Re: 48-Hour Consensus Call: InstateLongdesc CP Update

On Mon, 24 Sep 2012 12:32:26 +0200, Sam Ruby <rubys@intertwingly.net>  
wrote:

> On 09/24/2012 04:33 AM, Charles McCathie Nevile wrote:
>> On Sun, 23 Sep 2012 22:44:34 +0200, Sam Ruby <rubys@intertwingly.net>
>> wrote:
>>
>>> On 09/23/2012 04:16 PM, Geoff Freed wrote:
>>>>
>>>> Just for the record, I think that @longdesc *should* be improved.  If
>>>> the name remains the same, fine.  If it changes or is moved to ARIA,
>>>> fine.  I just don't want it to go away before that new Thing is
>>>> available.
>>>
>>> I must say that that's an eminently reasonable position to take.
>>
>> And is exactly my opinion...
>>
>>> Geoff: I gather that James hasn't convinced you that iframe is a
>>> superior solution to address the challenges you face.
>>
>> I'm not convinced for reasons I explained yesterday.
>>
>> Which is where the disagreement has arisen. The one thing I have seen  
>> that convinces me that it could solve the problems is the proposal for
>> aria-describedat.
>
> Can you explain why aria-describedat might be, in your opinion, an  
> adequate replacement?

Because it does the same things (i.e. meets the requirements I see). It  
changes the name,

> Is there any reason that such an specification could not be pursued in  
> parallel and meet CR exit criteria by 2014Q2?

Well, it needs to be implemented. Like longdesc, that ain't hard.

I believe a longdesc spec would meet CR exit criteria today, even while  
stating that any user agent that didn't directly expose the content to the  
user is not conformant.

>  Could it, like we are proposing here, be pursued as a separate  
> specification[1] and therefore not impact ARIA 1.1 or ARIA 2.0, but just  
> like we are proposing for HTML be integrated into ARIA once it reaches  
> maturity and consensus?

That depends on how PF manages the ARIA specs, but sure, it is feasible in  
principle.

> Care to comment on the position that David Bolter expressed[2], but I  
> have heard from others, namely that "part of the beauty of ARIA is that  
> it is purely annotative semantics that can be added to describe existing  
> UI without interfering with that UI"?

Sure. I think this is one of the major mistakes of ARIA, and I strongly  
recommend implementors ignore that particular aspect.

It provides for a disjointed implementation where user agents who *use*  
the annotations as designed might essentially turn a page into something  
completely different from the one rendered by user agents that don't let  
the ARIA interfere with their UI.

The current draft has an each-way bet, saying user agents *may* implement  
ARIA other than through accessibility APIs.
[...]
>>> Again I ask: is there any chance that we can get a consensus spec out  
>>> of this: one that doesn't attempt to portray publishing software that
>>> produces markup including longdesc as non-conforming; nor does it
>>> attempt to portray user agent software that doesn't natively implement
>>> longdesc as non-conforming?
>>>
>>> Geoff: if such an extension specification were written, could you live
>>> with that for now?
>>
>> In case we can get consensus on such an approach, I'll draft an  
>> extension spec.
>
> Excellent!

I have yet to add the use cases and requirements in, but here's a  
morning's work...

>> I believe it is a compromise. But it is one I could live with.

cheers

-- 
Charles McCathie Nevile - Consultant (web standards) CTO Office, Yandex
       chaals@yandex-team.ru         Find more at http://yandex.com