Editorial?: Fix for #759 #885
Conversation
There was a problem hiding this comment.
While I don't think that it's necessary to specify it includes the inherited role of presentation, I'm ok with that and believe that is indeed an editorial change.
I'm afraid I'm not cool with the "use its own heuristics" bit, however. We've already dealt with how presentational descendants which remain in the accessibility tree should be mapped in the Core-AAM. See the mappings for presentation and none. As you'll see in those mappings, the Core-AAM specifies what each platform expects. And, yes, those mappings are generic in nature, but they are specific in their expectations for each platform.
That said, arguably it's not sufficient to just remove your proposed change regarding heuristics, because the ARIA spec isn't right in its current form: The reason we made the aforementioned change to the Core AAM is because we didn't want what's currently described in the ARIA spec ("exposes the element according to its implicit native semantics") to happen. At least not in the case of required children. An accessible list item that is not inside an accessible list makes no sense. Ditto for table rows and table cells that are not in tables, which seems to be what is being reported in the original bug.
The current language states:
the user agent ignores the
presentationrole and exposes the element according to its implicit native semantics.
What about replacing that with something like:
the user agent exposes the element as described in the accessibility API
mappings for presentational objects which remain in the accessibility tree.
?
Preview | Diff