RE: Technique H45 procedure description incomplete

OK, but the phrase "Check that the img element has a longdesc element" to me means

"The img element SHOULD have a longdesc element".

If you intend it to mean

"If the img element has a longdesc element, then..."

then why not state this?

If it's true that the procedures are meant to be consistently described as though they apply to a single item then a) H37 should be rewritten, because it says "Check that *every* img element..." and b) in many cases there then needs to be a separate procedure for determining which items the technique applies to.  Personally I don't see any advantage in this - it would surely be less confusing to simply describe what a user or automated tool needs to do to determine whether the whole HTML page satisfies the technique.

In other words, the procedure for H45 would be something like:

1.  Check each IMG element to determine whether it can be adequately described with a short-text alternative (ALT text)
2.  If not, ensure that the IMG has a LONGDESC attribute
3.  Ensure that all LONGDESC attributes refer to an existing resource
4.  Check that such resources fully describe the content conveyed by the corresponding IMG element

That's more or less how any automated tool would have to function - though accepted step 1 is difficult to automate in the general case (a web developer may know, however, that all images on a particular site with a certain key string in the URL require a long description, and tweak the procedure appropriately).   And for WCAG 2.0 to be a useful set of guidelines that corporations can actually test large real-world sites (>100000 pages) against, it needs to be suitable for that sort of automated testing.  At this stage, it appears unfortunately that WCAG 2.0 is a step backwards from 1.0 as far as appropriateness for automated testing goes.

________________________________
From: Loretta Guarino Reid [lorettaguarino@google.com]
Sent: Tuesday, 2 December 2008 10:55 AM
To: Dylan Nicholson
Cc: public-comments-wcag20@w3.org
Subject: Re: Technique H45 procedure description incomplete

Hi, Dylan,

It feels like there is some confusion about the sufficient techniques. They are consistently written to apply to a single instance of content covered by the corresponding success criterion. It may turn out that the author always uses the same technique, but we have been careful not to make that assumption when describing the technique.

For instance, some images on an HTML page may need long descriptions and some may not. So some images may be using G94: Providing short text alternative for non-text content that serves the same purpose and presents the same information as the non-text content<http://www.w3.org/WAI/GL/WCAG20/WD-WCAG20-TECHS-20081105/G94>, using H36: Using alt attributes on images used as submit buttons<http://www.w3.org/WAI/GL/WCAG20/WD-WCAG20-TECHS-20081105/H36>  for the short text alternative. Other images on the same page may be using G95: Providing short text alternatives that provide a brief description of the non-text content<http://www.w3.org/WAI/GL/WCAG20/WD-WCAG20-TECHS-20081105/G95>, using H36: Using alt attributes on images used as submit buttons<http://www.w3.org/WAI/GL/WCAG20/WD-WCAG20-TECHS-20081105/H36>  for the short text alternative, and G92: Providing long description for non-text content that serves the same purpose and presents the same information<http://www.w3.org/WAI/GL/WCAG20/WD-WCAG20-TECHS-20081105/G92> , using H45: Using longdesc<http://www.w3.org/WAI/GL/WCAG20/WD-WCAG20-TECHS-20081105/H45> for the long description.

As  you noted, all the images are using H36, but they are using it in different ways, and the test procedure for the value of the alt attribute would be different in these two cases.

Success criterion 1.1.1 is a particularly complicated case, with a variety of different situations described in the Sufficient Techniques section. And, as we see, the ways of satisfying the success criterion may share techniques.

On Mon, Dec 1, 2008 at 3:30 PM, Dylan Nicholson <d.nicholson@hisoftware.com<mailto:d.nicholson@hisoftware.com>> wrote:
But that's tautologous - the technique IS the use of the LONGDESC element!

Correct. This technique describes the proper way to use LONGDESC for one image to satisfy Success Criterion 1.1.1.<http://1.1.1.>

The description of the procedure current strongly implies that every IMG requires a LONGDESC attribute, which is not correct.

The procedure describes the steps to apply to an image that is using this technique, i.e., check that the img element has a longdesc attribute. The procedure does not say to check that this is true for every image on the page.

This should to be contrasted with the ALT attribute, which is required for every IMG element, as per HTML/4.0 and XHTML/1.0 standards.  But oddly enough, the description of the procedure for technique H37 only says "Check that each img element which conveys meaning contains an alt attribute", which is also incorrect - it should say

"Check that every img element contains an alt attribute.  For img elements which convey mening, check that the attribute value is not blank".

As with H45, this is describing the check to be made for an image that is using this technique. The technique does not imply that it must be used for every image on the page. In the case of the alt attribute, validity would lead the author to use it for every image, but that is a separate consideration from the use the technique for a particular image.

This may be clearer when looking at techniques for other success criteria where the techniques may use attributes that are not required for validity. For instance, an author may mix a variety of techniques for abbreviations when satisfying SC 3.1.4.<http://3.1.4.>


________________________________
From: Loretta Guarino Reid [lorettaguarino@google.com<mailto:lorettaguarino@google.com>]
Sent: Tuesday, 2 December 2008 10:16 AM

To: Dylan Nicholson
Cc: public-comments-wcag20@w3.org<mailto:public-comments-wcag20@w3.org>
Subject: Re: Technique H45 procedure description incomplete

Any image *that uses this technique* must pass the check. However, different images, even on the same page, may use different techniques to meet the success criterion.

On Mon, Dec 1, 2008 at 3:13 PM, Dylan Nicholson <d.nicholson@hisoftware.com<mailto:d.nicholson@hisoftware.com>> wrote:
No, this is still problematic because it implies every IMG *should* have a LONGDESC attribute.

In reality, only a tiny percentage of real world IMG elements have LONGDESC attributes.

Having to check that the resource exists does make an automatic check somewhat less efficient, but given that LONGDESC is so rare, it shouldn't be an issue.

At this point I doubt we would bother with implementing an automatic check on the actual content of the page linked to.

________________________________
From: Loretta Guarino Reid [lorettaguarino@google.com<mailto:lorettaguarino@google.com>]
Sent: Tuesday, 2 December 2008 9:55 AM
To: Dylan Nicholson
Cc: public-comments-wcag20@w3.org<mailto:public-comments-wcag20@w3.org>
Subject: Re: Technique H45 procedure description incomplete

On Wed, Nov 12, 2008 at 9:20 PM, Dylan Nicholson <d.nicholson@hisoftware.com<mailto:d.nicholson@hisoftware.com>> wrote:
H45 is described as

Procedure

 1.  Check that a longdesc attribute exists.

 2.  Check that the link in the longdesc attribute is valid

 3.  Check that the long description describes the original non-text content associated with it.

Expected Results

 *   #1 through #3 are all true

But a) it doesn't explain on what elements we are to check for the longdesc attribute - presumably it should be IMG elements only in this case, and not FRAMES and IFRAMEs

b) surely it's not implying that to pass H45 every element that supports longdesc attributes must in fact do so??  That would cause virtually every page on the web to fail immediately, as I don't think I've ever seen a commercial site using LONGDESC - at least, not correctly.

c) what is meant by valid?  That it's a valid url?  Or it actually points to a resource that really exists?  If the former, fine (and a very good idea, seeing I've seen quite a few cases where the LONGDESC attribute actually contains the descriptive text in the value, rather than a URL pointing to where to find the text), but if the latter, then testing this automatically becomes potentially expensive on a large site (well, it would, if anyone actually used LONGDESC).

================================
Response from the Working Group
================================

Thank you, we have updated the test procedure to clarify these issues. See http://www.w3.org/WAI/GL/WCAG20/WD-WCAG20-TECHS/H45.html .

Could you let us know by Monday, December 8 whether you are satisfied with our resolution? If that date is not possible, please reply to this message indicating when you will be able to respond.

If we do not hear from you by Monday, December 8, we will assume that you are satisfied with the responses to your comments.

Thanks again for the interest that you have taken in these guidelines.

Loretta Guarino Reid, WCAG WG Co-Chair
Gregg Vanderheiden, WCAG WG Co-Chair
Michael Cooper, WCAG WG Staff Contact