minutes from today

Here they are in rough form...

SW - Susan Warren (Project manager for Visual Studio web forms - gave a
        presentation in the morning)

ASW - Andi Snow-Weaver (IBM - Stranded WCAG member)
LGR - Loretta Guarino Reid (Adobe, WCAG member)
CMN - Chaals McCathieNevile
WC - Wendy Chisholm
KHS - Katie Haritos-Shea
HS - Heather Swayne
GB - Giorgio Brajnik

on phone
JT - Jutta Treviranus
JR - Jan Richards
PJ - Phill Jenkins
WL - William Loughborough

NB there are a couple of drawings and screenshots referred to in here.
Screenshot 1 was lost, screenshot 2 was sent to the list, and the drawings
and remaining screenshot along with descriptions will be added to the minutes
in web form.

CMN This is a public, archived exercise, not under Non-disclosure Agreement.

SW These are a way to do web pages - sttatic content coming from a database,
but also programmable content, stuff that supports backend processing -
equivalent of java/asp. ASP isn't limited to HTML as a markup language.
http://www.asp.net
For mobile authoring it will author WML, or speech versions.
CMN Can you use it to take one set of content and output it in either HTML or
WML (for example)
SW Yes. We do it differently - encapsulate the transformation and the server
side logic. For example interaction with a database for an application.
[[Drawing 1]]. The rendering form is a decision that can be taken later.
Exception is HTML graphic apps, where high graphics means it doesn't morph
nicely.
JT Example - we have a database for courses at the university, and we then
render it on palm pilots, or on printed forms, or whatever
SW Yes, but there is another piece - "encapsulation". You're doing
transformation - it is a one-way trip. Encapsulation allows the information
to be rendered, and inside a "control" it kows how to interact with that
encapsulated unit. Typically the rendering is abstracted, so you can have
adaptable user interfaces.
GB Same thing that goes between a data structure and object to provide a
protocol for behaviour.
SW Yes, it is more about the UI level. We name a button, that could be
different for different renderings. The big difference is this encapsulation
of behaviour
CMN How does this go beyond Xforms?
SW Similar. (I am not an Xforms expert so can't give a detailed comparison).

==demo
SW Working with a webform. Because what is shown can change, navigating often
comes back to the same place - because it is possible to encapsulate a lot of
function

.. put in a ouple of text-entry things, add labels, make one a password entry
(obscured text).

CMN are labels attached to text-entry?
SW No.

.. add a validation control (require entry in both fields) Can but a marker
on the page, and a place for error messages to be returned.

JT You can assign a keystroke for a button control?
SW Yes. For example there is an accesskey property

[[Screenshot 1]]

ASW can you run this from the keyboard?
SW Not me, but we have a tester who does.
HS Visuall studio is built in an environment so it is tested to make sure it
should support this, as well as the individual app.

..  looking at code - this is like HTML but with some extra tagging in it to
give us terse control representation. These are transformed by the server at
run-time to HTML (or whatever).

[[slide deck 1]] http://www.swarren.net (slide not found - replaced by
[[drawing 2]])

.. Code is compiled and converted into real code - each of those things is
converted into a control added into the program. Programmers manipulate the
control objects, and then goes through a render phase, where we say what the
target format/capabililty is.

**WL joins

See it rendered in a browser
Trying to post without entry brings up error messages.
HTML source looks like HTML - because browser is targetted it uses
client-side script for validationi in this case, but could use server side.

====
CMN Does it do any accessibiltiy checking - for example identify that it made
a form control without a label
SW V-S allows the user to create anything. There isn't a utility in it for
checking accessibility at the moment.
HS It is still in development - will ship "sometime this year".
SW We are a couple of months from shipping, so we are in feature lock-down
HS This will be one of the things we take to the V-S team for the next
version.

SW Other piece is the controls. The framework itself renders a name on every
control - for every control there will be a name - may be autogenerated.

KHS what about URIs beng the same when the contents change.
CMN It is not possible to come back to the middle of a process - you need to
come back to the start.
SW Let me capture that as an issue.

CMN If there is a control there should be informatin for the user. For
example, check whether there is a label for controls, or whether there is
information provided when validation fails.
..discussion.

SW V-S can put warnings and labels into something - there is a tasklist so
you can decide when to do things. I think that is the appropriate place to
put errors, and label them as such so people can filter on them - access
errors and code syntax errors are different things...

Thanks to Susan for demo.
**SW leaves
JT Since we can render this on different devices is there a place where you
specify the logical sequence of these pieces, can that be nested?
Parent-child relationships can get quite lost - is it just dependent on how
it is visually laid out.
HS As I hear it - the developer is divorced from the end UI so how can they
get the tab-order right? I am not positive of the answer, and would like to
go to the developers.
JT more than tab order - stuff that goes with the control.
CMN I suspect that the "control logic" of interactions is encapsultaed in the
middle piece of [[drawing 2]] where there is a tree developed. But intersted
in the answer from HS. This is something that Voice applications have to have
under control, because of the interaction environment.
WC Is the positioning absolute?
HS As far as I know it is relative until it is compiled to a rendered form.
WC I would have liked to see how it was rendered into other formats...
JT have the results generated been tested with screenreaders?
HS The team uses internal tools to test the outcome and how it works with
MSAA. We have done some spot-testing, but tend to rely oin testing tools?
WC Testing tools area available?
HS All are available free in the MSAA development kit (except little
special-purpose tools developed by internal teams)
===============

Agenda:
Jan's evaluations 1 minute, Acrobat, Graphics/whiteboard, AERT.
Tomorrow:

Jan's couresware evaluations:

JR Looking to evaluate courseware, so this tmplate lays out ATAG, and lists
relevant WCAG guidelines, and underneath each point is a yes/no/anot
applicable with an explanantion piece. It is an intersting way of organising
the way that tools conform or not to relative priority checkpoints. In terms
of results there is progress in help as well as integration of features.
Didn't prevent users doing stuff, but there is a gaping hole in terms of
prompting and helping users proactively, and in terms of checking. Have not
published results yet.
JT Have sent them to developers - will talk to them to address issues. For a
template it is better than the ATAG guidelines - nice layout, links to useful
info.

================

Acrobat:
LGR: Couple of directions we can start in - looking at ways to generate
tagged PDF (accessible stuff) or we can look at tools within acrobat that let
you check for and try to fix accessibility problems.

(LGR talking unless otherwise noted)

From Word, Excel, and Powerpoint, you can generaate tagged PDF.  Macros are
installed for these products, when install Acrobat.  Adds menu and buttons.
In Word, change conversion settings dialog, we can see pdf maker settings.
This doc has a table, press PDF button to make it a PDF file.  Maintain as
much info about logical structure as it can.

WC When converting HTML to PDF, is there something in the HTML that you can
include, a comment or some sort of markup, to generate page breaks in the
PDF?

LGR not sure will find out. To generate them, Pagemaker 7, make accessible
plug-in (access.adobe.com), and the office suite, converting from html via
acrobat can create tagged pdf.

Back to demo:
Word has written the file.  I'm opening it up in acrobat.

JT What support are in the tool to support tagging.

LGR Which tool?

JT At any point, do you specify what the tags are?

HS My suspicion is that they are relyin gon the underlying data.
LGR Not sure about that.

JR What does the make accessible do when you turn it on.

LGR it controls whether tags are generated or not.  Possible to generate PDF
w/out tags if you turn off the switch.  e.g. to make as small as possible.
Default is to generate tags.

File open in acrobat, looks pretty similar.  I'm going to the windows
toolbar, selecting "tags" window.  We can see the tags that are there.  We
see a <sect> element with a table element, with the TR elements, then
drilling down further, see the TDs.  The conversion process did not recognize
the column headers.  Word did not identify them as such.  We can fix this up
by hand.

KHS That was a person who created it...is there a way in word to make it more
obvious.

CMN In here, marked as heading.  e.g., <TD><heading 3>

HS Pulling out the style of the font.

CMN In the second row, we see, <TD><LI>

LGR This is picking up the style that is used for that element.

HS this is a difference in that you want to mark the heading row as such, not
as a heading style

LGR I have nto been able to create a word in table so that it knows that the
first row is heading.

CMN Reckon we could figure it out.

LGR I'll manually fix it up.  Select elements in the table and could type in
the structured view, but will open up the properties.  Will turn the type
from TD to TH. Can associate a title.  If I need to attach alt-text can
associate to it. Can also associate actual text. Intended for a graphic that
is representing text. Like a drop cap or illuminated letter.  Alt-text is
more general.

CMN refering to equivalence and identity, actual text is equivalent, alt-text
is alternative.

GB Generic properties for all elements?

LGR .... reflow depends on that info.

HS Validation on the type?  Could an author create own type?

LGR No.  there are a set of basic types.  there is a way to let the author
map theirs into the basic types.  the original goal was for touching things
up rather than constructng from scratch. Can edit but not create new (roles)
today.  No validation along DTD lines.  Those are enhancements that could
come along.  Fewer restrictions than other markup.

This is the wrong place to do most of this editing, since each time you
regenerate you'll have to reedit.

CMN DOn't generate it back?  Not really an editing environment a donversion
process.

LGR And evaluation process.  Long term, this is not the way to do things. But
it is there, you can fix the problems.

HS You rely soly on 3rd party editors.

LGR Any thing generate PScript, few know how to generate tags.

HS any info tha tyou need from, e.g. Word, that we could provide?

LGR need them to generate tagged PDF.  There is a description of what is
needed today.  I can get that info to you from the team.

CMN An editor that a does not carry alt-text, nothing the tool can do to
create one.

LGR We would like to see the app itself generate the PDF.  <laughter>

The team that worked on these macros, took a lot of effort to get info out of
Word.  Word is probably better than most people.

HS Are they going into the native objecct model of Word?  Not sure what info
not contaiend there.

LGR I'll make the contacts there.

ASW The people that do the PsCript (the intermediate step) is there something
they can do?

LGR They would use pdfmarks (a pscript extension).

CMN Amaya generates postcript, they could use that to get tagged pdf.

LGR yes.

CMN I expect they won't, o nly 3 of them.  But nice to know it can be done.

ASW is that extension documented?

LGR On my list of things to find out when I get back.

This structure tree determines reading order, i.e. the order it gets passed
to a screen reader.

turn on associated content when click on tree will highlight in actual doc.

ASW all keyboard accessible right?

LGR not today.  we focused on the reader as the most important first step.
it is on the list, however.  pretty tricky.  if i want to turn something into
a link.  it is one chunk of text. i have to break it into 2 text items.  Can
create new content item in the tree.  Can go between the views, not intuitive
or as easy as might like.  but as i've said, this is for use by people
familiar with pdf.  this really should be handled by the editor, but if it's
not, at least it is possible to work at this level.

ASW How many usability engineers involved?

LGR The original goal was fixing minor things but it has evolved to editing.

HS Dev tool shipped to the public.

CMN It does checking things.  e.g., "did you really mean to do this."  cool.

LGR Not accessibility related.

CMN Interaction model that exists.  In Word, get squiggles everywhere.  Here
it pops up and asks.

LGR This is the view into the logical structure and fixing to the degree
possible.  But, this is the wrong place in the workflow to make accessibility
fixes. It's necessary today.

CMN In this doc, there is a table.  several rows, 4 columns, 1st row is
headers.  There is a sublist under row 3.  Each are rows. Is that captured in
the markup?  In XHTML, couldn't do it either. Breaking lists across table.
Use RDF to say "these are all part of the same beast."

HS Wonder if word object model can go across those borders.

ASW Ordered list vs. unordered list?

CMN Some model?

LGR It's in the text.  Or also use label it.  The representation will be in
the pdf, not genereated.

CMN Word object model? It does differentiate.

HS Right, i want to play with it. It does know how to continue  a list.  It
understands sub item or not.

JT I have a huge medical text with a lot of structure only in PDF, how do I
get at the structure?

LGR These elements will be mapped into MSAA, but we kind of lose a lot of
stuff

HS So, have you looked into iTextStore - with winXP new version of MSAA has
this - big advance, along with dynamic annotation. iTextStore iedntifies
document stuff.

LGR cool. We are also looking at a kind of DOM

CMN Can you attach semantics to these elements?

LGR You can create your own structure element types.  You want to map them
into the underlying types since we do special things w/them.

CMN You can say, "this is a type of paragraph."

HS My fear - you don't hafve a way to express that through MSAA.  Strict
roles and responsibilities.

LGR Defined roles.

CMN Can you annotate those roels?

LGR I will find out.

JT Can export to XML?

LGR Yes, a plug-in can export to HTML and export to XML.

JT Can take word, export to PDF, and export to XML?

LGR yes.

CMN Well-formed XML?

LGR believe so.  Can try these today.  Not too many tagged pdf files exist.

CMN increase in size?

LGR maybe 10% growth, although some have much more. really depends on the
content.  looking at how to make it better.  default excel not to tag since
penalty is so great.

GB Tagged files that are genereated already, by whom?

LGR The list of tools we listed earlier.

/* opening  a web page in Acrobat 5.  but it's an XHTML document.  found
something it couldn't process, but we do have it in PDF. */

CMN What is the article element?

WC Likely, since can collect a series of HTML documents into one PDF, each
article element contains each separate doc.

LGR i can find out more about it.

checking that character encodings are correct. we need to make sure these are
correct. depending on font, might not be able to do that.

CMN Make sure that characters we want are the ones we are using.

LGR Accessibility Checker Options.  Currently generate a log of issues, EARL
would be a possible format to save them in. Attach annotations in doc where
problems occur.

/* running accessibility checker. */

/* would be really cool if could pass back tagged pdf w/earl accessibility
report, however problem is that pdf does not have mapping to word file.  */

LGR Report comes up with errors - missing alt-text and chars that it can not
map to unicode characters.  Problem with bullets.

CMN These are bullets of list.  Check if marked as a list.

LGR Yes.  Which does it see missing as alt-text.  It's the HRs.  They should
be marked as artifacts.  At the checker level, can see that something being
drawn on the  page does not have text.

GB How does a screen reader read a bullet list out of PDF?

LGR Assuming get unicode translation, pass it that.  We have to look at itext
storage tech to see if way to communicate. Today can't communicate that.

CMN A good thing to solve.  List labeling, etc.

LGR Getting more of the doc info available is definitely good.

CMN Want to see untagged thing and make it accessible.

/* downloading  makeaccessible plug-ijn from adobe.com */

JT I would love to hear if anyone has  input on the topic of techniques
whiteboard graphics tools.  Do you have specific SVG thoughts?

CMN I have a couple of thoughts.  Mostly about prompting and hassling for
alternative content, structure into graphics that are generated.  One of the
techniques is to build pictures up by structure rather than drawing. Doesn't
always work. People would do, like me, who can't draw.

JT A pallette of structures.

CMN VIsio does this.  27 palletes of objects. Connect the pieces.  In fact
you can have a list of names.  You start with circles and arrows diagrams.
Then say this circle is cats, put in a cat symbol.  If very structured,
works.  e.g. process flow diagrams.

JT Another thing we've been looking at. crude image  recognition.
interpreting input gestures, translating into circes, etc.  Pattern
recognition.

CMN You end up w/the same rpocess of having a struct that is loose. have to
check it to determine if what really being communicated.

JT How to ecourage people to not write text free hand, but to use other text
tools to put text in appropriate place.

CMN A good range of fonts to select from, helps. People find it easier to
type than write.  Handwriting translation is crude but available.

JT A way to say that someone is part of this discussion that requires alt
format We've been coming up with text that is not too intrusive to encourage
people to annotate.

CMN 4 sharing whiteboard. someone can't see it.  sets a flag. turns on higher
prompting level.  others get prompted to create a text description.  doesn't
have to be the one drawing it.  these can contribute to scenarios and
techniques.

HS If we have a whiteboard, would it be next to them, across the table, or
person drawing.

CMN It could be any of those.

JT Redundancy is not bad.

WC Could also set up that someone takes responsibility.

CMN Yes, or could distribute the whole exercise. Like IRC.

JT Aside from having a human describer.  If relying on pallete of objects,
what mechanisms can we use to describe relationships between them?

CMN Look at SVG.  Typed relationships between pices of a picture.  Some are
relationships: inside, ontop,e tc. Some can be mathematically determined.
However, might have to say if putting something In the box or on top of the
box.  Danbri and Libby have "codepiction."  Everyone is 6 degrees separation
from person X.  To find that out, get two photos. Say who each is. Record in
rdf.  They draw maps of who knows who.  Grab a piece of image and say "this
piece is charles, this piece is giorgio."  Another way to get the info in
there.  In shared whiteboard context don't assign one person to label
everything.  Some will be prelabeled others won't.

JT Is there a convention to express that?

CMN Yes, most of what SVG does.  Pieces have position in drawing space.

JT I know they would need to express, but is transformed into something
semantically understandable?

CMN you mean, spacial coordinates to "on top, next to," etc.

JT right.

CMN SVG Only does 2d graphics, not always clear.  Can put into SVG or RDF
easily.

JT How many clusters can you create and create an uberlable for.

CMN Infinite, barrng memory constraint.

JT Fly on the nest on the tree ...

CMN In the bog on the ...exactly.

JT Nesting strucutre. label at each level.  We seem to have good techniques
for whiteboard.  We'll be working with courseware devs to test w/peple.
We'll inform you as results come in.

CMN Another piece is pattern reocnition stuff.  Guys in Switzerland taking
pictures and chunking them into pieces you can render in braille or feedback
mouse.  You can start w/high res andd rill down into.

JR Edge recognition stuff.

CMN edge recognition and finding, layered photographs. The rest of something
is invisible but expected to be there.  artificial vision.  e.g. rest of
person under table not gone.

JR Those techs we probably won't see n authoring tools any time soon.

LGR A few words about makeaccessible, since we cna't see it.  It will
generate tags tree. How good is the tags tree.

HKS I'll download it tonight for tomorrow.

==============
lunch. JT/JR leave
==============

AERT issues for AU:

1. The tool should look at the longdesc.
WC Tools should look at the file and ask the user if it makes sense.
  Resolved

2. Elements that might have audio
 embed/object should be checked for audio, and if so is there a text
equivalent?
  resolved

3. Checks for noscript
 script equivalents may be provided by links, or other HTML. This should be
checked by a user..
  resolved

4. allow user to change units of measure.
WC there is style attached to elements, and it needs to be possible to
transform them, even if it comes from a linked style sheet. @@chaals
technique for how to do this.
 resolved.
GB this becomes a P2?
HS for an authoring tool it should be ahigher priority to get things done
right the first time.
WC there are cases where you want absolute. e.g. sizing raster images. In
WCAG 2.0 think about what needs to be changed - padding shouldn't matter, but
text should be variable.
CMN We need to ask the WCAG group to publish an erratum to this effect.
WCAG 1.0 checkpoint 3.3 is noted as an issue where we would like resolution
according to proposal for WCAG 2.
CMN Getting back to heather's point,  the two are seperate, and kind of do
what you want.
HS There is a lot left up to the reader in the way that we lay off stuff as
relative priorities. In this case (CSS values) there are the templates,
canned stuff, and generated content that needs to be picked up. There are
other things at the same priority for the author to do it.
CMN Looks again and realises that he was going further than there is in the
document.
HS It is not explicitly clear that it is more important to get the author to
do it right the first time.
WC importing stuff and evaluating is a part of that.
CMN It is possible to make a tool that lets people do stuff badly, and then
asks them to fix  it afterwards, because that was what we were asked for in
getting to REC. We allow both. So question is whether it should be in
checkpoints or in techniques and introductory material.
HS Would prefer it in checkpoints. Makes it easy to sell to developers, too.
WC You think of editing and repairing is seperate?
HS Often, although we have a model like spellchecking where it happens all
the time and the user can know what is going on without haviong to do it and
one point or another.
WC I we modularise ATAG into creation, repair, checking, that ould be helpful
for developers.
GB What is wrong with ATAG suggesting this kind of interaction but not
requiring it? This allows people to implement a piece of it or all of it.
WC They are aimed at tools which it assumes will be the whole process.
CMN I don't think we need to chop this into pieces. Any set of tools can be
tested together, so developers can work on a piece rather than the whole.
HS FrontPage doesn't do it's own accessibility checking. I think that the
worry with EARL is that developers will see that as a solution that they can
rely on someone else doing and stop working on accessibility.
CMN EARL is a way of rcording results, not a panacea, If your tool doesn;'t
read and interact with EARL, it's a nightmare to make it work.
HS Right now we say authoring tools just need to read in EARL
CMN Not limited to just reading in EARL. EARL is an interchange language that
lets tools interact, but there are other formats that could be used
CMN: EARL is to get in evaluation results. Tools should also produce EARL, so
the evaluation tool can know what has already been done
WC: e.g. to record that the author has already been asked whether this item
needs longdesc
GB: does EARL include a pointer?
WC: primarilty using XPointer, but may need some way to include inline
GB: we need some way to refer to a portion of an XML page; find EARL of
limited usefulness
WC: if authoring tools always provided ID, this woulnd't be a problem. But
authoring tools don't want to require this
CMN: ids are easier to track than xpointers
CMN: Heather suggested that higher priority of getting it right the first
time should be reflected in checkpoints. Heather has an action item
to review the Wombat draft and investigate how this could be codified in
checkpoints.
GB: what is the rationale for associating priorities to guidelines? please
review
CMN: goals are that authoring tools be accessible, ordinary people using ATAG
generate accessible output. Essential, important, and beneficial to
these goals map to priorities
WC ABBR values should be shorter than 15 characters (issue). ABBR should be
prounceable.
CMN: this should be a WCAG decision.
Action for Wendy: move this to WCAG list
WC: testing embedded technologies
CMN: yet, we should take this on
WC: server-side to client-side image map conversion
CMN: we should include this technique
WC: provide general info about a site (related to 13.3 in WCAG); can't
automate evaluatino but some repairs could be automated
CMN: yes, we should include this technique; e.g. creating a table of contents
for consolidated content
WC: checking for quoted text - how can these be identified? e.g. lots of
italicized text. Identifying common tricks for identifying
quotes
CMN: can we find someone who can explain this idea? these don't seem like
good suggestions?
WC: we want someway to check this, but we think some research should be done
CMN: we can mark this as an open issue, but we don't know how to, either
WC: identifying OBJECTs
Issue 51, Action for Charles - make a proposal
WC: checking HTML elements for meta information. SHould there be a separate
one for RDF?
WC: site vs page -
CMN: depends on the scope of the tool.
WC: problem is within WCAG - some checkpoints are page based and some are
site based
CMN: yes there are two kinds of checking
Action for Wendy - close all the ATAG items except the one that goes to WCAG
=====================
Meeting adjourned.

Chaals and Giorgio are going to work on some techniques for evaluating
content, Wendy, Katie and Loretta are going to work on PDF.

Thanks to Heather for organising and reorganising everything...

Back again tomorrow.

-- 
Charles McCathieNevile    http://www.w3.org/People/Charles  phone: +61 409 134 136
W3C Web Accessibility Initiative     http://www.w3.org/WAI    fax: +1 617 258 5999
Location: 21 Mitchell street FOOTSCRAY Vic 3011, Australia
(or W3C INRIA, Route des Lucioles, BP 93, 06902 Sophia Antipolis Cedex, France)

Received on Wednesday, 12 September 2001 20:20:31 UTC