- From: Al Gilman <asgilman@iamdigex.net>
- Date: Mon, 03 Dec 2001 14:48:46 -0500
- To: <w3c-wai-ig@w3.org>
- Cc: RandR@SEC.GOV, Mike Scott <mscott@msfw.com>
At 10:46 AM 2001-12-03 , Phill Jenkins wrote: >Do you have a reference in the CSS spec that suggests that screen readers >read out loud content styled with Visibility: Hidden? AG:: Certainly not by default. >I couldn't find one >nor do I think screen readers should read anything marked hidden. AG:: As a 'try harder' option, the user agent should allow drilling down to expose the content marked hidden. This is a position that PF successfully defended in dialog with the SYMM working group as regards the semantics of the 'override' attribute on customTest variables. <http://www.w3.org/TR/smil20/smil-content.html#edef-customTest>http://www.w 3.org/TR/smil20/smil-content.html#edef-customTest I realize that the rationale is not public, but the result is. Compare the end result with the previous draft which allowed the author to impose a flat, final override-NOT condition. The reason is to maintain the user's trust. There must be no secrets which the User Agent and the Web Site share and the user is barred from knowing. I know this may sound silly to view system control codes as conspiratorial, but that is the way it comes across to the user. And authors _will blunder_ in the application of 'hide' codes. The Web has a long way to go to regain the user's trust in this respect. Further, people with disabilities have generally a more tenuous hold on control of the interactive dialog, so it will take more slavish devotion to "the user is always right" to keep them reassured and trusting than for the general population of users. The specification coverage is not that clear in CSS, it is spelled out more forcefully in UAAG Guideline 2. The content in question is intended as conditional content, anyway, not always-hidden content. Note that WCAG 1 requires the page to work with style sheets turned off, in which case the hiding will not take effect. So content hidden by style is not safely always hidden, and there is no reason to assume that screen readers will violate the 'hide' indication unless there is an explicit key in terms of media that indicates the hiding does not apply in audio. Now, the line we are pursuing in both the content guidelines working group and in Protocols and Formats seeking reforms in HTML, the desired plan is a) have a device-independent model of how all content fits with the other content. This means that you know the relationship of the 'expanded section intro' text to the rest of the content in the document. It is recognized as an introduction to the section which is likely to be valuable when the content is presented in a serial stream with only the user's memory for context, and likely to be unnecessary when the content is presented in the context of a screenful of other stuff which provides implicit orientation. But, following Tim Berners-Lee's injunction to "say what you mean, not what to do with it" this is identified first as a narrative introduction to the section, and later bound to show/hide semantics in hints that reflect the author's guesses as to what is appropriate in different delivery contexts [see Device Independence principles draft for this concept][See also the XML Accessibility Guidelines]. The author's guesses as to when it is appropriate to show or hide this narrative introduction are just guesses. At a certain point, the person using a screen magnifier will want these section introductions displayed in a visual presentation, and not just when the user is using a screen reader. The context cues will be off screen for a large enough magnification. Also, a context [you are here] display accessed through the context menu will be a mode of accessing this content that will make sense at all magnifications, and make the rationale for the structure more easy for all to follow. So show/hide indications per output media or delivery context indicated by the author have the status of hints. Good hints, because the author understands the content. But overridable hints, nonetheless, because the author cannot be expected to have understood the full range of delivery contexts in which the content will be accessed. Al
Received on Monday, 3 December 2001 14:39:57 UTC