[Accessibility] FSGA and AT-SPI meetings minutes for 3/29/06 and 3/31/06

Catherine Laws claws at us.ibm.com
Wed Apr 5 15:24:43 PDT 2006

In attendance on 3/29:
Janina Sajka (FSG)
Sun: Bill Haneman, Will Walker, Peter Korn
IBM: George Kraft, Cathy Laws, Larry Weiss, Aaron Leventhal, Pete Brunet,
Rich Schwerdtfeger
KDE: Gunnar and Olaf Schmidt
W3C: Al Gilman
Frank Carmicle
Ubuntu: Henrik Nilsen Omma

If interested, please register for FSGA membership to be listed.

Opening remarks by Bill about goals for exposing HTML/DOM document:
1. Expose document content that is structure
2. Serve the needs of the AT and users
3. Not a big burden on ATK provider. Requires minimum amount of DOM
remodeling before passing to ATK
4. Needs to be backwards binary compatible with existing apps and ATs. We
don't want it to break. Would have to have redundant APIs, but not ideal.
Also, don't want to add too much to ATK or have to wait for new ATK.
5. Time to market. Adding attributes to ATK objects also affects Java
accessibility APIs (Peter K.). But Bill thinks it could be shoehorned in
because Java supports rich bag of attributes. Time to market for Mozilla is
more critical.

Bill: Text is treated differently in HTML docs. ATK has node base DOM,
nodes are atomic, with children, but HTML can be overlapped - interleaved
text content can be handled with AtkHypertext.

Objections raised:
1. Getting Hypertext link, activation, etc is too complex - you have to go
in through the link, get number of links, etc.
 2. Hyperlinks are not intended to be ATK objects, but were interfaces to
get info about objects.

Bill: Change to encourage Hyperlinks to be AtkObjects - if a hyperlink has
only 1 anchor, then no change. Propose make a getLink call for the simple

Also, make Hyperlinks children of Hypertext. Use Hypertext to express
interleave of any embedded objects in text. Children of complex text
object, like a paragraph, can tell you where they are ordered. So you can
string the text and order of the text. Good for Collection API for finding

Bill - two other proposals -
1) Make all content (text) in leaf nodes. Makes it harder for Collection
becuase there is the question of which container each leaf node is in.
Also, will end up with roughly twice as many nodes as in HTML DOM.
 2) Use containers  and use embedded Unicode character in text. Is
reasonable but you have to rely on semantics of child order and decide if
text lives in parent anchor or in child. Previously text was always in
parent container. Breaks existing implementations. Better to keep the text
all in Hypertext parent and use Hypertext interface to create children, and
Hyperlink children will be objects.  Would not cause problems for ATs.

Aaron: Compare ATK with text in containers versus Hyperlink children
Andres: XML DOM proposal is like ATK with text in leaves - no longer being

Comments on examples from this document:

1. Use object attribute to clarify heading levels. Maybe "heading level 1"
spelled out. Same concept for other HTML tags.
Child image  - URI is filename. Hyperlink shows where image is using offset
range method.
Hypertext interface for all cases where embedded objects are interleaved
with text.
Distinguish ATK_ROLE_LINK for all links. Use other roles for non-link
embedded objects.
Question: How to position caret at image - before or after?
2. Put new line \n in for <br>
3. Hyperlink is really just an interface of an object. Get object. A
Hyperlink object can also implement Hypertext.
4. All parents that have text content and embedded children implement
5. OK
6. All text should  be in parent. Child hyperlink shows range. Should child
Hyperlink dup AtkText implementation that is in parent? Seems more
convenient to dup the text.
Aaron: Does gok think something like an image of label is a link if
Hyperlink is implemented?
Bill: Real links implement Action and have role of ATK_ROLE_LINK.

Next meeting: Friday, March 31, at 10 AM CDT on regular FSG call number to
finish review.

Some notes from the Friday review:

The use of embedded Unicode characters in text to show children is
dependent on the user agent.
Discussion of links spanning paragraphs.
Should text attributes for links be duplicated in both parent and child?
If an image link has no text, should we still have a parent with no content
and a child with the image? Or does the AT prefer for it to be collapsed?
Will: Concern with image taking no space. Would prefer an embed character
to represent caret moving to image.
Lists: Omit the bullet characters in a list, but not numbers for numbered
lists. ODF, OpenOffice uses Unicode characters for bullets.

See the  "ATK using Hypertext and Hyperlink children" column for the
recommended ATK implementation notes for each HTML example in

Thanks, Bill, Aaron, and everyone for all of your hard work on these

Cathy Laws

Manager - IBM Software Group (SWG) Accessibility Architecture and
11501 Burnet Road,  Bldg 902 Office 5C016, Austin, Texas 78758
Phone: (512) 838-4595 or (512)838-2308, FAX: (512) 246-8502
E-mail: claws at us.ibm.com, Web: http://www.ibm.com/able

Vision without action is merely a dream, action without vision is merely
passing time, but action with vision can change the world.” Nelson Mandela

More information about the Accessibility mailing list