- From: Al Gilman <Alfred.S.Gilman@IEEE.org>
- Date: Wed, 10 Sep 2008 10:33:06 -0400
- To: "wai-xtech@w3.org WAI-XTECH" <wai-xtech@w3.org>
[personal opinion disclaimer...] On 10 Sep 2008, at 9:08 AM, Al Gilman wrote: > > WCAG WG has requested help from PFWG in responding to the following > comment: > > <quote > cite="http://lists.w3.org/Archives/Public/public-comments- > wcag20/2008Aug/0013.html"> > > I propose aligning with the HTML5 draft by saying that an image be > marked as omissible from non-visual rendering if it illustrates what > the surrounding prose already says. PFWG has generally viewed "the surrounding prose" as fatally vague in this statement. In other words, the function "assistive technology uses the proximate prose as the text alternative" is not readily achievable. The pertinent prose needs to be isolated or the user takes a big hit. Listening to everything to find the pertinent stuff (unless there are readily machinable rules for the case where there is only a little and it's all about the picture) is an undue burden on the user, and isolating the germane stuff without help from the markup places an undue burden on the AT. We would like to improve the performance of adapted user experiences in this case, however, where an image is tightly related to prose that is near it. One way to make the association machine-recognizable would be to wrap one or more <span id=...> elements around the pertinent prose and cite this/these in @aria-describedby on the image. Even when such a long discourse in avaiable and associated, I personally doubt that "marked to ignore" is the highest and best markup of the image itself. My suspicion is that in the audio rendering it would be better to have some sort of acknowledgement of the position of the image in the reading of the information. This can be terse and perhaps as simple as "the image." That's best left to WCAG techniques and tutorials, or perhaps the "Author's guide to HTML5" but not the specification of the format. On the WCAG side, I don't believe that this case requires any change in success criteria. > I propose aligning with the HTML5 draft by saying that if the > generator of markup does not have a text alternative available, it > should use the natural-language expression describing the kind of > content (to the precision known to the generator) as the text > alternative and use a mechanism provided by the host format for > marking the text as not really being a text alternative but an > indication of what kind of non-text content is in question. In the case that a fully successful text alternative is not available, but there is recommendable repair text that the markup-producing tool can synthesize one should not say that "a text alternative is not available." Say "the best available text alternative is not (known / expected) to meet WCAG success criteria..." Knowing that the text alternative offered is a repair synthesized by a tool without benefit of human review is potentially of some value on the client side. But not that important to drive the whole decision process. In particular, although WCAG is transparent to the means the host language chooses to impart this information, PFWG should be recognized as a knowledgeable source for assessing the burden on AT represented by what hypertext markup means are used to denote this information. In particular, the insertion of curly braces within the text value of the @alt attribute has negatives that seem stronger than the associated positives, albeit non-trivial positives apply. How to address this: summary: not in WCAG proper (Success Criteria); maybe in WCAG techniques, maybe in HTML5 author guide. And TBD in HTML5 spec for "this is a repair" indication. There should be no barrier to HTML5 telling authors how to do the best available practice in cases where success is not achieved. If there is a transcoder/formatter tool that is generating markup without the interactive cooperation of an author, it may encounter cases where it can generate a "probably better than nothing, but probably not success by WCAG terms" text alternative. In such a case the HTML5 specification and author guide documents should make clear that the indicated encoding is not up to WCAG success criteria, but what to do when you can't achieve that. Whether the instance documents should contain this as metadata is unclear, pending a compelling proposal. My personal opinion is that this is better handled in EARL metadata, probably outside the HTML page. http://lists.w3.org/Archives/Public/wai-xtech/2008Aug/0190.html There is an alternative that WCAG Techniques be expanded to include this "close, but no cigar" case. WCAG could take this on, but if WCAG declines, it should be with the agreement that the HTML5 author guide at the very least may include this case. [and WCAG WG can leave to HTML WG, ATAG WG, UAWG and PFWG to work out if and how to indicate in the document that a repair has been machine-generated and is not trusted by the generator to meet WCAG requirements.] € 0.02 Al /self, no hat implied > </quote> > > discussion and status in WCAG: > http://trace.wisc.edu/bugzilla_wcag/issuereports/issue_ind.php?id=2615 > > I'm opening this thread on XTECH because WCAG works on the public > record, and > to have a cross-group conversation to explore issues and options. > > Formal feedback will be based on consensus among the PFWG > participants, but > we welcome input by discussion here. > > Al > > >
Received on Wednesday, 10 September 2008 14:33:51 UTC