Re: what type of document do we want?

Paul is correct in his analysis, which is exactly in line with my own
thinking. The action item resulting from the teleconference, however, does
not call for a crude categorisation of checkpoints into "subjective" and
"objective", respectively. Rather, it demands a more explicit
acknowledgment of which aspects of each checkpoint can be verified in
accordance with clear and definite criteria, and which can not. Of course,
this judgment in itself depends partially on the technical mechanism
employed. To take the example of checkpoint 1.1, most markup languages of
contemporary interest which permit the inclusion of auditory and graphical
content, provide elements or attributes (or comparable mechanisms) whereby
the text equivalent is directly associated with the medium-specific
material to which it applies. Examples including the (X)HTML ALT attribute
and the SVG DESC element.

Now suppose that the text equivalent is not directly associated with the
auditory or graphical content, but is instead presented in, for example, a
logically separate CAPTION element which has no relation, in the markup,
to the non-text content, or is given in a separate document (e.g., a
transcript). Under these circumstances it will usually be possible to
determine, through the exercise of human judgment and in a relatively
uncontroversial manner, whether or not the text equivalent has been
provided. Any concern regarding the accessibility of the content, rather,
would centre upon the different but related question of whether checkpoint
1.1, is (or should be) intended to require a logical connection,
discernible in the markup, between the equivalent and the content to which
it relates; or whether the existence of a separate caption, a transcript
linked to the principal content, etc., would suffice.

Taking the analysis one stage further, suppose that instead of
establishing a one to one mapping between auditory and graphical
presentations, on the one side, and text equivalents, on the other, the
author has decided to rewrite the content, thereby creating a second,
completely textual, version which provides the same information but
without reference to auditory or graphical material. Needless to say, such
a rewritten version may be easier for individuals who are unable to
perceive the non-text content, to understand (for example it might
offer explanations that can be readily comprehended without reliance on
auditory/visual concepts and metaphors that could create confusion).

Now the question arises of whether this "alternative version" of the
content can genuinely be regarded as an "equivalent" for purposes of
checkpoint 1.1. Even if the checkpoint were construed as allowing such
purported "equivalents" to be counted, for purposes of the guidelines,
opinions may, and probably would, legitimately vary as to whether, in such
a case, the two separate versions of the content were genuinely
equivalent (whether, for example, the same meaning or information, so far
as possible, was being conveyed in both). Under these conditions it is not
even clear, without further argument, whether there is a text equivalent
associated with the auditory/graphical presentations, let alone its
adequacy. The more flexibility that the guidelines offer in regard to what
can count as satisfying a requirement, the more difficult verification
becomes, a problem which Len has pointed out on several occasions. He also
raised the pertinent question, which is yet to be answered definitively by
this working group, of whether, and if so to what extent, the availability
of clear verification criteria should be taken into consideration in
deciding what does, and does not, amount to adequate implementation of a
checkpoint. For example, if verification were considered important, then
one would be inclined to interpret checkpoint 1.1 as requiring an explicit
connection, in markup, a data model or metadata, between the text
equivalent and the auditory or graphical content to which it applies. If
verification were less of a concern, then one could say that the author
who has provided two different versions of the content (one completely
textual and the other employing auditory/graphical presentations) can be
regarded as having satisfied the access requirement.

The guidelines, I think, are similar to natural language, in that with
respect to each checkpoint there tends to be a core of definite,
paradigmatic cases, surrounded by a penumbra of dubious, or at least
disputable instances. This is likewise true of the meanings of words in
natural language. To mention another example from the guidelines, it is
usually clear (to someone familiar with the specification of a markup
language, for instance HTML 4.0) whether elements and attributes have been
used in accordance with the definitions provided in the specification; but
there are always questionable cases. For instance, does the use of HTML
4.0 definition lists in the guidelines document itself, to distinguish
between the text of each checkpoint and accompanying explanatory material,
constitute a proper use of "definition lists" within the definition
provided by the HTML specification? I do not want to start a debate on
this specific point, but rather to indicate that it is a question on which
there may be legitimate disagreement; and other illustrations could
readily be cited.

As Paul has suggested, there is indeed an unresolved question regarding
precision with which the guidelines should define what does and does not
constitute compliance with each checkpoint and the role which
considerations of testing and verification should play in the making of
that determination.

Received on Friday, 30 March 2001 19:57:40 UTC