Re: aria-describedat

Silvia Pfeiffer, Fri, 23 Mar 2012 08:25:09 +1100:
> On Fri, Mar 23, 2012 at 2:54 AM, Charles McCathieNevile:
>> On Thu, 22 Mar 2012 15:54:07 +0100, Leif Halvard Silli:
>>> Benjamin Hawkes-Lewis, Thu, 22 Mar 2012 13:28:11 +0000:
>>>> On Thu, Mar 22, 2012 at 9:28 AM, Charles McCathieNevile:

>>>> It might (?) make sense to restrict the HTML or ARIA semantics that
>>>> conformingly act as description fragments so as to exclude headings.
>>> 
>>> I think we 'just' need to say what authors *MUST* do: point to an
>>> existing fragment. And then, if they point to an non-existing fragment,
>>> then - in fact - per the way browsers handle it, the entire page would
>>> be in scope, no?
>> 
>> Sure. But some authors, whatever we say, will point to the h3 element and
>> not realise that it doesn't include the following stuff that is the
>> description they worked so hard on.

I always like Benjamin's proposal much better the second time I read 
them .. :-D So, hm, it might not be so bad to exclude headings the way 
he suggested.

>> I am prepared to live with a certain failure rate in order to make an
>> improvement, but it is worth trying to maximise the benefit and so worth
>> thinking about how things will go wrong when people are trying to do the
>> right thing.

Indeed. But it seems Silvia's proposal about a simple way for authors 
to test it, is the best way to solve that:

> If the browsers provide a way to visually display the description -
> e.g. in an overlay - then the authors can check the result of their
> work for themselves.
> 
> I would also suggest that a link to an element would just display the
> page fragment rooted at that element. If an author wants to include
> more than that, then they would put a div or p or so around the set of
> elements that they want to link to and reference that. It won't break
> HTML or fragment URIs. It's just a semantic of this attribute.

Just one thing to you all: We need to think at @aria-describedAT and 
@longdesc in tandem. Or else I think that one of them will fail. Which 
makes me think about another thing:

Rich, 

may I propose that you somewhere in your draft define that that AT can 
use native attributes if aria-describedAT is not present? 
-- 
Leif Halvard Silli

Received on Thursday, 22 March 2012 22:25:54 UTC