The W3C Web Accessibility Initiative Protocols & Formats Working Group has posted a working draft of Accessible Rich Internet Applications (WAI-ARIA) Version 1.0:
The domain of Web accessibility defines how to make Web content usable by people with disabilities. People with some types of disabilities use Assistive Technology (AT) to interact with content. AT can transform the presentation of content into a format more suitable to the user, and can allow the user to interact in different ways than the author designed. In order to accomplish this, AT must understand the semantics of the content. Semantics are knowledge of roles, states, and properties, as a person would understand them, that apply to elements within the content. For instance, if a paragraph is semantically identified as such, AT can interact with it as a unit separable from the rest of the content, knowing the exact boundaries of that paragraph. A slider or tree widget is a more complex example, in which various parts of a widget each have semantics that must be properly identified for the computer to support effective interaction.
Established content technologies define semantics for elements commonly used in those technologies. However, new technologies can overlook some of the semantics required for accessibility. Furthermore, new authoring practices evolve which override the intended semantics—elements that have one defined semantic meaning in the technology are used with a different semantic meaning intended to be understood by the user.
For example, Rich Internet Applications developers can create a tree widget in HTML using CSS and JavaScript even though HTML lacks a semantic element for that. A different element must be used, possibly a list element with display instructions to make it look and behave like a tree widget. Assistive technology, however, must present the element in a different modality and the display instructions may not be applicable. The AT will present it as a list, which has very different display and interaction from a tree widget, and the user may be unsuccessful at understanding and operating the widget.
The incorporation of WAI-ARIA is a way for an author to provide proper type semantics on custom widgets (elements with repurposed semantics) to make these widgets accessible, usable and interoperable with assistive technologies. This specification identifies the types of widgets and structures that are recognized by accessibility products, by providing an ontology of corresponding roles that can be attached to content. This allows elements with a given role to be understood as a particular widget or structural type regardless of any semantic inherited from the implementing technology. Roles are a common property of platform Accessibility APIs which applications use to support assistive technologies. Assistive technology can then use the role information to provide effective presentation and interaction with these elements.
This role taxonomy currently includes interaction widget (user interface widget) and structural document (content organization) types of objects. The role taxonomy describes inheritance (widgets that are types of other widgets) and details what states and properties each role supports. When possible, information is provided about mapping of roles to accessibility APIs.
Roles are element types and should not change with time or user actions. Changing the role on an element from its inital value will be treated, via accessibility API events, as the removal of the old element and insertion of a new element with the new role.
Changeable states and properties of elements are also defined in this specification. States and Properties are used to declare important properties of an element that affect and describe interaction. These properties enable the user agent or operating system to properly handle the element even when these properties are altered dynamically by scripts. For example, alternative input and output technology such as screen readers, speech dictation software and on-screen keyboards must recognize the state of an element (such as: if an object is disabled, checked, focused, collapsed, hidden, etc.).
While it is possible for assistive technologies to access these properties through the Document Object Model [DOM], the preferred mechanism is for the user agent to map the States and Properties to the accessibility API of the operating system.
