Re: [Bug 12587] AT shouldn't see img@alt="" and img@role=presentation as 100% synonyms

Steve Faulkner, Tue, 3 May 2011 12:28:55 +0100:

> we will certainly consider this in light of your bug details, but I 
> would suggest that the first place to raise this as a bug is the 
> html5 spec as that is where the mapping between alt and 
> role=presentation is normatively defined.

>> http://www.w3.org/Bugs/Public/show_bug.cgi?id=12587

I'm in the process on filing the bug you ask for. Some comments though:

You should separate @role from validity. What if an IMG has <img 
role=img alt=""> ? What if it has <img role=presentation alt="Lorem 
Ipsum"> ? What if <img aria-labelledby=something alt="" > ? The 'HTML 
to Platform Accessibility APIs implementation guide' is about *role*, 
and not about validity. The Decisions we have had are also about 
validity and not about role.

Bug 12587 relates to you and Richard's view on whether 
role=presentation makes the IMG valid without presence of @alt 
attribute. You yourself also fight the Decision which says that a 
non-empty @title makes the IMG valid even if it lacks @alt. Doesn't the 
Decision on @title show why it is bad when one A11Y attribute inflicts 
on the validity of another?

The @title decision creates doubt: what if the IMG has a non-empty 
@title in combination with empty @alt? VoiceOver then considers that 
the image is non-presentational and reads the @title to the user. (I 
don't know about other AT.)

As I said in the poll: before deciding whether @alt can be empty of 
non-empty, AT as well as author should calculate the role of the IMG. 
As expressed my my 'validity map' [1], the valid value of @alt should 
depend on the outcome of that calculation, rather than the opposite way 
- that the role of the img depends on @alt. Only if *nothing else* 
affects the role, should the @alt affect the role of the IMG.

Thus: if there is a @title, then the IMG should - by default [that is: 
unless there is a role=presentation] - be considered 
non-presentational. It follows (as I see it), that - for validity - 
there should also be a non-empty @alt. This is also compatible with 
your view on @title.

For AT, the most important thing to decide, is whether the IMG is 
non-presentational - or not. And not whether it is valid - or not. AT 
should definitely use @title. But we must separate 'should use @title' 
from 'valid as long as there is @title'.

[1] http://lists.w3.org/Archives/Public/public-html/2011Mar/0750
-- 
Leif H Silli

Received on Tuesday, 3 May 2011 13:11:21 UTC