Meeting minutes
<jamesn> - [Registration Open](https://
<jamesn> - [Ideas for Joint meetings](https://
<jamesn> - This is blocking re-entering CR: need answers for Anne's comment: "Was an effort made to get implementations to change here? Are there tests? This change was made only a year ago too. Furthermore, it's not clear to me that the current text works as reflecting only supports nullable strings if the attribute is an enumerated attribute and meets the requirements thereof. See also the discussion in the issue @saschanaz referenced."
[New Issue Triage](https://bit.ly/3u3z9Xq)
jnurthen: two issues. First "fix hidden text". Filed it because it doesn't seem correct.
… child of visibility hidden can be visible.
… visibility:hidden is still rendered in HTML sense.
… probably dependent on accname resolution for hidden content.
<spectranaut> present_
jcraig: seems reasonable, just like the related accname change
jnurthen: add to 1.3
jnurthen: next issue: role for carousel.
… we pointed them to openUI.
… seems good?
matt_king: rather not have a carousel role.
jnurthen: fine but if open UI comes back we can consider it.
jcraig: there's an openui piece that's more generic.
matt_king: APG has some examples as well.
jcraig: +1 for closing. I'll add a comment
cyns: we should encourage openUI to consider it.
… we've seen too many bad carousel.
matt_king: right, not necessarily a bad concept.
jnurthen: would need answer: 'what is a carousel??'
… that's why we shouldn't add a role now because we don't know what it would mean.
matt_king: +100
cyns: +1. somebody commented on outdated APG.
matt_king: but we just did.
jnurthen: the link goes to an ancient version. we need to publish the latest
… jon already pointed to editor's draft.
bryan: people mix the terms carousel and slider.
matt_king: slider is some kind of input.
bryan: just the concept.
… not the role.
jnurthen: that's where openUI does a great job at collecting real world examples.
[New PR Triage](https://bit.ly/3EMoKEo)
jnurthen: 1 PR. I agree with scott that APG should update wording
… @matt could it be?
… APG had changed "item" to "data"
matt_king: sounds reasonable.
jnurthen: curious if we should figure out why the wording diverged.
matt_king: I don't know the origin.
jnurthen: ok, will file an issue.
matt_king: might have been surrounding APG wording, trying to simplify language is a focus.
jnurthen: ok, maybe we should drop "column or row" entirely?
matt_king: is it allowed?
jnurthen: yes. but I've never seen it used.
cyns: I can think of some areas but not very common.
jnurthen: maybe just "items can be sorted in ascending order"?
cyns: sounds a bit circular.
matt_king: probably doesn't matter as much. it's really the cells anyway.
jnurthen: but not all of the cell might be relevant for sorting
cyns: items seems better. data as plural tends to bugs people.
matt_king: let's raise an issue on APG.
jnurthen: and remove "column or row"?
… objections?
matt_king: cell still seems good for me.
jnurthen: but let's keep it editorial.
cyns: agree to remove "column or row"
jcraig: +1
jory: "items are sorted in ascending order"
jnurthen: yes.
[Deep Dive planning](https://bit.ly/aria-meaty-topic-candidates)
jnurthen: we don't have any. low attendance this week.
cyns: would like to talk about canvas
jnurthen: sounds good. anything specific?
cyns: how to handle word processor, shape, bounding rect on canvas
… maybe a TPAC item.
jnurthen: maybe beginning Nov.
… tpac in oct
siri: modal vs non-modal. still doing research, there's scott's PR
… tab trap etc.
matt_king: we had a deep dive about default value for attribute
siri: keyboard and screenreader in virtual mode.
jnurthen: virtual mode, you'd want to allow escape
matt_king: discussion could be a deep dive but screenreader behavior is outside of ARIA, maybe ARIA-AT can help
… we don't have a non-modal in APG yet.
siri: what about tab trap?
matt_king: that's could be a deep dive.
… almost more an APG issue
jnurthen: if you're trapping, it's a modal?
… maybe TPAC discussion?
… could you put something together?
matt_king: we have stuff written, question is if people agree.
jnurthen: feels almost not an ARIA issue.
matt_king: no industry agreement on the definition.
… we talk about managed focus in the spec though.
jnurthen: everyone uses non-modal dialogs in many different ways, trapping depends on use case.
matt_king: but if it's unclear, then the role loses usefulness
… there's never been a clear definition of expectation
jnurthen: don't see not a lot of difference from landmark, visually speaking
matt_king: Aaron's example of comments, people don't want to fall outside.
pkra: focusgroup might be interesting for a deep dive.
jnurthen: sounds good. maybe tpac
TPAC 2021
jnurthen: registration is open
<jamesn> https://
jnurthen: please register. no cost.
… we need to decide if we want to have break out meetings or joint meetings.
… please add to the issue
cyns: canvas.
… where would that go?
jnurthen: I understood you to be more generic rather than the actual element?
cyns: true. webgl, relationship to SVG. We have old discussion from 10 years ago. Lots of use cases, maybe solved 2/10
… what other WGs should we talk to?
jnurthen: good question. Let's add it to the issue.
<IrfanA> https://
irfanA: spoken-html spec. we have identified the gaps, want to move towards exposing it.
… joanie had mentioned need for vendor inpu.
… not sure if we need a dedicated meeting.
jnurthen: sounds good. not everyone has to show up.
irfanA: will do.
[Updating ARIA 1.2 due to IDL implementations (exit and re-enter CR?)](https://github.com/w3c/aria/issues/1598)
<jamesn> "Was an effort made to get implementations to change here? Are there tests? This change was made only a year ago too.
<jamesn> Furthermore, it's not clear to me that the current text works as reflecting only supports nullable strings if the attribute is an enumerated attribute and meets the requirements thereof. See also the discussion in the issue @saschanaz referenced."
jnurthen: we need to answer it but I'm not sure.
cyns: I'm not understanding the concern
jcraig: but agreement that we should match browsers unless we are sure it's wrong.
… which is what we want.
… I will respond.
jnurthen: thank you.
… as soon as we have agreement, we can do a CfC for the new PR
… other than that, it's ready to go.
… cyns, jcraig could you respond and follow up?
jcraig: +1
cyns: +1
[When is hidden content taken into calculation of name and description?](https://github.com/w3c/accname/issues/57)
jnurthen: we have some comments in the issue and PR
<jamesn> https://
jnurthen: James Teh added a new comment recently.
… the change only happens if 'not rendered' in the HTML sense
… so visibility hidden goes the same route as before
cyns: core issue is that visibility hidden is rendered.
… is this the same problem for engines?
jcraig: not sure without checking
jnurthen: my problem with visibility hidden. if there's a visible child, how does that get parented in accessibility tree?
jcraig: that would have a render object. webkit would expose a generic group for any render object so that it can pass along the bounds etc.
aaron: in blink, the parent in the accessibility tree depends whether you're walking the full tree (DOM based) or unignored (filtered) tree.
… visibility hidden parent is not in the filtered tree, then the child's parent would be the parent's parent.
jcraig: this is more or less the same in webkit. sometimes we don't have DOM elements (e.g., generated content) but in the tree.
… but e.g. div scrollable you'd get a scroll view with a render object
aaron: right. blink tries to have 1-1 relationship between DOM and AX tree. But pseudo-elements, list markers don't.
… for those subtrees we walk the render tree.
jnurthen: how do we make progress?
… someone suggested collecting examples and test for consistency.
aaron: we're happy to try the rule with collapse, we can still have them in the full accessibility tree.
jcraig: that's what we were discussing yesterday. Concern was that it might break something.
… I summarized my answer on github.
jnurthen: I would suggest to get many (simple) code examples, work through them, identify if the result is good.
jcraig: coming back to my point. aaron was concerned with a specific example. I suggested google might be able to say if that's a prevalent problem.
aaron: does webkit support *-by to visibility hidden object?
jcraig: direct pointing or to parent?
aaron: direct.
… prevalent might not be that great, if it used to work and goes without warning.
jnurthen: but with "rendered" then this wouldn't apply.
jcraig: right.
cyns: then we can use APG to inform authors that they can go different paths.
jnurthen: and display none is more performant
… but examples are key
aaron: but we need to be very clear about "render"
cyns: and make the examples test cases.
jcraig: are we suggesting this is part of the PR in question?
… waiting for lots of examples might take a while.
jnurthen: some examples at least so that readers know that it works differently.
cyns: to show the main differences.
aaron: from CSS, display:none is the only "not rendered"?
cyns: hidden example, too.
… visibility, offscreen and probably enough
jcraig: one for rendered as people expect, one for none, one for offscreen, one for visibility hidden
cyns: sounds good.
pkra: pointing to script tag
<jamesn> "An element is being rendered if it has any associated CSS layout boxes, SVG layout boxes, or some equivalent in other styling languages."
jcraig: that's display:none. style might be interesting.