Minutes ov 4 November Meeting

Here are the Minutes taken by Al Gilman

Wed., Nov, 4th, 4-6 pm EST  (i.e. Boston Time)

** On the call:
Len Kasday - chair
Chris Ridpath
Al Giman - scribe
William Loughborough
Gregg Vanderheiden
Daniel Dardailler
David Clark
Larry Goldberg
Harvey Bingham

** Decisions/actions:

Next meeting: penciled in for 3: p.m. EST, December 3, 1998

Invite webwatch-l to review trouble reporting site: Len will
extend a special invitation on webwatch-l for people to review
the propose site setup, when Daniel says it's ready for that.

Investigate ATRC UseNet applicability: Len will approach Jutta
about our interest in this through WAI-CG.

Mail out access information for A-prompt:  Len email distribution
list to Chris Redpath.  Chris mail our A-prompt access instructions.

Table linearization algoriths: [Daniel?] Refer "what
linearization algorithm will WAI-UA and browsers support?" to
inter-guidelines issues list (high priority right now).


Work Areas

1. The web report form.

Discussion:
- being done by summer student who is now gone
- there is working code for the first one-pass concept
- the present concept is a multistep process for which there is
	nothing behind the mockups on the website
- Daniel strongly leaning to a mailing list as the connection
	between the original form filing and follow-up discussions
	and data collection.

The subscribers to the webwatch-l list were discussed as possible
users of the form.  These are people who care about cleaning up
the accessibility of the web site.  If it works for them, it's 
good for something.

Len will extend a special invitation on webwatch-l for people
to review the propose site setup, when Daniel says it's ready
for that.

Also discussed using reports gathered in this way to seed more
systematic surveys of the state of accessibility across the web.

Robots came up in this context and were deferred 'til 
the end of the list.

Discussed possible wrinkle where site that a public visitor
complained about would then be reviewed by a web technician 
from a stable of continuing workers e.g. ER IG troops.


2. Bottom Line evaluation (pass/fail).

Lots of consumers want it.  Want to know have they done enough.

It's good to try to give such a decision.  Not clear that we have
the necessary knowledge in hand at this time.  Want distribution
across disabilities, real user testing to be sure.

Form results would be a way to get started, even if not so
scientific as we would want for an "approved" cutoff.

Toronto has a UseNet setup with a ~600 user pool of evaulators.
Seems like the killer evaluation setup; would want some of this
perhaps before we declare a pass/fail threshold.  Al said that
he thinks using this resource takes money.  Len will approach
Jutta about our interest in this through WAI-CG.

3. ALT text inserters for users

Several people voiced skepticism that these techniques could
be made to work.

The idea came up of using targeted robot studies: pick a
technique and see how often the technique could come up with
anything useful.  Depends on someone stepping forward to
prototype the methods.  Mike V. is planning to do some of this,
but he was not on the call to comment.  Daniel pointed out that
the basic robot engine is available as part of the W3C maintained
wwwlib software library.  Doesn't take much to extend this to do
one or two experiments.

The targeted robot studies idea seemed worth keeping on the
table.

4. ALT text inserters for authors

There was strong comment against something that would auto-insert
an automatially-generated ALT and make it easy for authors to
simply accept it.  This would totally fool automatic tools like
Bobby and leave the user listening to garbage, all too likely.
Author should have to do something, should have to think at least
a little.  

On the other hand, if the easy cases (e.g. bullets) are
automated, users are more likely to take the time for the harder
cases.  Some of these are already implemented by the authoring
tools when you insert the same bullet again.

Al talked a bit about the idea of a synthetic context that would
"frame the question right" for authors when they are writing ALT
strings.  Turns out that what he is really thinking about is
links.  The text that previews the other end of the link may
belong in the anchor TITLE, not the image ALT. Will have to check
with the PA guidelines on that.  This was another item where the
conceptual robot study is pretty simple.  How many pages have a
<TITLE> or <META SUMMARY> and is it any use?  Or what algorithms
do the search pages use?

New notion: ER IG pull together criteria (ALT critic) for
authoring tools to apply that would question obvious bad choices
(not a word, ALT=SRC on IMG, etc.).

Chris Ridpath had had an earful and volunteered to inject some of
these ideas into the WAI-AU working group.

Aside: Len asked if Chris would mail out A-prompt URL information
if Len gave him a list of the ER-IG subscribers.  Chris said yes.
Action: Len mail the list to Chris.

5. Audio browsing simulators

Al described a possible way it could work, with text initially
greeked (obscured) and then as the user tabbed through the links
the current link would be clear and the previously read text be
grayed out.  The point is to try to convey via visual tricks the
flow of the audio browsing process.  Not have to convert graphics
jockeys to paying that much attention to their ears.  Don't know
if it will work until we try it.  Author tools might pick it up;
some of them now offer browser-emulation modes.

A server site hosting a phone browser is an alternative way to
get the idea across.  Al/Len to work offline to explore
possibilities.

6. Tables.

Table navigator as DOM application is one approach.  Has been
some work on this by Mark Novak at Trace.  Too rough to share
as yet.

Another is to spell out the navigation options in HTML hotlinks
in an exploded view of the page.  Len and Daniel have looked
at this a bit.

Tables need to be written to make linearization work.  Discussed
nested tables as a technique to force column-wise reading of a
table, instead of using a tool to transform it.  This raised the
question of what WAI-UA and browser implementors can agree is the
algorithm.  There is an algorithm in HTML 4.0.  But is that what
browsers are going to implement?  Refer this issue to the
"guidelines coordination" process via WAI-CG.

Ran out of time.   Set meeting for 3: p.m. on Dec. 3.  Topics
left:

7. Heristics tests for "good" ALT text

8. Structural fixups, e.g. formalizing closet list structures,
DIV structures, etc.  See results of existing M. Vorburger program.

9. Robots (actually discussed some under 3. and 4.)


-------
Leonard R. Kasday
Institute on Disabilities/UAP at Temple University, Philadelphia PA
email:     kasday@acm.org
telephone: (215} 204 2247

Received on Tuesday, 17 November 1998 09:11:13 UTC