technologies consist of software or hardware that has been specifically designed to assist people with disabilities in carrying out daily activities. These technologies include wheelchairs, reading machines, devices for grasping, text telephones, and vibrating pagers. For example, the following very general definition of "assistive technology device" comes from the (U.S.) Assistive Technology Act of 1998 [AT1998]:

Any item, piece of equipment, or product system, whether acquired commercially, modified, or customized, that is used to increase, maintain, or improve functional capabilities of individuals with disabilities.

Attribute

This document uses the term "attribute" in the XML sense: an element may have a set of attribute specifications (refer to the XML 1.0 specification [XML] section 3).

Audio

In this document, the term "audio" refers to content that encodes prerecorded sound.

Audio-only presentation

An audio-only presentation is content consisting exclusively of one or more audio tracks presented concurrently or in series. Examples of an audio-only presentation include a musical performance, a radio-style news broadcast, and a narration.

Audio track

An audio object is content rendered as sound through an audio viewport. An audio track is an audio object that is intended as a whole or partial presentation. An audio track may, but is not required to, correspond to a single audio channel (left or right audio channel).

Audio description

An audio description (called an "auditory description" in the Web Content Accessibility Guidelines 1.0 [WCAG10]) is either a prerecorded human voice or a synthesized voice (recorded or generated dynamically) describing the key visual elements of a movie or other animation. The audio description is synchronized with (and possibly included as part of) the audio track of the presentation, usually during natural pauses in the audio track. Audio descriptions include information about actions, body language, graphics, and scene changes.

Author styles

Authors styles are style property values that come from content (e.g., style sheets within a document, that are associated with a document, or that are generated by a server).

Captions

Captions are text transcripts that are synchronized with other audio tracks or visual tracks. Captions convey information about spoken words and non-spoken sounds such as sound effects. They benefit people who are deaf or hard-of-hearing, and anyone who cannot hear the audio (e.g., someone in a noisy environment). Captions are generally rendered graphically superimposed ("on top of") the synchronized visual track.

The term "open captions" generally refers to captions that are always rendered with a visual track; they cannot be turned off. The term "closed captions" generally refers to captions that may be turned on and off. The captions requirements of this document assume that the user agent can recognize the captions as such; see the section on applicability for more information.

Note: Other terms that include the word "caption" may have different meanings in this document. For instance, a "table caption" is a title for the table, often positioned graphically above or below the table. In this document, the intended meaning of "caption" will be clear from context.

Character encoding

A "character encoding" is a mapping from a character set definition to the actual code units used to represent the data. Refer to the Unicode specification [UNICODE] for more information about character encodings. Refer to "Character Model for the World Wide Web" [CHARMOD] for additional information about characters and character encodings.

Collated text transcript

A collated text transcript is a text equivalent of a movie or other animation. More specifically, it is the combination of the text transcript of the audio track and the text equivalent of the visual track. For example, a collated text transcript typically includes segments of spoken dialogue interspersed with text descriptions of the key visual elements of a presentation (actions, body language, graphics, and scene changes). See also the definitions of text transcript and audio description. Collated text transcripts are essential for individuals who are deaf-blind.

Conditional content

Conditional content is content that, by format specification, should be made available to users through the user interface, generally under certain conditions (e.g., based on user preferences or operating environment limitations). Some examples of conditional content mechanisms include:

• The "alt" attribute of the IMG element in HTML 4. According to section 13.2 of the HTML 4 specification ([HTML4]): "User agents must render alternate text when they cannot support images, they cannot support a certain image type or when they are configured not to display images.
• OBJECT elements in HTML 4. Section 13.3.1 of the HTML 4 specification ([HTML4]) explains the conditional rendering rules of (nested) OBJECT elements. The rules select among ordered alternatives according to user preferences or error conditions.
• The switch element and test attributes in SMIL 1.0. Sections 4.3 and 4.4, respectively, of SMIL 1.0 [SMIL] explain the conditional rendering rules of these features.
• SVG 1.0 [SVG] also includes a switch element and several attributes for conditional processing.
• The NOSCRIPT and NOFRAMES elements in HTML 4 [HTML4] allow the author to provide content under conditions when the user agent does not support scripts or frames, or the user has turned off support for scripts or frames.

Specifications vary in how completely they define how and when to render conditional content. For instance, the HTML 4 specification includes the rendering conditions for the "alt" attribute, but not for the "title" attribute. The HTML 4 specification does indicate that the "title" attribute should be available to users through the user interface ("Values of the title attribute may be rendered by user agents in a variety of ways...").

Note: The Web Content Accessibility Guidelines 1.0 requires that authors provide text equivalents for non-text content. This is generally done by using the conditional content mechanisms of a markup language. Since conditional content may not be rendered by default, the current document requires the user agent to provide access to unrendered conditional content (checkpoint 2.3 and checkpoint 2.9) as it may have been provided to promote accessibility.

Configure, control

In the context of this document, the verbs "to control" and "to configure" share in common the idea of governance such as a user may exercise over interface layout, user agent behavior, rendering style, and other parameters required by this document. Generally, the difference in the terms centers on the idea of persistence. When a user makes a change by "controlling" a setting, that change usually does not persist beyond that user session. On the other hand, when a user "configures" a setting, that setting typically persists into later user sessions. Furthermore, the term "control" typically means that the change can be made easily (such as through a keyboard shortcut) and that the results of the change occur immediately. The term "configure" typically means that making the change requires more time and effort (such as making the change via a series of menus leading to a dialog box, or via style sheets or scripts). The results of "configuration" might not take effect immediately (e.g., due to time spent reinitializing the system, initiating a new session, or rebooting the system).

In order to be able to configure and control the user agent, the user needs to be able to "write" as well as "read" values for these parameters. Configuration settings may be stored in a profile. The range and granularity of the changes that can be controlled or configured by the user may depend on limitations of the operating environment or hardware.

Both configuration and control may apply at different "levels": across Web resources (i.e., at the user agent level, or inherited from the operating environment), to the entirety of a Web resource, or to components of a Web resource (e.g., on a per-element basis).

A global configuration is one that applies across elements of the same Web resource, as well as across Web resources. A global configuration may be implemented by more than one setting (e.g., per component of the user agent). For instance, when a user agent consists of a browser that renders HTML and a plug-in that renders SVG, to satisfy the global configuration requirements of this document, the browser may provide one setting and the plug-in another.

User agents may allow users to choose configurations based on various parameters, such as hardware capabilities or natural language preferences.

Note: In this document, the noun "control" refers to a user interface control.

Content

In this specification, the noun "content" is used in three ways:

1. It is used to mean the document object as a whole or in parts.
2. It is used to mean the content of an HTML or XML element, in the sense employed by the XML 1.0 specification ([XML], section 3.1): "The text between the start-tag and end-tag is called the element's content." Context should indicate that the term content is being used in this sense.
3. It is used in the terms non-text content and text content.

Empty content (which may be conditional content) is either a null value or an empty string (i.e., one that is zero characters long). For instance, in HTML, alt="" sets the value of the "alt" attribute to the empty string. In some markup languages, an element may have empty content (e.g., the HR element in HTML).

Device-independence

Device-independence refers to the ability to make use of software with any appropriate supported input or output device.

Document object, Document Object Model (DOM)

In general usage, the term "document object" refers to the user agent's representation of data (e.g., a document). This data generally comes from the document source, but may also be generated (e.g., from style sheets, scripts, or transformations), produced as a result of preferences set within the user agent, or added as the result of a repair performed automatically by the user agent. Some data that is part of the document object is routinely rendered (e.g., in HTML, what appears between the start and end tags of elements and the values of attributes such as "alt", "title", and "summary"). Other parts of the document object are generally processed by the user agent without user awareness, such as DTD- or schema-defined names of element types and attributes, and other attribute values such as "href" and "id." These guidelines require that users have access to both kinds of data through the user interface. Most of the requirements of this document apply to the document object after its construction. However, a few checkpoints (e.g., checkpoint 2.7 and checkpoint 2.10) may affect the construction of the document object.

A "document object model" is the abstraction that governs the construction of the user agent's document object. The document object model employed by different user agents may vary in implementation and sometimes in scope. This specification requires that user agents implement the APIs defined in Document Object Model (DOM) Level 2 specifications ([DOM2CORE] and [DOM2STYLE]) for access to HTML, XML, and CSS content. These DOM APIs allow authors to access and modify the content via a scripting language (e.g., JavaScript) in a consistent manner across different scripting languages. As a standard interface, the DOM APIs make it easier not just for authors, but for assistive technology developers to extract information and render it in ways most suited to the needs of particular users.

Document character set

A document character set (a concept taken from SGML) is a sequence of abstract characters that may appear in Web content represented in a particular format (such as HTML, SVG, or SMIL). A document character set consists of:

• A "repertoire": A set of abstract characters, such as the Latin letter "A", the Cyrillic letter "I", and the Chinese character meaning "water."
• Code positions: A set of integer references to characters in the repertoire.

For instance, the character set required by the HTML 4 specification [HTML4] is defined in the Unicode specification [UNICODE]. Refer to "Character Model for the World Wide Web" [CHARMOD] for more information about document character sets.

Document source, text source

In this document, the term "document source" refers to the data that the user agent receives as the direct result of a request for a Web resource (e.g., as the result of an HTTP/1.1 [RFC2616] "GET", or as the result of viewing a resource on the local file system). The document source generally refers to the "payload" of the user agent's request, and does not generally include information exchanged as part of the transfer protocol. The document source is data that is prior to any repair by the user agent (e.g., prior to repairing invalid markup). "Text source" refers to document source that is composed of text.

Documentation

Documentation refers to information that supports the use of a user agent. This information may be found, for example, in manuals, installation instructions, the help system, and tutorials. Documentation may be distributed (e.g., some parts may be delivered on CD-ROM, others on the Web). Refer to guideline 12 for information about documentation requirements.

Element, element type

This document uses the terms "element" and "element type" in the sense employed by the XML 1.0 specification ([XML], section 3): an element type is a syntactic construct of a document type definition (DTD) for its application. This sense is also relevant to structures defined by XML schemas. The document also uses the term "element" more generally to mean a type of content (such as video or sound) or a logical construct (such as a header or list).

Enabled element, disabled element

An enabled element is a piece of content with associated behaviors that may be activated through the user interface or through an API. The set of elements that a user agent enables is generally derived from, but is not limited to, the set of interactive elements defined by implemented markup languages.

Some elements may only be enabled elements for part of a user session. For instance, an element may be disabled by a script as the result of user interaction. Or, an element may only be enabled during a given time period (e.g., during part of a SMIL 1.0 [SMIL] presentation). Or, the user may be viewing content in "read-only" mode, which may disable some elements.

A disabled element is a piece of content that is potentially an enabled element, but is not in the current session. One example of a disabled element is a menu item that is unavailable in the current session; it might be "greyed out" to show that it is disabled. Generally, disabled elements will be interactive elements that are not enabled in the current session. This document distinguishes disabled elements (not currently enabled) from non-interactive elements (never enabled).

For the requirements of this document, user selection does not constitute user interaction with enabled elements. See the definition of content focus.

Note: Enabled and disabled elements come from content; they are not part of the user agent user interface.

Note: The term "active element" is not used in this document since it may suggest several different concepts, including: interactive element, enabled element, an element "in the process of being activated" (which is the meaning of ':active' in CSS2 [CSS2], for example).

Equivalent (for content)

The term "equivalent" is used in this document as it is used in the Web Content Accessibility Guidelines 1.0 [WCAG10]:

Content is "equivalent" to other content when both fulfill essentially the same function or purpose upon presentation to the user. In the context of this document, the equivalent must fulfill essentially the same function for the person with a disability (at least insofar as is feasible, given the nature of the disability and the state of technology), as the primary content does for the person without any disability.

Equivalents include text equivalents (e.g., text equivalents for images, text transcripts for audio tracks, or collated text transcripts for a movie) and non-text equivalents (e.g., a prerecorded audio description of a visual track of a movie, or a sign language video rendition of a written text).

Each markup language defines its own mechanisms for specifying conditional content, and these mechanisms may be used by authors to provide text equivalents. For instance, in HTML 4 [HTML4] or SMIL 1.0 [SMIL], authors may use the "alt" attribute to specify a text equivalent for some elements. In HTML 4, authors may provide equivalents and other conditional content in attribute values (e.g., the "summary" attribute for the TABLE element), in element content (e.g., OBJECT for external content it specifies, NOFRAMES for frame equivalents, and NOSCRIPT for script equivalents), and in prose. Please consult the Web Content Accessibility Guidelines 1.0 [WCAG10] and its associated Techniques document [WCAG10-TECHS] for more information about equivalents.

Events and scripting, event handler, event type

User agents often perform a task when an event having a particular "event type" occurs, including user interface events, changes to content, loading of content, and requests from the operating environment. Some markup languages allow authors to specify that a script, called an event handler, be executed when an event of a given type occurs. An event handler is explicitly associated with an element when the event handler is associated with that element through markup or the DOM. The term "event bubbling" describes a programming style where a single event handler dispatches events to more than one element. In this case, the event handlers are not explicitly associated with the elements receiving the events (except for the single element that dispatches the events).

Note: The combination of HTML, style sheets, the Document Object Model (DOM), and scripting is commonly referred to as "Dynamic HTML" or DHTML. However, as there is no W3C specification that formally defines DHTML, this document only refers to event handlers and scripts.

Explicit user request

In this document, the term "explicit user request" refers to any user interaction through the user agent user interface (not through rendered content), the focus, or the selection. User requests are made, for example, through user agent user interface controls and keyboard bindings.

Some examples of explicit user requests include when the user selects "New viewport", responds "Yes" to a prompt in the user agent's user interface, configures the user agent to behave in a certain way, or changes the selection or focus with the keyboard or pointing device.

Note: Users make mistakes. For example, a user may inadvertently respond "yes" to a prompt when they meant "no." In this document, this type of mistake is still considered an explicit user request.

Focus, content focus, user interface focus, current focus

In this document, the term "content focus" (required by checkpoint 9.1) refers to a user agent mechanism that has all of the following properties:

1. It designates zero or one element in content that is either enabled or disabled. In general, the focus should only designate enabled elements, but it may also designate disabled elements.
2. It has state: the user may "set it" (programmatically or through the user interface) on an enabled element. Events may be triggered when the focus is set (or unset). Which events are triggered depends on the content (e.g., HTML events and CSS pseudo-classes) or user interface settings.
3. Once it has been set, it may be used to trigger other behaviors associated with the enabled element (e.g., the user may activate a link or change the state of a form control). These behaviors may be triggered programmatically or through the user interface (e.g., through keyboard events).

User interface mechanisms may resemble content focus, but do not satisfy all of the properties. For example, text editors often implement a "caret" that indicates the current location of text input or editing. The caret may have state and may respond to input device events, but it does not enable users to activate the behaviors associated with enabled elements.

The user interface focus shares the properties of the content focus except that, rather than designating pieces of content, it designates zero or one control of the user agent user interface that has associated behaviors (e.g., a radio button, text box, or menu).

On the screen, the content focus may be highlighted in a variety of ways, including through colors, fonts, graphics, and magnification. The content focus may also be highlighted when rendered as synthesized speech, for example through changes in speech prosody. The dimensions of the rendered content focus may exceed those of the viewport.

In this document, each viewport is expected to have at most one content focus and at most one user interface focus. This document includes requirements for content focus only, for user interface focus only, and for both. When a requirement refers to both, the term "focus" is used.

When several viewports coexist, at most one viewport's content focus or user interface focus responds to input events; this is called the current focus.

Graphical

In this document, the term "graphical" refers to information (including text, colors, graphics, images, and animations) rendered for visual consumption.

Highlight

In this document, "to highlight" means to emphasize through the user interface. For example, user agents highlight which content is selected or focused. Graphical highlight mechanisms include dotted boxes, underlining, and reverse video. Synthesized speech highlight mechanisms include alterations of voice pitch and volume ("speech prosody").

Image

This document uses the term "image" to refer (as is commonly the case) to pictorial content. However, in this document, term image is limited to static (i.e., unmoving) visual information. See also the definition of animation.

Input configuration

An input configuration is the set of "bindings" between user agent functionalities and user interface input mechanisms (e.g., menus, buttons, keyboard keys, and voice commands). The default input configuration is the set of bindings the user finds after installation of the software; it must be documented (per checkpoint 12.3]). Input configurations may be affected by author-specified bindings (e.g., through the "accesskey" attribute of HTML 4 [HTML4]).

Interactive element, non-interactive element,

An interactive element is piece of content that, by specification, may have associated behaviors to be executed or carried out as a result of user or programmatic interaction. For instance, the interactive elements of HTML 4 [HTML4] include: links, image maps, form elements, elements with a value for the "longdesc" attribute, and elements with event handlers explicitly associated with them (e.g., through the various "on" attributes). The role of an element as an interactive element is subject to applicability. A non-interactive element is an element that, by format specification, does not have associated behaviors. The expectation of this document is that interactive elements become enabled elements in some sessions, and non-interactive elements never become enabled elements.

Natural language

Natural language is spoken, written, or signed human language such as French, Japanese, and American Sign Language. On the Web, the natural language of content may be specified by markup or HTTP headers. Some examples include the "lang" attribute in HTML 4 ([HTML4] section 8.1), the "xml:lang" attribute in XML 1.0 ([XML], section 2.12), the HTML 4 "hreflang" attribute for links in HTML 4 ([HTML4], section 12.1.5), the HTTP Content-Language header ([RFC2616], section 14.12) and the Accept-Language request header ([RFC2616], section 14.4). See also the definition of script.

Normative, informative

What is identified as "normative" is required for conformance (noting that one may conform in a variety of well-defined ways to this document). What is identified as "informative" (sometimes, "non-normative") is never required for conformance.

Operating environment

The term "operating environment" refers to the environment that governs the user agent's operation, whether it is an operating system or a programming language environment such as Java.

Override

In this document, the term "override" means that one configuration or behavior preference prevails over another. Generally, the requirements of this document involve user preferences prevailing over author preferences and user agent default settings and behaviors. Preferences may be multi-valued in general (e.g., the user prefers blue over red or yellow), and include the special case of two values (e.g., turn on or off blinking text content).

Placeholder

A placeholder is content generated by the user agent to replace author-supplied content. A placeholder may be generated as the result of a user preference (e.g., to not render images) or as repair content (e.g., when an image cannot be found). Placeholders can be any type of content, including text, images, and audio cues.

This document includes requirements that the user be able to view the original author-supplied content associated with a placeholder. To satisfy these requirements, the user agent might render the content in place of the placeholder or in a separate viewport (leaving the placeholder as is). A request to view the original content associated with a placeholder is considered an explicit user request to render that content.

This document does not require user agents to include placeholders in the document object. A placeholder that is inserted in the document object should conform to the Web Content Accessibility Guidelines 1.0 [WCAG10]. If a placeholder is not part of the document object, it is part of the user interface only (and subject, for example, to checkpoint 1.3).

Plug-in

A plug-in is a program that runs as part of the user agent and that is not part of content. Users generally choose to include or exclude plug-ins from their user agent.

Point of regard

The point of regard is a position in rendered content that the user is presumed to be viewing. The dimensions of the point of regard may vary. For example, it may be a point (e.g., a moment in an audio rendering or a cursor in a graphical rendering), or a range of text (e.g., focused text), or a two-dimensional area (e.g., content rendered through a two-dimensional graphical viewport). The point of regard is almost always within the viewport, but it may exceed the spatial or temporal dimensions of the viewport (see the definition of rendered content for more information about viewport dimensions). The point of regard may also refer to a particular moment in time for content that changes over time (e.g., an audio-only presentation). User agents may determine the point of regard in a number of ways, including based on viewport position in content, content focus, and selection. A user agent should not change the point of regard unexpectedly as this may disorient the user. The point of regard should be available programmatically (e.g., for assistive technologies).

Profile

A profile is a named and persistent representation of user preferences that may be used to configure a user agent. Preferences include input configurations, style preferences, and natural language preferences. In operating environments with distinct user accounts, profiles enable users to reconfigure software quickly when they log on, and profiles may be shared by several users. Platform-independent profiles are useful for those who use the same user agent on different platforms.

Prompt

In this document, "to prompt" means to require input from the user. The user agent should allow users to configure how they wish to be prompted. For instance, for a user agent functionality X, configurations might include: "always prompt me before doing X," "always do X without prompting me," "never do X but tell me when you could have," and "never do X and never tell me that you could have."

Properties, values, and defaults

A user agent renders a document by applying formatting algorithms and style information to the document's elements. Formatting depends on a number of factors, including where the document is rendered: on screen, on paper, through loudspeakers, on a braille display, or on a mobile device. Style information (e.g., fonts, colors, and synthesized speech prosody) may come from the elements themselves (e.g., certain font and phrase elements in HTML), from style sheets, or from user agent settings. For the purposes of these guidelines, each formatting or style option is governed by a property and each property may take one value from a set of legal values. Generally in this document, the term "property" has the meaning defined in CSS 2 ([CSS2], section 3). A reference to "styles" in this document means a set of style-related properties.

The value given to a property by a user agent when it is installed is called the property's default value.

Recognize

Authors encode information in many ways, including in markup languages, style sheet languages, scripting languages, and protocols. When the information is encoded in a manner that allows the user agent to process it with certainty, the user agent can "recognize" the information. For instance, HTML allows authors to specify a heading with the H1 element, so a user agent that implements HTML can recognize that content as a heading. If the author creates headings using a visual effect alone (e.g., by increasing the font size), then the author has encoded the heading in a manner that does not allow the user agent to recognize it as a heading.

Some requirements of this document depend on content roles, content relationships, timing relationships, and other information supplied by the author. These requirements only apply when the author has encoded that information in a manner that the user agent can recognize. See the section on conformance in User Agent Accessibility Guidelines 1.0 [UAAG10] for more information about applicability.

In practice, user agents will rely heavily on information that the author has encoded in a markup language or style sheet language. On the other hand, behaviors, style, meaning encoded in a script, and markup in an unfamiliar XML namespace may not be recognized by the user agent as easily or at all.

Rendered content, rendered text

Rendered content is the part of content that the user agent makes available to the user's senses of sight and hearing (and only those senses for the purposes of this document). Any content that causes an effect that may be perceived through these senses constitutes rendered content. This includes text characters, images, style sheets, scripts, and anything else in content that, once processed, may be perceived through sight and hearing.

The term "rendered text" refers to text content that is rendered in a way that communicates information about the characters themselves, whether visually or as synthesized speech.

In the context of this document, invisible content is content that influences graphical rendering of other content but is not rendered itself. Similarly, silent content is content that influences audio rendering of other content but is not rendered itself. Neither invisible nor silent content is considered rendered content.

Repair content, repair text

In this document, the term "repair content" refers to content generated by the user agent in order to correct an error condition. "Repair text" means repair content consisting only of text. Some error conditions that may lead to the generation of repair content include:

• Erroneous or incomplete content (e.g., ill-formed markup, invalid markup, missing conditional content that is required by format specification);
• Missing resources for handling or rendering content (e.g., the user agent lacks a font family to display some characters, the user agent does not implement a particular scripting language).

This document does not require user agents to include repair content in the document object. Repair content inserted in the document object should conform to the Web Content Accessibility Guidelines 1.0 [WCAG10]. For more information about repair techniques for Web content and software, refer to "Techniques for Authoring Tool Accessibility Guidelines 1.0" [ATAG10-TECHS].

Script

In this document, the term "script" almost always refers to a scripting (programming) language used to create dynamic Web content. However, in checkpoints referring to the written (natural) language of content, the term "script" is used as in Unicode [UNICODE] to mean "A collection of symbols used to represent textual information in one or more writing systems."

Information encoded in scripts may be difficult for a user agent to recognize. For instance, a user agent is not expected to recognize that, when executed, a script will calculate a factorial. The user agent will be able to recognize some information in a script by virtue of implementing the scripting language or a known program library (e.g., the user agent is expected to recognize when a script will open a viewport or retrieve a resource from the Web).

Selection, current selection

In this document, the term "selection" refers to a user agent mechanism for identifying a range of content (e.g., text and images). Generally, user agents limit selection to text content (e.g., one or more fragments of text). The selection may be structured (based on the document tree) or unstructured (e.g., text-based). The range may be empty.

On the screen, the selection may be highlighted in a variety of ways, including through colors, fonts, graphics, and magnification. The selection may also be highlighted when rendered as synthesized speech, for example through changes in speech prosody. The dimensions of the rendered selection may exceed those of the viewport.

The selection may be used for a variety of purposes, including for cut and paste operations, to designate a specific element in a document for the purposes of a query, and as an indication of point of regard.

The selection has state, and the user may "set it" (programmatically or through the user interface).

In this document, each viewport is expected to have at most one selection. When several viewports coexist, at most one viewport's selection responds to input events; this is called the current selection.

See the section on the Selection label for information about implementing a selection and conformance.

Note: Some user agents may also implement a selection for designating a range of information in the user agent user interface. The current document only includes requirements for a content selection mechanism.

Serial access, sequential navigation

In this document, the expression "serial access" refers to one-dimensional access to rendered content. Some examples of serial access include listening to an audio stream or watching a video (both of which involve one temporal dimension), or reading a series of lines of braille one line at a time (one spatial dimension). Many users with blindness have serial access to content rendered as audio, synthesized speech, or lines of braille.

The expression "sequential navigation" refers to navigation through an ordered set of items (e.g., the enabled elements in a document, a sequence of lines or pages, or a sequence of menu options). Sequential navigation implies that the user cannot skip directly from one member of the set to another, in contrast to direct or structured navigation (see guideline 9 for information about these types of navigation). Users with blindness or some users with a physical disability may navigate content sequentially (e.g., by navigating through links, one by one, in a graphical viewport with or without the aid of an assistive technology). Sequential navigation is important to users who cannot scan rendered content visually for context and also benefits users unfamiliar with content. The increments of sequential navigation may be determined by a number of factors, including element type (e.g., links only), content structure (e.g., navigation from heading to heading), and the current navigation context (e.g., having navigated to a table, allow navigation among the table cells).

Users with serial access to content or who navigate sequentially may require more time to access content than users who use direct or structured navigation.

Support, implement, conform

In this document, the terms "support", "implement", and "conform" all refer to what a developer has designed a user agent to do, but they represent different degrees of specificity. A user agent "supports" general classes of objects, such as "images" or "Japanese". A user agent "implements" a specification (e.g., the PNG and SVG image format specifications or a particular scripting language), or an API (e.g., the DOM API) when it has been programmed to follow all or part of a specification. A user agent "conforms to" a specification when it implements the specification and satisfies its conformance criteria. This document includes some conformance requirements to other specifications (e.g., to a particular level of the "Web Content Accessibility Guidelines 1.0" [WCAG10]).

Synchronize

In this document, "to synchronize" refers to the act of time-coordinating two or more presentation components (e.g., in a multimedia presentation, a visual track with captions). For Web content developers, the requirement to synchronize means to provide the data that will permit sensible time-coordinated rendering by a user agent. For example, Web content developers can ensure that the segments of caption text are neither too long nor too short, and that they map to segments of the visual track that are appropriate in length. For user agent developers, the requirement to synchronize means to present the content in a sensible time-coordinated fashion under a wide range of circumstances including technology constraints (e.g., small text-only displays), user limitations (slow reading speeds, large font sizes, high need for review or repeat functions), and content that is sub-optimal in terms of accessibility.

Text

In this document, the term "text" used by itself refers to a sequence of characters from a markup language's document character set. Refer to the "Character Model for the World Wide Web " [CHARMOD] for more information about text and characters. Note: This document makes use of other terms that include the word "text" that have highly specialized meanings: collated text transcript, non-text content, text content, non-text element, text element, text equivalent, and text transcript.

Text content, non-text content, text element, non-text element, text equivalent, non-text equivalent

As used in this document a "text element" adds text characters to either content or the user interface. Both in the Web Content Accessibility Guidelines 1.0 [WCAG10] and in this document, text elements are presumed to produce text that can be understood when rendered visually, as synthesized speech, or as Braille. Such text elements benefit at least these three groups of users:

1. visually-displayed text benefits users who are deaf and adept in reading visually-displayed text;
2. synthesized speech benefits users who are blind and adept in use of synthesized speech;
3. braille benefits users who are blind, and possibly deaf-blind, and adept at reading braille.

A text element may consist of both text and non-text data. For instance, a text element may contain markup for style (e.g., font size or color), structure (e.g., heading levels), and other semantics. The essential function of the text element should be retained even if style information happens to be lost in rendering.

A user agent may have to process a text element in order to have access to the text characters. For instance, a text element may consist of markup, it may be encrypted or compressed, or it may include embedded text in a binary format (e.g., JPEG).

"Text content" is content that is composed of one or more text elements. A "text equivalent" (whether in content or the user interface) is an equivalent composed of one or more text elements. Authors generally provide text equivalents for content by using the conditional content mechanisms of a specification.

A "non-text element" is an element (in content or the user interface) that does not have the qualities of a text element. "Non-text content" is composed of one or more non-text elements. A "non-text equivalent" (whether in content or the user interface) is an equivalent composed of one or more non-text elements.

Text decoration

In this document, a "text decoration" is any stylistic effect that the user agent may apply to visually rendered text that does not affect the layout of the document (i.e., does not require reformatting when applied or removed). Text decoration mechanisms include underline, overline, and strike-through.

Text transcript

A text transcript is a text equivalent of audio information (e.g., an audio-only presentation or the audio track of a movie or other animation). It provides text for both spoken words and non-spoken sounds such as sound effects. Text transcripts make audio information accessible to people who have hearing disabilities and to people who cannot play the audio. Text transcripts are usually created by hand but may be generated on the fly (e.g., by voice-to-text converters). See also the definitions of captions and collated text transcripts.

User agent

In this document, the term "user agent" is used in two ways:

1. The software and documentation components that together, conform to the requirements of User Agent Accessibility Guidelines 1.0 [UAAG10]. This is the most common use of the term in this document and is the usage in the checkpoints.
2. Any software that retrieves and renders Web content for users. This may include Web browsers, media players, plug-ins, and other programs – including assistive technologies -- that help in retrieving and rendering Web content.

User agent default styles

User agent default styles are style property values applied in the absence of any author or user styles. Some markup languages specify a default rendering for documents in that markup language. Other specifications may not specify default styles. For example, XML 1.0 [XML] does not specify default styles for XML documents. HTML 4 [HTML4] does not specify default styles for HTML documents, but the CSS 2 [CSS2] specification suggests a sample default style sheet for HTML 4 based on current practice.

User interface, user interface control

For the purposes of this document, user interface includes both:

1. the user agent user interface, i.e., the controls (e.g., menus, buttons, prompts, and other components for input and output) and mechanisms (e.g., selection and focus) provided by the user agent ("out of the box") that are not created by content.
2. the "content user interface", i.e., the enabled elements that are part of content, such as form controls, links, and applets.

The document distinguishes them only where required for clarity. For more information, see the section on requirements for content, for user agent features, or both in User Agent Accessibility Guidelines 1.0 [UAAG10].

The term "user interface control" refers to a component of the user agent user interface or the content user interface, distinguished where necessary.

User styles

User styles are style property values that come from user interface settings, user style sheets, or other user interactions.

Views, viewports

The user agent renders content through one or more viewports. Viewports include windows, frames, pieces of paper, loudspeakers, and virtual magnifying glasses. A viewport may contain another viewport (e.g., nested frames). User agent user interface controls such as prompts, menus, and alerts are not viewports.

Graphical and tactile viewports have two spatial dimensions. A viewport may also have temporal dimensions, for instance when audio, speech, animations, and movies are rendered. When the dimensions (spatial or temporal) of rendered content exceed the dimensions of the viewport, the user agent provides mechanisms such as scroll bars and advance and rewind controls so that the user can access the rendered content "outside" the viewport. Examples include: when the user can only view a portion of a large document through a small graphical viewport, or when audio content has already been played.

When several viewports coexist, only one has the current focus at a given moment. This viewport is highlighted to make it stand out.

User agents may render the same content in a variety of ways; each rendering is called a view. For instance, a user agent may allow users to view an entire document or just a list of the document's headers. These are two different views of the document.

Visual-only presentation

A visual-only presentation is content consisting exclusively of one or more visual tracks presented concurrently or in series. A silent movie is an example of a visual-only presentation.

Visual track

A visual object is content rendered through a graphical viewport. Visual objects include graphics, text, and visual portions of movies and other animations. A visual track is a visual object that is intended as a whole or partial presentation. A visual track does not necessarily correspond to a single physical object or software object.

Voice browser

From "Introduction and Overview of W3C Speech Interface Framework" [VOICEBROWSER]: "A voice browser is a device (hardware and software) that interprets voice markup languages to generate voice output, interpret voice input, and possibly accept and produce other modalities of input and output."

Web resource

The term "Web resource" is used in this document in accordance with Web Characterization Terminology and Definitions Sheet [WEBCHAR] to mean anything that can be identified by a Uniform Resource Identifier (URI); refer to RFC 2396 [RFC2396].

5 References

For the latest version of any W3C specification please consult the list of W3C Technical Reports at http://www.w3.org/TR/. Some documents listed below may have been superseded since the publication of this document.

Note: In this document, bracketed labels such as "[HTML4]" link to the corresponding entries in this section. These labels are also identified as references through markup.

5.1 How to refer to this document

There are two recommended ways to refer to the "Techniques for User Agent Accessibility Guidelines 1.0" (and to W3C documents in general):

1. References to a specific version of "Techniques for User Agent Accessibility Guidelines 1.0". For example, use the "this version" URI to refer to the current document: http://www.w3.org/TR/2002/WD-UAAG10-TECHS-20021016/.
2. References to the latest version of "Techniques for User Agent Accessibility Guidelines 1.0". Use the "latest version" URI to refer to the most recently published document in the series: http://www.w3.org/TR/UAAG10-TECHS/.

In almost all cases, references (either by name or by link) should be to a specific version of the document. W3C will make every effort to make this document indefinitely available at its original address in its original form. The top of this document includes the relevant catalog metadata for specific references (including title, publication date, "this version" URI, editors' names, and copyright information).

<p>
<cite><a href="http://www.w3.org/TR/2002/WD-UAAG10-TECHS-20021016/">
"Techniques for User Agent Accessibility Guidelines 1.0"</a></cite>,
I. Jacobs, J. Gunderson, E. Hansen, eds.,
W3C Working Draft, 16 October 2002.
The <a href="http://www.w3.org/TR/UAAG10-TECHS/">latest
version</a> of this document is available at
http://www.w3.org/TR/UAAG10-TECHS/.</p>

For very general references to this document (where stability of content and anchors is not required), it may be appropriate to refer to the latest version of this document. In this case, please use the "latest version" URI at the top of this document.

5.2 Normative references

[CSS1]

"Cascading Style Sheets (CSS1) Level 1 Specification", B. Bos, H. Wium Lie, eds., 17 December 1996, revised 11 January 1999. This W3C Recommendation is http://www.w3.org/TR/1999/REC-CSS1-19990111.

[CSS2]

"Cascading Style Sheets, level 2 (CSS2) Specification", B. Bos, H. Wium Lie, C. Lilley, and I. Jacobs, eds., 12 May 1998. This W3C Recommendation is http://www.w3.org/TR/1998/REC-CSS2-19980512/.

[DOM2CORE]

"Document Object Model (DOM) Level 2 Core Specification", A. Le Hors, P. Le Hégaret, L. Wood, G. Nicol, J. Robie, M. Champion, S. Byrne, eds., 13 November 2000. This W3C Recommendation is http://www.w3.org/TR/2000/REC-DOM-Level-2-Core-20001113/.

[DOM2STYLE]

"Document Object Model (DOM) Level 2 Style Specification", V. Apparao, P. Le Hégaret, C. Wilson, eds., 13 November 2000. This W3C Recommendation is http://www.w3.org/TR/2000/REC-DOM-Level-2-Style-20001113/.

[ECMASCRIPT]

"ECMAScript Language Specification", European Computer Manufacturers Association, December 1999. This specification is available at http://www.ecma.ch/ecma1/STAND/ECMA-262.HTM.

[INFOSET]

"XML Information Set", J. Cowan and R. Tobin, eds., 24 October 2001. This W3C Recommendation is http://www.w3.org/TR/2001/REC-xml-infoset-20011024/.

[JAVA]

"The Java Language Specification", Sun Microsystems Inc., J. Gosling, B. Joy, and G. Steele, September 1996. The specification is available at http://java.sun.com/docs/books/jls.

[RFC2046]

"Multipurpose Internet Mail Extensions (MIME) Part Two: Media Types", N. Freed, N. Borenstein, November 1996.

[UAAG10]

"User Agent Accessibility Guidelines 1.0", I. Jacobs, J. Gunderson, E. Hansen, eds. The latest draft of the guidelines is available at http://www.w3.org/TR/UAAG10/.

[WCAG10]

"Web Content Accessibility Guidelines 1.0", W. Chisholm, G. Vanderheiden, and I. Jacobs, eds., 5 May 1999. This W3C Recommendation is http://www.w3.org/TR/1999/WAI-WEBCONTENT-19990505/.

[XML]

"Extensible Markup Language (XML) 1.0 (Second Edition)", T. Bray, J. Paoli, C.M. Sperberg-McQueen, eds., 6 October 2000. This W3C Recommendation is http://www.w3.org/TR/2000/REC-xml-20001006.

5.3 Informative references

Some of the references in this section become normative if they are used to satisfy the requirements of guideline 6 and guideline 8.

[AT1998]

The Assistive Technology Act of 1998, 13 November 1998, United States P.L. 105-394.

[ATAG10]

"Authoring Tool Accessibility Guidelines 1.0", J. Treviranus, C. McCathieNevile, I. Jacobs, and J. Richards, eds., 3 February 2000. This W3C Recommendation is http://www.w3.org/TR/2000/REC-ATAG10-20000203/.

[ATAG10-TECHS]

"Techniques for Authoring Tool Accessibility Guidelines 1.0", J. Treviranus, C. McCathieNevile, I. Jacobs, and J. Richards, eds., 4 May 2000. This W3C Note is http://www.w3.org/TR/2000/NOTE-ATAG10-TECHS-20000504/.

[CHARMOD]

"Character Model for the World Wide Web", M. Dürst and F. Yergeau, eds., 30 April 2002. This W3C Working Draft is http://www.w3.org/TR/2002/WD-charmod-20020430/. The latest version is available at http://www.w3.org/TR/charmod/.

[CSS-ACCESS]

"Accessibility Features of CSS", I. Jacobs, J. Brewer, 4 August 1999. This W3C Note is http://www.w3.org/1999/08/NOTE-CSS-access-19990804.

[DOM2EVENTS]

Document Object Model (DOM) Level 2 Events Specification, T. Pixley, ed., 13 November 2000. This W3C Recommendation is http://www.w3.org/TR/2000/REC-DOM-Level-2-Events-20001113/.

[DOM2RANGE]

Document Object Model (DOM) Level 2 Traversal and Range Specification, J. Kesselman, J. Robie, M. Champion, P. Sharpe, V. Apparao, and L. Wood, eds., 13 November 2000. This W3C Recommendation is http://www.w3.org/TR/2000/REC-DOM-Level-2-Traversal-Range-20001113/.

[DOM3EVENTS]

Document Object Model (DOM) Level 3 Events Specification, T. Pixley, ed., 8 February 2002. This W3C Working Draft is http://www.w3.org/TR/2002/WD-DOM-Level-3-Events-20020208/. The latest version is available at http://www.w3.org/TR/DOM-Level-3-Events/.

[HTML4]

"HTML 4.01 Recommendation", D. Raggett, A. Le Hors, and I. Jacobs, eds., 24 December 1999. This W3C Recommendation is http://www.w3.org/TR/1999/REC-html401-19991224/.

[MATHML20]

"Mathematical Markup Language (MathML) Version 2.0", D. Carlisle, P. Ion, R. Miner, N. Poppelier, et al., 21 February 2001. This W3C Recommendation is http://www.w3.org/TR/2001/REC-MathML2-20010221/.

[PNG]

"PNG (Portable Network Graphics) Specification 1.0", T. Boutell, ed., 1 October 1996. This W3C Recommendation is http://www.w3.org/TR/REC-png.

[PWD-USE-WEB]

The "How People with Disabilities Use the Web", J. Brewer. This document provides an introduction to use of the Web by people with disabilities. It is not yet a formal W3C Working Draft.

[RFC2396]

"Uniform Resource Identifiers (URI): Generic Syntax", T. Berners-Lee, R. Fielding, L. Masinter, August 1998.

[RFC2616]

"Hypertext Transfer Protocol – HTTP/1.1", J. Gettys, J. Mogul, H. Frystyk, L. Masinter, P. Leach, T. Berners-Lee, June 1999.

[RFC3023]

"XML Media Types", M. Murata, S. St. Laurent, D. Kohn, January 2001.

[SMIL]

"Synchronized Multimedia Integration Language (SMIL) 1.0 Specification", P. Hoschka, ed., 15 June 1998. This W3C Recommendation is http://www.w3.org/TR/1998/REC-smil-19980615/.

[SMIL-ACCESS]

"Accessibility Features of SMIL", M-R. Koivunen, I. Jacobs, 21 September 1999. This W3C Note is http://www.w3.org/TR/1999/NOTE-SMIL-access-19990921/.

[SMIL20]

Synchronized Multimedia Integration Language (SMIL 2.0) Specification, J. Ayars, et al., eds., 7 August 2001. This W3C Recommendation is http://www.w3.org/TR/2001/REC-smil20-20010807/.

[SVG]

"Scalable Vector Graphics (SVG) 1.0 Specification", J. Ferraiolo, ed., 2 August 2000. This W3C Candidate Recommendation is http://www.w3.org/TR/2000/CR-SVG-20000802/.

[SVG-ACCESS]

"Accessibility Features of SVG", C. McCathieNevile and M.-R. Koivunen, 7 August 2000. This W3C Note is http://www.w3.org/TR/2000/NOTE-SVG-access-20000807/.

[UNICODE]

"The Unicode Standard, Version 3.2". This technical report of the Unicode Consortium is available at http://www.unicode.org/unicode/reports/tr28/. This is a revision of "The Unicode Standard, Version 3.0", The Unicode Consortium, Addison-Wesley Developers Press, 2000. ISBN 0-201-61633-5. Refer also to http://www.unicode.org/unicode/standard/versions/. For information about character encodings, refer to Unicode Technical Report #17 "Character Encoding Model".

[VOICEBROWSER]

"Introduction and Overview of W3C Speech Interface Framework", J. Larson, 4 December 2000. This W3C Working Draft is http://www.w3.org/TR/2000/WD-voice-intro-20001204/. The latest version is available at http://www.w3.org/TR/voice-intro/. This document includes references to additional W3C specifications about voice browser technology.

[WCAG10-TECHS]

"Techniques for Web Content Accessibility Guidelines 1.0", W. Chisholm, G. Vanderheiden, and I. Jacobs, eds., 6 November 2000. This W3C Note is http://www.w3.org/TR/1999/WAI-WEBCONTENT-TECHS-19990505/. The latest version is available at http://www.w3.org/TR/WCAG10-TECHS/. Additional format-specific techniques documents are available from this Note.

[WEBCHAR]

"Web Characterization Terminology and Definitions Sheet", B. Lavoie, H. F. Nielsen, eds., 24 May 1999. This is a W3C Working Draft that defines some terms to establish a common understanding about key Web concepts. This W3C Working Draft is http://www.w3.org/1999/05/WCA-terms/01.

[XFORMS10]

"XForms 1.0", M. Dubinko, J. Dietl, R. Merrick, T.V. Raman, 18 January 2002. This W3C Working Draft is http://www.w3.org/TR/2002/WD-xforms-20020118/. The latest version of XForms 1.0 is available at http://www.w3.org/TR/xforms/.

[XHTML10]

"XHTML[tm] 1.0: The Extensible HyperText Markup Language", S. Pemberton, et al., 26 January 2000. This W3C Recommendation is http://www.w3.org/TR/2000/REC-xhtml1-20000126/.

[XLINK]

"XML Linking Language (XLink) Version 1.0", S. DeRose, E. Maler, D. Orchard, B. Trafford, eds., 3 July 2000. This XML 1.0 Candidate Recommendation is http://www.w3.org/TR/2000/CR-xlink-20000703/.

[XMLSTYLE]

"Associating Style Sheets with XML documents Version 1.0", J. Clark, ed., 29 June 1999. This W3C Recommendation is http://www.w3.org/1999/06/REC-xml-stylesheet-19990629/

[XSLT]

"XSL Transformations (XSLT) Version 1.0", J. Clark, 16 November 1999. This W3C Recommendation is http://www.w3.org/TR/1999/REC-xslt-19991116.

6 Resources

Note: W3C does not guarantee the stability of any of the following references outside of its control. These references are included for convenience. References to products are not endorsements of those products by W3C.