Re: From the HTML-WG about aria-hidden

Janina Sajka, Fri, 17 Feb 2012 13:35:54 -0500:
> Leif Halvard Silli writes:
>> I gather that you have faith in ARIA as such, but not in this detail. 
>> It would be interesting to know why.
> 
> The ARIA TF is still considering LC comments, especially on the UAIG.
> The Status Sections in our documents clearly say not to rely on them.
> Yet, the docs are being cited as though the guidance there is settled.
> That's wrong and can easily be disruptive.

ARIA 1.0 present itself as a Candidate Recommendation. I could not find 
support for the claim that CRs 'clearly say not to rely on them': 
http://www.w3.org/2004/02/Process-20040205/tr.html#RecsCR

Anyway: I see that my question was a bit unclear: What I wanted to ask 
was about that detail in ARIA which says that elements reference by 
aria-describedby SHOULD be presented with *native semantics*: It 
sounded as if you, for elements that have been hidden with the @hidden 
attribute did not believe that AT and UAs would be able to present the 
native semantics. 

But I guess it was also a pretty pointless question: During the working 
group's work with ISSUE-204, the evidence - or lack thereof - will be 
key.

> The proper course, imho, would be to reach out the the ARIA TF (or PF
> itself) and ask.

The fact hat aria-describedby can point to sections that are hidden - 
in ARIA's sense of 'hidden' - is something I have reconciled myself 
with long ago: Having accepted how ARIA defines it, then the question 
of whether ARIA-describedby should be permitted to point to elements 
that are hidden with the use of the host language's @hidden attribute, 
seems like more of an HTML5-conformance issue than an ARIA issue.

PS: As far as I can tell, then ARIA 1.0 does not put *any* restriction 
on point to hidden elements. The only thing it puts a restriction on is 
the use of aria-hidden=true: Unless the element really *is* hidden - 
with some technique or another - for all medias and user groups, then 
there is only a MAY permission to use it. That is: If the element is 
visible, but should be removed from the A11Y tree, the there is only 
MAY permission to do that.

> In fact none of this discussion was even cross-posted
> to the HTML-A11Y TF though we've repeatedly asked the HTML Chairs to
> please do so when issues of particular interest arose on the main WG
> list. So, OK, I did spot the issue and advised our group. We will
> discuss the HTML approach, together with the various approaches we've
> been considering.

They chairs must speak for themselves. But I too was not aware of 
HTML5's permission to point to elements that have been hidden with the 
@hidden attribute. It was only when Laura informed the group, that this 
was known to me. The chairs did, however, in December, note that Jonas 
had this in his proposal - and asked him to justify it. From my own 
point of view, I supported Laura's recent revert request. But I 
expressed my support in terms of pointing to the chair's request for 
justification. Since the part of ARIA 1.0 where it speaks about 
aria-describedby, is pretty well known to me, I was difficult to be 
shocked by the thing itself, though.

Frankly: It does not seem fruitful to discuss ISSUE-204 in the light of 
what impact it could have on @longdesc: Such a thing only looks as if 
one is so married to an issue that one refuses to discuss other issues 
on their own merits.

> The point is that the whole of ARIA is at stake, not just how to get
> around longdesc with clever tricks.

3 questions - from least to most important:

A: Why do we view ISSUE-204 in light of its *possible* @longdesc impact?
B: Which problems other than the those related to @longdesc?
C: What is *the* problem?

> ARIA-Hidden is there for a number of reasons and use cases that are
> fairly complicated to think through clearly. And yet, that is what we
> have to do to insure that ARIA is a robust and reliable specification.
> The HTML-WG doesn't, of course, have that kind of stake in ARIA success.

While I have not specced ARIA, I have read parts of many times, I have 
offered feedback in a last call period and I have struggled with 
understanding it - including testing how AT interprets it. So, at least 
I am not *completely* unfamiliar with aria-hidden and its use cases.

I think it is a bit wrong to focus on the fact that elements that are 
hidden with aria-hidden are still presented to the user via 
aria-describedby. It seems more correct to focus on - and to say - that 
there is nothing that escapes @aria-describedby: Everything 
aria-describedby points to, becomes gold - ah - a description. Because, 
there are many, many ways to hide an element, apart from using 
aria-hidden. E.g. if an element has role=img, then its children are 
automatically hidden too. But aria-describedby and aria-labelledby 
don't care: All they point to get presented.

> So, my answer to you, the thing I keep asking myself in all of this--yet
> another Latin phrase--Cui bono.?

To whom does *what* benefit?
-- 
Leif H Silli

Received on Saturday, 18 February 2012 00:30:50 UTC