Re: Proposed rewording of ISSUE-204 text to address outstanding technical issues

Hi John,

Thanks again for participating in the discussion.

On Sep 10, 2012, at 7:30 PM, John Foliot <john@foliot.ca> wrote:

> Maciej Stachowiak wrote:
>> 
>> To provide further context: how is this context being referenced? Is it
>> via a <td headers=""> attribute? Via <img usemap="">? Or something
>> else? If it just appears in the document without being referenced at
>> all, the proposal is pretty clear that it would not be presented in any
>> modality.
>> 
>> I think that might affect the answers to your questions.
> 
> Sure.  Here is some real, live code:

[snip]

Thanks! Here is a saved Live DOM Viewer version of that markup, with the headers marked hidden and borders added for clarity:
http://software.hixie.ch/utilities/js/live-dom-viewer/?saved=1755

I'm not sure it entirely makes sense to hide the headers in this case, as they are not manifest from context, but perhaps it is still sensible to talk about the behavior when they are referenced. I hope Ted or James can chime in here.


> Moving on here, Ted and James have taken pains to not include ARIA in the
> new text, simply suggesting that something that is hidden could still take
> semantic 'fullness'.  They have brought forth 2 examples: <td headers="">,
> <img usemap="">), but leave the door open for others - but what others would
> there be?  What other elements use a mechanism that references IDs without
> invoking ARIA?  If all we have is headers= and usemap=, then why not just
> say so?

The original text cited aria-describedby as an example. The new text uses some non-ARIA examples, to avoid the PFWG's jurisdictional concerns. I think it may be valuable to do a survey of indirect reference mechanisms across HTML and potentially include them in this guidance. I believe that there are a number of mechanisms like this in HTML, but generating a complete and correct list would take considerable time. I hope you will agree with me that this should not be a showstopper for improving the text at all, as the current proposal replaces non-exhaustive example with non-exhaustive example.

> 
> And what *if* an author did something as above, and used a @hidden header
> with a hyperlink in it? Would Safari/VoiceOver expose the link in the modal
> dialog you explained earlier? Is this what we should expect consistently in
> User Agents that would support this technique? 
> 
> As I have previously noted, there is a tension between under-specification
> and over-specification, and left too under-specified here delivers us to a
> situation similar to @longdesc: a good idea that was initially
> under-specified in its implementation, and thus pilloried and derided down
> the road.  
> 
> I don't want to see that happen to something that is supposed to benefit
> accessibility ever again!

I agree that functionality should not be under-specified. This is in tension with the fact that assistive technologies are generally given fairly broad latitude in the details of their UI choices and how they adapt content for particular user needs. Can you think of language that sets the right direction without overconstraining future implementor choices?

BTW I don't know if I've fully conveyed to you how the VoiceOver experience works for sighted users. From your text above I am not sure you are fully getting it. Did you watch the videos I pointed you to? I would be glad to do a demo in person for interested parties.

Cheers,
Maciej

Received on Tuesday, 11 September 2012 06:33:44 UTC