Re: WAI-CG Consensus Recommendations on Alternative Text in HTML 5

Hi Henri,

Comments in-line...

Henri Sivonen wrote:
> (Added public-html to CC per HTML WG telecon minutes.)
> 
> On Jul 9, 2009, at 16:28, Jan Richards wrote:
> 
>> Henri Sivonen wrote:
>>> On Jun 11, 2009, at 04:30, Janina Sajka wrote:
>>>> This document is also available at the following URI;
>>>> http://www.w3.org/2009/06/Text-Alternatives-in-HTML5
>>>>
>>>> We look forward to working with you to make HTML 5 the best
>>>> accessibility solution yet.
>>> Thank you for documenting the WAI CG consensus.
>>> I have three questions related to software generating <img> elements:
>>> 1) The user creates a new HTML document in an HTML5-compliant future 
>>> version of a Dreamweaver-like product and saves it with a name. The 
>>> user then drags an image file from the Finder onto the document 
>>> window and the image is inserted into the document. The user presses 
>>> command-S and then command-Q. What attributes should the generated 
>>> <img> element have?
>>
>> JR: The consensus document does not attempt to dictate syntax for this 
>> case, but the following options agree with the consensus and Public WD 
>> of ATAG 2.0 (http://www.w3.org/TR/2009/WD-ATAG20-20090521/):
>>
>> a. The tool leaves @alt out entirely, so that the code produced is 
>> invalid.
> 
> That doesn't make sense. Invalid means that authors or tools acting on 
> their behalf must not generate what is defined to be invalid.
> 
> It doesn't make sense to have a situation where ATAG 2.0 says "do X" and 
> HTML 5 says "you must not do X". If there's consensus that it's good for 
> ATAG 2.0 to say "do X", then we should make sure HTML 5 allows doing 
> X--i.e. doesn't make it invalid.
> 
> (It's unclear to me where ATAG 2.0 actually says that alt can/should be 
> omitted in this case. Specifically, B.2.4.3 is too abstract for me to 
> apply concretely to HTML5 with confidence that my application is what 
> ATAG 2.0 meant.)

JR: ATAG 2.0 does NOT say "include @alt" and it does NOT say "don't 
include @alt". What it does say (in B.2.4.3) is that, if the human 
author is no longer available (e.g., they closed Dreamweaver, their 
phone auto-uploaded an image without their intervention), then the 
authoring tool shouldn't automatically fill in any alternative content 
(including @alt) using any text content that the user agent could just 
as easily use BECAUSE doing so misleads the user agent and end user into 
thinking that the @alt is more reliable than it is.


>> b. The tool makes use of some kind of 'autogenerated' or 'missing' 
>> attributes as per this paragraph in the consensus:
>> "In order to address both the validity and human generation concerns, 
>> we do not oppose the creation of 'autogenerated' and 'missing' 
>> attributes where either one of these could be used to make an image 
>> that does not have any human-generated text alternatives valid. (Note: 
>> It is important that this marker is not included in the alternative 
>> text string itself.)"
> 
> Is there a WAI CG consensus on what the marker should concretely be in 
> the case I outlined? Do you mean an attribute named "autogenerated" or 
> "missing"?

JR: There was no concrete proposal on this because the task force wanted 
to avoid being seen to dictate syntax to the HTML5 group wherever possible.


> ATAG 2.0 B.2.4 doesn't help me understand what ATAG 2.0 wants HTML 
> generators to do when there aren't "relevant sources" that B.2.4.3 
> doesn't exclude and the user doesn't cooperate affirmatively per 
> B.2.4.2(a).

JR: Remember: ATAG 2.0 applies to formats other than HTML5. For this 
reason there is no validity requirement in ATAG 2.0 and therefore, in 
the general case, alternative content attributes could just be left out. 
  In the HTML5 case (where validity is required), you are looking for a 
construct that means "this @alt is included for validity but can't be 
trusted as alternative content" and that is what you could get with 
attributes such as "autogenerated" or "missing". Note: Content doing so 
might be valid to HTML5 but would not meet WCAG2.0.


>>> 2) The user takes a photo with his/her cameraphone and uploads it to 
>>> a service such as Brightkite. The user has the option to add an 
>>> SMS-like short caption that will be rendered in addition to the photo 
>>> in visual browsing settings but is not forced to add a caption. What 
>>> attributes should the generated <img> element have when the photo 
>>> appears immediately on the Web? (Brightkite now says 
>>> alt="Photo-feed". Reference: 
>>> http://brightkite.com/objects/80d0f32c11e311deab31003048c10834)
>>
>> JR: If the photo appears "immediately on the Web" (without any 
>> "authoring session") or if the author ignores prompts for the short 
>> caption during their authoring session, then the tool may add an @alt 
>> value as long as it is a value that the user agent wouldn't know. So 
>> because the fact that this is a "photo" is information the user agent 
>> can't easily detect,
> 
> I would guess that detecting that a bitmap is a photo with a pretty good 
> confidence would of the easier kind of signal processing tasks, so I'm 
> suspecting it's a case of "don't" rather than "can't", but I haven't 
> programmed it to find out easiness myself, so OK.
> 
>> alt="Photo-feed" is ok (though of course very suboptimal from a user 
>> perspective)
> 
> OK.
> 
>>> 3) Point #5 under "Use Case 2 (author using a photo sharing site)" 
>>> instructs photo sharing sites to generating alt="Photo 1 of 50 of 
>>> album Paris 2009". Why is this approach recommended instead of 
>>> generating aria-posinset="1" aria-setsize="50" and an alt attribute 
>>> that doesn't repeat the set position data, e.g. alt="Paris 2009"?
>>
>> JR: To clarify, this isn't an "instruction". This is an example of the 
>> kind of information that "allowed" because it is more readily 
>> available to the authoring tool than to a user agent (hopefully with 
>> an accompanying 'autogenerated' attribute).
> 
> OK. If a W3C publication were to give this example, would using 
> aria-posinset="1" aria-setsize="50" set a better example than putting 
> the "n of m" data into the alt attribute?

JR: Sure. Another example of where the authoring tool might know 
something the user agent wouldn't would be a person's profile picture on 
a social networking site.

Cheers,
Jan

-- 
Jan Richards, M.Sc.
User Interface Design Lead
Adaptive Technology Resource Centre (ATRC)
Faculty of Information
University of Toronto

   Email: jan.richards@utoronto.ca
   Web:   http://jan.atrc.utoronto.ca
   Phone: 416-946-7060
   Fax:   416-971-2896

Received on Friday, 10 July 2009 14:01:51 UTC