Open Zhang-Junzhi opened 2 months ago
Nothing specifically for the popover attribute itself
popovertarget on a button should not have an aria-expanded declared with it.
Nothing specifically for the popover attribute itself
popovertarget on a button should not have an aria-expanded declared with it.
Might popovertarget
have a native semantics equivalent to aria-controls
?
popovertarget
forbid the attribute aria-controls
?aria-controls
(in case the button controlls other elements besides the popover target)?And there's also popovertargetaction
, which needs to be considered.
It doesn’t have an implicit aria-controls.
popovertargetaction doesn’t add any a11y semantics
It doesn’t have an implicit aria-controls.
I think maybe we cannot safely exclude (at least in any case) the possibility of an implicit aria-controls
from popovertarget
.
Since I found that the ARIA spec says the following about aria-expanded
:
If a grouping container that can be expanded or collapsed is not the accessibility child of the element that has the aria-expanded attribute, the author SHOULD identify the controlling relationship by referencing the container from the element that has aria-expanded with the aria-controls property.
That is to say, it might be proper to have an implicit aria-controls
(or an ID implicitly added to aria-controls
) if the popover
element is not the child of the element with popovertarget
.
Being that I helped spec what was to be implemented for popover and its related attributes, I'm very confident that aria-controls is not used, nor is it going to be implicitly used.
It'd be very rare to have a popover as a child of the popovertarget - since that attribute is only valid on buttons. In the few cases where that could happen, an aria-controls relationship would be useless (beyond the fact that aria-controls is already pretty useless outside of with comboboxes).
The current spec hasn't mentioned anything about
popover
, which is a new HTML feature. The relationship between native semantics of HTMLpopover
feature and ARIA might need to be normalized.