+++ This bug was initially created as a clone of Bug #13726 +++

http://www.w3.org/mid/B6CB855C5769484F862F4FB2CCFA50F402D545A7@VHAISHMSGJ2.vha.med.va.gov

A phrase or paragraph with an alternative graphical representation:
charts, diagrams, graphs, maps, illustrations

This section is clearly meant to fill the role of the longdesc
attribute.  Unfortunately, the reasoning used is flawed.  It assumes
several things:

1. That all users require the same level of description for complex
information at all times;

2. That the alt attribute is adequate to the task of conveying textual
descriptions of graphical information.

At the VHA, we encounter graphical information in the form of
screen-captures of what would otherwise be tabular data, or charts with
complex interrelationships of information.  If the suggested
implementation here were to be used, a user (such as a screen reader
user) could encounter something like: 

<img src="screenshot.jpg" alt="Table with five rows and five columns
listing types of conditions and their severity. The first cell (column
1, row 1) is empty. Column 2, Row 1, says Pulmonary; Column 3, Row 1
says Circulatory; Column 4, Row 1 says Muscular; Column 5, Row 1 says
Neurological. Column 1, Row 2 says, Minor Concerns ... Column 5, Row 5
says coma"> 

Not only would this be tedious to listen to for anyone who just needed
to know that the image was a screenshot with lists of different types of
ailments, it would not allow them to explore the layout of the table in
a 2-dimensional way, to understand the relationship of the tabular data.

The argument could be made that this information should have been
provided in tabular form anyway, without the graphic, except that the
purpose of the graphic is not just to convey the data itself, but to
show how it would be presented in a particular computer application for
a user to select from.

It would be much more useful to provide multi-layered description (such
as could be available with a modified and updated longdesc attribute) so
that the image would be presented as:

<img src="screenshot.jpg" alt="screenshot of ailment selection screen"
longdesc="LinkToTableInformationForThoseWhoWantIt">.  

This would also be true of complex charts describing multiple factors
that interact, or in less formal items like clothing catalogs where both
an alt="short-sleeved shirt" could be supplemented with a
long-description that gives details to only those who need more
description.  Those who were not looking for ANY type of short-sleeved
shirt would not have to deal with the description of the fabric and
color, etc.

However, if a new form of long description is implemented, the
description of how to utilize it should be made much clearer, and
user-agents should be able to make that information available to any
user who wants it; not just screen-reader users.  In some ways, it would
seem that adapting the details element to this purpose could work nicely
(perhaps also becoming an attribute of img, or there should be a way of
adapting it to an "additional information about images" role).  Coming
up with a details-as-attribute or being able to apply a details element
to an image could alleviate one of the problems with longdesc - far too
few people knew what it was meant for.  Since the details element is
described as being something that is used to provide further information
on text, being able to apply it or something like it to img would
basically do the same thing - provide a clickable element that indicates
that there is more information available.

[split out from bug 13590]