Re: From the HTML-WG about aria-hidden

Jason White, Fri, 17 Feb 2012 16:03:33 +1100:
> Leif Halvard Silli wrote:
>  
>> Yes, if 'specified properly' means 'specified in a way that vendors 
>> will implement'.
> 
> By "specified properly" I mean, specified so that it solves all of the
> problems which it is intended to address, and the requirements on 
> authors and user-agents are sufficiently well defined that it can
> be implemented consistently.

I hope that the longdesc CP will convince vendors. And also, that the 
vendors make it simple for authors to test @longdesc links, because 
easy testing seems necessary in order to get authors to use it - and 
use it correctly. Unfortunately, I don't think the CP *requires* that 
longdesc URLs can be accessed without AT. Another thing that bothers 
me, is that my bug reports for @longdesc support in text browsers, have 
so far not lead to results.

>> First, I will say that ARIA can also be described as something that we 
>> unfortunately must live with.
> 
> Actually, no, it can be progressively deprecated by moving what it expresses
> into HTML and other markup languages. The content model of IMG, by contrast,
> can't be changed without breaking compatibility with decades of software and
> HTML documents.

The ARIA spec includes only one kind of deprecation, I think. Namely 
that kind of 'deprecation' that happens when ARIA's "host language 
provides a semantic element for that type of object": In that 
situation, then ARIA says that the correct thing is to use the host 
feature: <http://www.w3.org/TR/wai-aria/introduction#co-evolution>

So I agree, that for the purpose of offering a long description, then 
'an img element with the longdesc attribute' would be the correct 
feature to use - provided HTML5 includes it. But one *could* also say 
that the native host feature to use, is <object data=image>.

The section of ARIA that I linked to above, offers this example: "For 
example, it's better to use an h1 element in HTML than to use the 
heading role on a div element." Which means that ARIA considers it the 
duty of the author to choose of the host language element that has the 
best native accessibility. [Though note ARIA says that even for native 
objects, "WAI-ARIA can improve the accessibility of these objects" - 
see same link as above. So much for 'deprecation' ...]

But one could say that, for those cases when the authors is unable or 
unwilling to replace <img> with a better element - <object>, then 
@aria-describedby [eventually plus <a hidden href>] or the future 
@aria-longdesc, is the way to go.

All this said: The historic problems and perhaps resistance towards 
using <object>, speaks in favor of a native @longdesc, I agree with 
that.

>> Second: W.r.t. alternative content as child element, then I have news 
>> for you, in that regard: Support for <object> is improving. And I have 
>> recently written a change proposal for ISSUE-158 which will change the 
>> content model of <object> to be more similar to what it was in HTML4. 
   ... snip ...

> I've read the proposal and I think it's entirely reasonable. I was a strong
> supporter of OBJECT when it was introduced into what became HTML 4, and I
> still think it's a better solution thank kludges such as @longdesc.

Thanks! Very glad to hear that.

> One possibility would be to fix the OBJECT content model as proposed, then
> advise authors and authoring tool developers that if they need anything more
> than @alt, OBJECT must be used. My suspicion, though, is that the desire for
> backward-compatibility would kill that proposal in practice, at least in the
> short term, which brings us back to properly specifying @longdesc, 
> preferably in addition to the OBJECT proposal rather than in place of it.

I agree with you with regard to the short term. But until the vendors 
step up, it seems not exactly unlikely, that the short term effect 
would only be that @longdesc becomes conforming - but without improved 
support in UAs in the nearest time frame. It is quite possible that 
legacy UAs with bad support for <object>, will be replaced with new UAs 
with good <object> support faster than new UAs with @longdesc support 
can be introduced.
-- 
Leif Halvard Silli

Received on Friday, 17 February 2012 16:25:58 UTC