Re: feedback requested on WAI CG Consensus Resolutions on Text alternatives in HTML 5 document

On Aug 19, 2009, at 23:43, Jan Richards wrote:

>>>>> When used, a "missing" signal (however this might be encoded -  
>>>>> as long as it is not in the @alt string) would communicate to  
>>>>> the user agent that the @alt value should not be trusted.
>>>> What concretely is this envisioned to mean for user agent behavior?
>>> Since the semantic meaning would be "don't trust the @alt string"  
>>> I would think the behaviour would be the same as it is now when  
>>> @alt is fully missing.
>> This seems more problematic than a pure validator pragma. Why would  
>> a generator generate an alt string and a signal for UAs to  
>> completely ignore it? More importantly, having an "user agents  
>> should ignore alt completely" marker would pose a backward and  
>> forward compatibility problem: If the markup generator wants UAs to  
>> ignore the alt completely, generating some alt and an "ignore"  
>> marker would make old UAs not ignore the alt. If early adopters use  
>> the "ignore" marker while testing in current UAs, future UAs would  
>> lose potentially useful text alternatives.
> I can see the backwards compatibility problem, but not the forwards  
> compatibility problem. Why would testers add the "ignore" marker  
> while testing in current UAs?

Given enough rope, Web authors do the wildest things for the craziest  
reasons. However, here's a plausible non-crazy failure scenario:

Author A creates a document using an authoring tool and fails to make  
the document accessible. The authoring tool inserts the "ignore"  
marker. Later, author B in author A's organization addresses the  
problem that the document is inaccessible. Author B adds sensible alt  
text using a text editor. While author B has read introductions to Web  
accessibility to know about alt, author B isn't aware of the finer  
points of HTML syntax and fails to remove the "ignore" marker. Now the  
document is still inaccessible in UAs that honor the "ignore" marker.  
Author B could even test the document in older UAs without noticing  
the problem.

>> If the user of the tool inserts an image (outside <figure>, etc.  
>> special constructs) and doesn't provide a text alternative or  
>> rejects (which must be an option under ATAG 2 B.2.4.2.a) a tool- 
>> suggested text alternative but doesn't flag the image as omissible  
>> from AT presentation, as far as I can tell, generating  
>> role=presentation would be wrong by analogy with ATAG 2 B.2.4.4 and  
>> generating any alt would be wrong per ATAG 2 B.2.4.3 when the tool  
>> has no relevant sources that the UA wouldn't have.
> If the author rejects the suggested @alt value, ignores the prompts/ 
> checks for their own @alt, and turns off the "missing" mechanism  
> then....yes, invalidity will result, but at some point, isn't that  
> the author's choice?

I meant in the absence of a "missing" mechanism. If a "missing"  
mechanism is introduced, it makes no sense for authoring tools to have  
UI for micromanaging the "missing" mechanism.

Henri Sivonen

Received on Thursday, 20 August 2009 06:19:14 UTC