Re: Revert Request

Matthew Turvey, Sun, 29 Jan 2012 18:15:35 +0000:
> On 28 January 2012 04:57, John Foliot <john@foliot.ca> wrote:
>> Matthew Turvey wrote:

> Since this change to the HTML5 spec seems to be just bringing it into
> line with what is already specified in ARIA,

Whether it should be *conforming* to point to sections that are hidden 
with the @hidden attribute, has nothing - in itself - to do with ARIA. 
That it *works* - per ARIA - to point to hidden sections - regardless 
of how they are hidden - is the only link to ARIA here, as far as I can 
see.

> It seems there are parts of WAI-ARIA 1.0 where no one can agree on
> what it actually specifies, or whether what it does apparently specify
> is actually possible or even desirable.

Which are the sections of ARIA 1.0 that you have in mind?

> This simple solution meets all the requirements:
> 
> <a href=foo><img src=pic alt="*the purpose of the link*"></a>

Most or all AT present the above as a link and not as an image.
 
> If users need to be able to determine programmatically that the link
> is a long description of the image, or authors want to put two links
> on one image:
> 
> <a href=foo rel=longdesc><img src=pic alt="*a programmatically
> determinable long description link*"></a>

No UA/AT support rel=longdesc. Hence: same problems as described above.

> <img src=pic alt="*text alternative*" usemap=#map>
> <map name=map>
> <area shape=rect coords=0,0,100,50 href=foo rel=longdesc alt="*a
> programmatically determinable long description link*">
> <area shape=rect coords=0,50,100,100 href=bar alt="*on an image that
> is already a link*">
> </map>

In Safari+VoiceOver, the @alt text of the <img> in this image map, is 
not presented to the user at all. Only the alt text of the <area> 
elements is presented. Last I checked, there were similar problems in 
other AT as well.

> This universal design approach works for everyone, right now,

Given what I described above, I fail to see how img@longdesc can 
currently be replicated with your above mentioned techniques.

> and
> doesn’t require changes to accessibility APIs, software upgrades,
> browser add-ons, user training, author training, or employing the
> services of an accessibility specialist. Unlike longdesc (and ARIA)
> this technique currently works in all screen readers, including
> VoiceOver, Orca and NVDA, as well as all other AT, including screen
> magnifiers, and all browsers.

Doesn't match the terrain - see my note above Safari+VoiceOver. There 
seems to be a general problem related to handling something as both a 
link and an image. And likewise, when something is considered an image, 
then the element is seen as being empty, even if it - such as for 
<object data=image> Text. </object> - is non-empty. The only solution 
that has solved that problem, so far, is @longdesc. 

In a summary:

If UAs/ATs supported rel=longdesc ...
If UAs/ATs would present the alt text of <img usemap=m alt=imageText >
If UAs/ATs would present fallback content of <object type=image/png> …

... then we could have used those methods for presenting a link to a 
long description.

None of the above has any thing to do with ARIA. 

> Since nearly everyone agrees we should include ARIA support in HTML5,
> it makes sense to include the ARIA functionality the authors of
> WAI-ARIA 1.0 clearly believe is essential for AT users, and improve
> existing accessibility APIs as required to support that functionality
> as Silvia previously suggested, and/or patch it in at the UA level as
> Jonas has previously demonstrated with his proof-of-concept code to
> make Firefox expose a link pointed to with aria-describedby.

It would be nice if @aria-describedby could point to a section 
*without* flattening the content, semantically. Then we could, for 
example, do a thing like this:

  <object aria-describedby=fallback data=image id=fallback >
    <i>Image description</i>
    <a href=link>Long description</a>
  </object>

But unfortunately, this does not work.
-- 
Leif Halvard Silli

Received on Sunday, 29 January 2012 22:31:19 UTC