RE: Drop longdesc, get aria-describedat?

Geoff Freed, Thu, 8 Mar 2012 01:21:45 +0000:

Hi Geoff,

> Apologies for jumping in late.  I haven't read all the messages in 
> this thread yet, but at the risk of repeating what someone may have 
> already pointed out, take note that the DIAGRAM project 
> ( has begun coordinating with the EPUB 
> working group on an epub:describedAt attribute for image descriptions:
> Before going down the CG path, or any other W3C path, I suggest you 
> take a look here first and then see how we can coordinate with the 
> work that DIAGRAM has already begun.

Thanks for the link. That page also points to Laura's research page on 
@longdesc. So it is pretty obvious that the describedAT attribute is 
thought of as a replacement for @longdesc. Which in turns means that - 
in the end - @longdesc's burial is being prepared. It is also stated 
that the problem with @longdesc is that it is only good for <img> [and 
<iframe>, but no one has seriously been suggesting @longdesc to be 
valid for <iframe>].

So, this only confirms that Janina's preferred solution - obsolete but 
conforming - would be the most *accurate* description where we are 
heading with @longdesc.

In a way, it is good that @describedAT is so directly linked to 
@longdesc, though. Because, the closer @aria-describedAT is to 
@longdesc, the better reason there is - in my opinion - to make 
@longdesc 'obsolete but conforming': Could it not make sense, for 
instance, to say that @longdesc is conforming, but only whenever it is 
duplicated with a @aria-describedAT?

As for the conformance issue - and the trouble that @aria-describedAT 
currently isn't whether specced or implemented, consider that, per 
HTML5, it is currently conforming to do this:

 <div><style scope>div{color:red}</style></div>

One could argue that it should not be conforming, yet, because <style 
scope> has zero implementation, so far. In fact, if you look in 
public-html - in this moment, then you'll see that it isn't even clear 
how it will work ...  the CSS cascade is unclear:

So we do not necessarily need so *much* before we can put in HTML5 ... 
At the same time: It would be futile, in an directly accessibility 
related issue, to recommend people to use an attribute with zero 
implementation. And hence, yet another argument in favor of saying that 
the validity of @longdesc should depend on duplication with the 
@aria-describedAT attribute.
leif halvard silli

Received on Thursday, 8 March 2012 02:35:19 UTC