428954 - Wrong accessible name for iframe-based ARIA widgets

Status: VERIFIED FIXED

(Core :: Disability Access APIs, defect)

Description

[Aaron Leventhal]

When a Javascript custom widget is made from an iframe, it must be possible to define the accessible name (label) on the iframe itself, where the widget is instantiated.

For example, the following widget should have "List of colors:" as the label.
http://www.mozilla.org/access/dhtml/listbox-owner

Comment 1

[Aaron Leventhal]

Attachment: Remove code so ::GetName() is from base class. We don't need the child's name on the parent. The parent knows more about the reason the widget was instantiated and what the label should be.

Comment 2

[alexander :surkov (:asurkov)]

Aaron, could you provide an example?

Comment 3

[alexander :surkov (:asurkov)]

Should we try to get ARIA name firstly instead to drop the name getting from title and url?

Comment 4

[Aaron Leventhal]

Alexander, I'm not sure I understand your question.

The way this works is that the iframe's accessible label/name is used if it's a good one. Otherwise the body/docaccessible gets a chance.

Comment 5

[Aaron Leventhal]

I understand the question now.

Basically I think, no -- this patch is the right way to do this. The inner widget may wish to define a default accessible name via ARIA and I don't want the title or url of the inner document to block that.

Comment 6

[alexander :surkov (:asurkov)]

Ok, let me look in details:

nsOuterAcc::GetName

1) title from inner (nsDocAcc)
2) ARIA name
3) url from inner (nsDocAcc)

nsDocAcc::GetName

1) name from outer (nsOuterAcc)
2) ARIA name
3) title from self

nsOuterAcc:
You suggest to save 2) item of nsOuterAcc and to loose 1) and 3) and I still don't catch you why we don't need this any more (how can title or url block ARIA name of the inner?).

nsDocAcc:
Why nsDocAcc prefers to get name from outer rather than own ARIA name?

Do we want to try url if title (see 3d item) is empty?

Comment 7

[Aaron Leventhal]

> Why nsDocAcc prefers to get name from outer rather than own ARIA name?
Here's what happens. A developer creates a mashup widget in an iframe. Other developers use the iframe. The iframe widget author doesn't know what context it will be used for each time, but the outer document author does know. So the iframe element really represents the specific instance data. The specific instance is more specific and should be able to set the specific a11y properties of that usage.
See http://developer.mozilla.org/en/docs/ARIA_User_Agent_Implementors_Guide#11.3.16_Documents.2C_Handling_.3Cframe.3Es_and_.3Ciframe.3Es

These iframe-based widgets are becoming more common.

I don't want to use the full URL to repair a missing accessible name. It doesn't make much sense to hear that a listbox is called http://www.blahblah.org/widgets/listbox.htm -- we already know it's a listbox from the role or it's not accessible at all.

Does that make sense?

Comment 8

[alexander :surkov (:asurkov)]

Ok, since now inner gets name from outer and outer check url then should we add url checking to inner?

Comment 9

[Aaron Leventhal]

Maybe. However, the inner exposes the URL via the accessible value and the document interface so it would in fact be redundant.

Marco, if a doc has no title or other name, does the URL need to be exposed in accName?

Comment 10

[Marco Zehe (:MarcoZ)]

(In reply to comment #9)
> Marco, if a doc has no title or other name, does the URL need to be exposed in
> accName?

Well it should be present as a last resort if everything else fails, yes. Otherwise one would get empty-named frames, which is just as irritating.

Comment 11

[Aaron Leventhal]

Attachment: Allow URL as last resort

Comment 12

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

Comment on attachment 315762 [details] [diff] [review]
Allow URL as last resort

a1.9=beltzner

Comment 13

[Marco Zehe (:MarcoZ)]

Verified fixed using Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9pre) Gecko/2008041606 Minefield/3.0pre.

Comment 14

[Marco Zehe (:MarcoZ)]

Flagging for intestsuite since I'll be adding a test for an absent name in bug 441519.

Comment 15

[Marco Zehe (:MarcoZ)]

Test in fix for bug 441519, test_nsOuterDocAccessible.html.