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

On Aug 17, 2009, at 00:11, Leif Halvard Silli wrote:

> I agree that it should not insert role="presentation" by default.  
> However, since we both agree with Consensus in that <img> without  
> @role defaults to role="img", it could insert role="img".

What problem would this solve?

> Tools do not need to ask "Do you want to insert an <img>?" They  
> could offer choice between IMG@role=presentation and normal IMG.  
> Tools should not bug users about lack of alternative text unless the  
> <img> has  a non-presentational role ...

What kind of markup and UI do you envision for the case where in a  
future HTML5-compliant version of Dreamweaver, the user creates a new  
document (File: New), drags an image file to the document from the  
Finder and saves the document?

> We should treat lack of @alt and empty alt="" as semantically  
> identical.

That's not how existing client software behaves. Previously, it has  
been stated on the list that it takes a long time to upgrade the  

> The Consensus Documents goes in that direction when it states that  
> it doesn't mater if an <IMG> with role="presentation" has an empty  
> alt="" or no alt at all. But it goes slightly in the opposite  
> direction when it recommends that validators should say that an  
> <IMG> with an empty alt="" but not @role should automatically get a  
> role="presentation".

My biggest concern with the proposed normative warning is that  
role=presentation wouldn't be the path of least resistance for  
dismissing the warning. Putting a space in the value of alt would be.

On Aug 17, 2009, at 02:25, Maciej Stachowiak wrote:

> If you're saying that authoring tools should produce missing (not  
> empty) alt, without any of the alternatives suggested by HTML5 such  
> as @title or <legend>, and that this should be conforming, then I  
> believe you disagree with the current state of the HTML5 spec.

Yes. I pointed out in my previous email that I disagree with the  
current state of the HTML 5 spec. (I believe Hixie knew this, but for  
all-round productivity reasons I didn't make a big deal about it on  
this mailing list while the alt discussion was otherwise dormant.)

The reason I disagree with it is that I haven't seen a credible  
expectation of how a Dreamweaver-like product should implement the  
requirements of HTML5 as drafted without failing to conform to ATAG 2  
as drafted (or vice versa).

Anyway, I think discussing what should be conforming before coming to  
consensus on desirable authoring tool behavior will rathole this  
thread. Therefore, instead of discussing my conclusions, I'd like to  
state my premises and invite anyone who disagrees with any of my  
premises to come forward. If it turns out that one of my premises is  
wrong, my conclusion is most likely wrong.

Here are my premises:

  * "Authoring tools and markup generators must generate conforming  
documents." ("Authoring tools are exempt from the strict requirements  
of     using elements only for their specified purpose, but only to  
the extent that authoring tools are not yet able to determine author  
intent." "In terms of conformance checking, an editor is therefore  
required to output documents that conform to the same extent that a  
conformance checker will verify.") (Quoted from HTML 5.)

  * "After the end of an authoring session, the authoring tool does  
not attempt to repair alternative content for non-text content using  
text content that is equally available to user agents (e.g., the  
filename is not used)." (Quoted from ATAG 2)

  * Autogenerated alt="image", alt="" and alt=" " violate the ATAG 2  
language quoted in the previous point.

  * Autogenerated alt="photo" might be spun not to violate it but  
practically isn't materially different from alt="image".

  * Autogenerated role=presentation doesn't violate the ATAG 2 point  
literally but does in spirit.

  * An HTML authoring tool should conform to both HTML 5 and ATAG 2.

  * In a GUI editor, the user should be able to insert and delete  
images and section/figure headers/captions/legends independently of  
each other, because gluing them together would violate long-standing  
GUI behavior expectations.

  * Most authors don't respond to prompts in a meaningful way.  
(Contrast with ATAG 2 B.1.3 applicability notes.)

  * Dreamweaver and BlueGriffon-type products, Microsoft Word and products (when exporting HTML) and Flickr and  
Brightkite-type services are legitimate classes of services and  
products that are within scope for HTML 5 and ATAG 2.

Does anyone disagree with any of these points?

On Aug 17, 2009, at 02:45, Maciej Stachowiak wrote:

> (To be specific, I think the best option in this case that would be  
> conforming to the current draft would be to put a title attribute on  
> the image, giving the best information the authoring tool has  
> available, even if it is a low-value description like "Photo 1 of  
> 15".)

The "1 of 15" part should be aria-posinset=1 aria-setsize=15. For the  
remaining "Photo" part, see above.

Henri Sivonen

Received on Monday, 17 August 2009 07:12:18 UTC