338337 - Update DHTML Accessibility roles

Status: RESOLVED FIXED

(Core :: Disability Access APIs, defect)

Description

[Mark Pilgrim (inactive)]

New draft of DHTML Accessibility role taxonomy.  Need to check for changes and update gWAIRoleMap as needed.

http://lxr.mozilla.org/seamonkey/source/accessible/src/base/nsAccessible.cpp#1640

Comment 1

[Aaron Leventhal]

New properties as well. We have to decide whether it's worth updating the Firefox support until we have something approaching a stable spec.

Comment 2

[Aaron Leventhal]

States/properties as well

Comment 3

[Aaron Leventhal]

Need to add option, select as roles.

Comment 4

[Aaron Leventhal]

Attachment: Update to be in line with upcoming working draft

Comment 5

[Aaron Leventhal]

Attachment: Update both roles and relations, and fix some issues with relations

Comment 6

[Aaron Leventhal]

Attachment: File with describedby, labelledby, controls and flowto relations

Comment 7

[Mike Connor [:mconnor]]

given the question in comment 1, and the lack of clarity as to whether this is a stable spec, not going to block on this.

Comment 8

[Aaron Leventhal]

It's worth harmonizing with the current spec because it does not break backwards compatibility and will allow authors to experiment with the new capabilities. Especially important for AJAX apps.

Comment 9

[Aaron Leventhal]

Comment on attachment 229980 [details] [diff] [review]
Update both roles and relations, and fix some issues with relations

I'll add a comment that explains that the max depth is an optimization for labels and desriptions, which we believe will be relatively nearby their controls in the DOM.

Comment 10

[Aaron Leventhal]

Attachment: 1) Null check before CallQI, 2) remove magic number of 5000

Comment 11

[Darin Fisher]

This patch seems non-trivial, so I worry about regressions at this stage of the release.  I'm concerned about the removal of some of the atoms.  Won't that impact people who are already depending on those atoms?  Or, were those atoms never implemented?  Can you give us more to go on here, so we understand the risk / reward trade-offs.  Thanks!

Comment 12

[Aaron Leventhal]

(In reply to comment #11)
> This patch seems non-trivial, so I worry about regressions at this stage of 
> the release.
> I'm concerned about the removal of some of the atoms.  Won't that
> impact people who are already depending on those atoms?  Or, were those atoms
> never implemented?

The new roles and relations we are adding certainly won't cause regressions. 
The only thing removed is "icon", which is not part of the evolving standard at all, and never was. Not sure how it got in there. In any case, no screen readers or assistive tech actually make use of "icon".

The only other changes are for description for, label for and labelled by. There were some obvious mistakes in that code. I was careful to test those changes.

For the most part, this code is just 1) additions to the role table, 2) fixing errors in the above mention relations, 3) moving the reverse relation finding intelligence for reuse purposes, 4) using that code for new relations

The benefit of those patches is that we will correct and fill out our support for roles and relations, and this is a critical thing to do as we try to get industry adoption for DHTML accessibility. The controlledby/controllerfor relations are especially important for AJAX applications, which is really a new area we're looking at. A live region's content can be controlled by a widget. Thus, when the widget has focus and the liveregion changes, a screen reader knows that the changes are related to that widget.

Flowsto and flowsfrom are not necessarily crucial at this point, but the way I implemented the patch it made sense to put them in at the same time.

Comment 13

[Mike Beltzner [:beltzner, not reading bugmail]]

Comment on attachment 229996 [details] [diff] [review]
1) Null check before CallQI, 2) remove magic number of 5000

a=drivers. Please go ahead and land this on the branch.