1501273 - Expose an html:form element or an element with role "form" as a landmark role if it has an accessible name

Status: RESOLVED FIXED

(Core :: Disability Access APIs, enhancement, P2)

Description

[Marco Zehe (:MarcoZ)]

This complements bug 1442196, which introduced the XML role mapping for the form element. This bug is about the exposure of the platform APIs LANDMARK role. Also, if it is a non-form element but with a WAI-ARIA role of "form", the landmark role will now be exposed.

Comment 1

[Marco Zehe (:MarcoZ)]

Attachment: Bug 1501273 - Expose an html:form element as a landmark role if it has an accessible name, r=jamie

If html:form has a name given via ARIA, expose it as a landmark to platform APIS.

Likewise, if an element gets an ARIA form role and a name, expose it as a landmark accessible.

Comment 2

[James Teh [:Jamie]]

Can you point to where you determined this to be the correct mapping? As far as I can see, this is not correct according to the Core AAM (either 1.1 or 1.2):
https://www.w3.org/TR/wai-aria-1.2/#form
The Core AAM says IA2_ROLE_FORM but ATK ROLE_FORM. I think the reasoning for this difference is that there was contention (from Alex and me) as to whether the landmark role really should override the semantics in this case. For ATK, that's necessary in order to find landmarks, but this isn't necessary in Windows. It's also a backwards compatibility problem. See this IA2 thread:
https://lists.w3.org/Archives/Public/public-aria/2016Aug/0233.html
Interestingly, the Core AAM also doesn't say anything about differentiating the role depending on whether there is a label. The HTML AAM does for <form>, but not the Core AAM for role form.

Of course, it's entirely possible I missed something. Happy to be corrected if so.

Comment 3

[Marco Zehe (:MarcoZ)]

You're right, I was over-enthusiastic in making sure both implementations produce the same result. So the question is: Should a form with an accessible name still get a form role in IA2, but a Landmark role in ATK, then, in other words, keep existing behavior in IA2 regardless of whether there's a name or not? The statement to map to the WAI-ARIA landmark role behavior in the HTML AM spec seems to suggest that.

Comment 4

[alexander :surkov (:asurkov)]

Besides I don't like we compute name for each role request from perf point of view, it also might be a problem that role may be changing during accessible object life cycle. I realize the idea is to filter out irrelevant form elements, but sort of concerned about it.

(In reply to James Teh [:Jamie] from comment #2)
> The Core AAM says IA2_ROLE_FORM but ATK ROLE_FORM. I think the reasoning for
> this difference is that there was contention (from Alex and me) as to
> whether the landmark role really should override the semantics in this case.
> For ATK, that's necessary in order to find landmarks, but this isn't
> necessary in Windows. It's also a backwards compatibility problem. 

I second it. HTML-AAM says either form role (name case) or grouping role (no name case), not a landmark/form role.

Btw, it feels weird about linux requirements, because they want to use less specific role than ATK provides, but it is how it is.

Comment 5

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/2a050ad70afb
Expose an html:form element as a landmark role if it has an accessible name, r=Jamie

Comment 6

[Andrei Ciure[:aciure]]

https://hg.mozilla.org/mozilla-central/rev/2a050ad70afb