Re: update of WAI-ARIA Graphics Module and aria-gtype

Hello Fred & all,

Thanks for all your work separating out the two graphics modules, Fred.

Below are a description of some edits I have made Level 1 module, and then
a response to Fred's concerns about gaps and limitations of the restricted
role model.

A reminder that I will not be on the call Sept 26.  Feel free to resolve in
my absence that all of my changes are awful, but please don't then assign
me ten actions to fix everything!

Amelia


Summary of Edits:
--------------------------

My changes to the aria/graphics.html spec and related example files are
covered by the following pull request:
https://github.com/w3c/aria/pull/93

Until the changes get integrated in the main repo, you can view the
compiled file from my fork:
http://rawgit.com/AmeliaBR/aria/aria-graphics/aria/graphics.html

Many of the text changes are editorial, trying to soften the jargon.

Please carefully review, however, all instances of SHOULD and SHOULD NOT
statements, some of which Fred introduced and some of which were mine.

I've added more explanatory text for my "house" example.  I think this
effectively demonstrates the distinction between semantic and stylistic
groups which you were emphasizing in the new "circle of months" example,
Fred.  I've therefore commented out the months example, but haven't deleted
it in case you wanted to argue for its inclusion.

Finally, I've expanded the "Other Roles for Graphics" section, with a
series of examples comparing img, figure, and graphics-doc.  I've also
re-instated the figure role description as an editorial note for now, so
that we have a reference copy of it until it can be integrated in ARIA 1.1.


Remaining gaps:
------------------------

Re aria-gtype: As we've discussed on the call before, I am concerned that
you're trying to make one property do too many things, Fred.  It cannot be
both a flexible author-supplied description and an AT-interpretable
category for controlling reading order.

While I appreciate that effective navigation of a chart requires a reading
order separate from the normal drawing order, that's why we will follow up
this simple graphics role model with a chart-specific taxonomy.  For
generic graphics, I think we should be relying on aria-flowto and aria-owns
to convey the correct reading order.

Re labels being potentially read twice:  This is an important issue.  I'm
sure it could quickly get frustrating to navigate a diagram with both
graphics and labels causing the same text to be read out.   Yet, this isn't
a new issue: HTML forms make extensive use of labels, and I think
screen-readers are able to keep track when reading both the visible text
and the descriptions of the corresponding inputs.

Perhaps the regular screen reader users in the group can offer suggestions
for how this should be handled?  Then we can offer better guidance in the
spec, and decide whether a separate role is required.