Re: Desc and Alt

In response to Al's query, it should be pointed out that although the
content of OBJECT is normally ignored if the resource which it designates
(for example an image or a script) can be processed, this is not the case
with an image map for which the boolean SHAPES attribute is specified.
Under these circumstances, the visual user agent processes the content of
OBJECT, searching for anchors that provide the coordinates of the various
regions of the image map. The text of the HTML specification, as currently
drafted, does not make this point sufficiently clear (it states that a
visual user agent will process the content of OBJECT, looking for anchors;
whereas it should qualify this by saying "anchors with COORDS
attributes"). This is a minor point relating to the accuracy of the
specification which is not directly an accessibility concern.

Moving to Al's more substantive concern, Daniel and I have given some
consideration to how LONGDESC should be presented to the user. The general
conclusion is that it should be controlled by style sheets. In practice,
of course, LONGDESC behaviour would most often be specified by the user's
default style sheet, since most document authors are unlikely to make
specific provision in their style sheets for non-visual media, even if
they take advantage of LONGDESC itself. LONGDESC would, in general, act as
a link to the description, which would be placed at the end of the current
document or, more typically, in a separate document. The presence of
LONGDESC could be signaled to the user by changing the speech or braille
formatting of the ALT text associated with the image, or by inserting text
next to the ALT text (for example the familiar "[D]" to indicate the
availability of a description).

There are, of course, limitations of CSS 2 syntax which prevent the
rendering of LONGDESC from being controlled effectively by styles. Daniel
is investigating these problems at the moment to find out whether they
will be resolved as part of the ongoing CSS development work. An
alternative would be to require the user agent to convert an image with
LONGDESC into an equivalent OBJECT element prior to applying styles to the
document. Such an OBJECT would have a class of LONGDESC, and automatically
generated content (for example, an anchor containing the text of the ALT
attribute from the original image). The OBJECT could then be addressed by
familiar CSS selectors.

Personally, I would prefer to extend CSS so as to permit the rendering of
LONGDESC to be specified directly, instead of relying on user agents to
transform <img src="image.gif" alt="alt text" longdesc="description.html">
into <object data="image.gif"> <a class="longdesc"
href="description.html"> ALT text </a> </object> or similar markup.

Note that an alternative approach would be to import the contents of
description.html in the above example directly into the document.

To meet short term requirements, the types of behaviour mentioned above
could all be proposed as alternatives which text-based user agents could
implement, even if they do not support style sheets.

Received on Thursday, 9 October 1997 19:37:12 UTC