This is an archived snapshot of W3C's public bugzilla bug tracker, decommissioned in April 2019. Please see the home page for more details.
The spec reads: "For traditional static Web pages, assistive technologies, such as screen readers, interact with user agents using the DOM. For UI elements that are known to be interactive, such as HTML form elements and desktop applications, assistive technologies may use platform accessibility APIs." This language is out of date and should be deprecated. As is clear from other bugs, simply reading the DOM is not sufficient, even for a static web page with no interactive content. That misses out on CSS generated content, and other rendering changes affected by CSS, for example. The spec should mandate that user agents provide full access to an accessibility tree that contains an accessible representation of all content actually displayed and rendered.
FWIW, for ATK/AT-SPI2-based assistive technologies, all access is provided through the platform's accessibility APIs. Even for static web pages with no CSS whatsoever. The DOM is never used.
I wouldn't restrict the assistive technologies developer in ways to operate the web content. I'm aware of at least one recent project preferring to work with DOM over high level a11y API. I'd say if they want to go their own way then we shouldn't make them out of law. On the another hand we don't have cross browser web accessibility API so any a11y web app (like ChromeVox?) has to deal with DOM this or that way. I agree the spec wording should be updated but direct DOM access shouldn't be deprecated or denied (we may state though it's not recommended).
Cannot locate the passage in question in the current HTML5.1 spec. Suspect this bug has been superceded by time. If this is incorrect, please file an issue against the 5.1 spec: http://github.com/w3c/html/issues/