Comments on 30 July 2004 WCAG 2.0

Dear WCAG WG, 

I have a few comments on the 30 July draft of WCAG 2.0
for your consideration [1]. 

Thank you for all of your hard work,

 _ Ian

P.S. I did not review the glossary.


[1] http://www.w3.org/TR/2004/WD-WCAG20-20040730/


======================================================

1) I think the "Overview of Design Principles" should be
   closer to the front of the document. It's a nice overview
   of what is to come.

2) 1.1, Level 1 SC:

   - a, b, and c are not mutually exclusive. Are several text
     equivalents required if all are true?

   - "non-text content" is not a closed definition and therefore
     cannot be tested. I think this term needs a much stronger
     definition.

3) 1.1, Level 1 SC:

   - I don't know how "real-time" and "not time-sensitive"
     relate. Please clarify.

   - In 5., I don't think the distinction between "providing a 
     substitute" and "providing a link to a substitute" is
     significant. If, in general, I cannot satisfy the
     requirements by providing a set of linked resources, then
     I think that's a larger problem. 

4) 1.3, Level 1 SC:

   - "Structures and relationships within the content can be 
      derived programmatically". Does that mean _all_
      relationships? If I say "See paragraph 2 for more
      information" that is a semantic relationship that I express
      in English; in the general case, such semantic
      relationships cannot be derived programmatically. 
      What is a structure?  I'm not sure how one would test this
      success criterion.

5) 1.4, Level 3 SC:

   - "to determine if the background makes it difficult to
     identify individual characters" is not part of the
     requirement. The requirement is: 
 
      "Text is not presented over a background image or pattern,
      or if a background image or pattern is present, the text is
      easily readable when the content is viewed in grayscale."

     Even written that way, "easily readable" will be highly
     subjective.

6) A number of the requirements still seem to me to be
   requirements for user agents, not content authors. These
   include, but are not limited to, parts of Guideline 2 and 3.2
   SC 2 bullets 2 and 3.

7) 2.4, Level 3 SC, bullet 5. This is one of those assertions
   within an assertion. I still believe that one conformance
   claim suffices, and subsumes all other assertions such as
   this one. The requirement is "break up content into
   paragraphs". Claiming conformance means saying "I claim to
   have satisfied a lot of requirements including the one
   for breaking up content into paragraphs."

   I believe the WCAG WG is using this mechanism of
   assertion-rather-than-requirement so that they have something
   to test: "Is this sentence present, yes or no?" However,
   writing the guidelines this way does not actually help authors
   to understand what they have to do to make the content more
   accessible. In the end, if this approach is adopted, I believe
   that people will simply make the statements without actually
   making the content more accessible.

   You will serve the community better by ensuring that the
   requirements are clear, and where they cannot be made clear, 
   removing them.

   This comment also applies to 3.1, SC 3, bullet three.

8) 3.1, SC 3, bullet one: What is an "associated dictionary"? If
   I understand this requirement correctly, I would be required,
   for every page that I write, to verify that for every word I
   use, the meaning I intend is the first one that appears in
   a chosen dictionary. That seems entirely unmanageable.

9) 3.1, SC 3, bullet two: "Section headings and link text are
   understandable when read by themselves as a group".

    a) the phrase "by themselves as a group" should be made
       to sound less self-contradictory.
 
    b) Does this mean "All section headings and link text",
       e.g., if I have 75 in my document?

10) 3.1, SC 2, bullet 1.

    - "The meanings and pronunciations of all words in the
       content can be programmatically located."  The definition
       of "programmatically located" is: 

        "the meaning can be found, though there may be multiple
        meanings for a word." 

       I think that the definition is untestable. I note that it
       is in development.

       Natural language is a moving target. I fear you are not
       going to be able to ensure that every term in a human text
       is understandable by all readers, even with online
       dictionaries. And if meanings change, content that was
       once conformed may no longer conform.

       By the way, is that property of content discussed: If your
       content conforms and you don't change it it will continue
       to conform. If that's not a property of conforming
       content, please be sure to say so up front.

-- 
Ian Jacobs (ij@w3.org)   http://www.w3.org/People/Jacobs
Tel:                     +1 718 260-9447