Meeting minutes
New PR Triage
<github-bot> I can't comment on that because it doesn't look like a github issue to me.
spectranaut_: Have some editorial issues
pkra: This first one is good if someone wants to look through the spec.
pkra: See if sufficient or not. Old issues.
pkra: If somebody can do a longer review that would be helpful.
pkra: Tried to link it in role, state, prop definitions for the first time they appear.
spectranaut_: ANyone interested? Good way to get exposed to spec.
HaTheo: I'll take
front-endian-jane: I'll also take
spectranaut_: The other is role info.
pkra: Just waiting for final approval.
WPT Open PRs
<github-bot> I can't comment on that because it doesn't look like a github issue to me.
spectranaut_: Rego is working on AAM, I'll review those.
jcraig: First two are for you, and fourth one as well.
jcraig: Third one is for Adam. He and I spoke on Tuesday, so on me now.
jcraig: Fifth one is a recent update, wanted to talk with Giacomo if there's a block.
jcraig: Is this a spurious failure?
giacomo-petri: I'll double check the linter error.
jcraig: If there's still a spurious failure after that, you can close and then re-open an issue to kick the CI Bots to re-run the tests.
HaTheo: In the GitHub app itself, you can individually re-run things.
TPAC planning
<github-bot> I can't comment on that because it doesn't look like a github issue to me.
spectranaut_: Reminder -- less than two months to TPAC!
spectranaut_: If you have a topic to discuss, put it here so we can work on the agenda.
Status Update: aria-actions: handling focus when actions are synthetically triggered
spectranaut_: Was supposed to be a status update.
<spectranaut_> w3c/
spectranaut_: I don't know if there's much to say without Sara.
spectranaut_: Jamie Teh reviewed 12 hours ago, so hopefully that will move along.
jcraig: I have not looked since his review, will look now.
pkra: Sarah responded as well.
Matt_King: I will take a look today.
AT support for meter
pkra: I put agenda on this to revisit.
pkra: Matt, you had shared test results for ARIA min/max
pkra: Adrian re-tested and it matched Steve's, Tetra results.
pkra: It seems meter is not behaving.
pkra: Maybe a TPAC conversation.
Matt_King: WHy focusing on meter, versus everything else using min/max.
pkra: Came up with people not using HTML because of support.
pkra: Hence wanting to retest, set up for f2f
Matt_King: I was suggesting we reframe it around properties in general instead of meter in particular.
Matt_King: There are other elements that have min/max.
pkra: Makes sense, not beholden to meter.
spectranaut_: Making an F2F candidate for larger attribute support.
pkra: It's really about the element with slider patterns.
pkra: The ARIA pattern appears to work at least for slider.
Matt_King: ARIA AT shows inconsistent support for min/max. NVDA has little, VO has some, JAWS consistent.
jcraig: Adjacent question about ARIA-AT.
jcraig: Maybe that session at TPAC should be about sharing more of what hidden interface of what ARIA-AT looks like.
jcraig: I had suggested this stuff be visible to user as well.
jcraig: Owing to challenges with reports.
jcraig: Would help others connect dots, gather new contributors.
jcraig: Other confusing part is main page says VO has 33% of value-min.
jcraig: But when I look at admin interface, there's nothing about value-min needing review.
jcraig: What test expectations are there.
jcraig: I see claims that VO doesn't work, but that may not be how VO works.
jcraig: Example with unpressed button.
jcraig: Tests feel written for JAWS, applied universally.
jcraig: Suggesting to pull curtain back, show WG how it works, what's exposed.
jcraig: I haven't been diligent about ARIA-AT reviews, I know.
Matt_King: Happy to spend more time. I want to be careful.
Matt_King: There isn't any info you can't see. It's mostly to approve, move things on to next stage.
Matt_King: I think we need to talk more about rationale for expectations.
Matt_King: Based on CG input, the expectations are about interoperability, not specific SR expectations.
jcraig: There's a Candidate Review section visible only when logged in.
Matt_King: Yes, that's for SR users to give their stamp. All other info available through reports, GitHub pages.
jcraig: We may need a full hour working session.
jcraig: To see GitHub issues linked to results.
jcraig: To identify what needs to change -- test, spec, etc.
siri: So the ARIA value is not in HTML or the meter role?
siri: I remember working with progress bar and it worked fine with NVDA.
siri: I used meter as well and worked fine.
siri: Do we need aria-pressed on sliders?
jcraig: I was using that as an exmple for ARIA-AT to mark failures that weren't failures.
pkra: THe issue is about support for the HTML element versus the mappings, and how it's exposed in AT and the accTree.
spectranaut_: Please fill out the larger topic for TPAC.
"accessibility descendant" in presentational role inheritance and other concerns
pkra: This is the more fun one!
POssibly.
pkra: I saw role presentation inheritance for invalid HTML in a GitHub convo.
pkra: And now that I've looked further, the language in that part of the spec is outdated.
pkra: It doesn't make a lot of sense to me anymore. Some sentences seem over-edited.
pkra: It discusses host language elements are only complete when descendants are provided.
pkra: Examples given are lists, tables.
pkra: Since those aren't required in HTML or even ARIA, it feels weird.
pkra: The language doesn't match the reality.
pkra: The best reference might be looking at the content model, which doesn't require this.
pkra: But then it gets tricky with fluid content models.
pkra: Such as figure element, which can have figcaption as first or last. And then flow content.
pkra: Should the caption be role none? What about the image? No browser does this.
pkra: There seemed to be something more complicated. I looked at summary / details.
pkra: Summary doesn't become presentation when details is set to role none.
pkra: Maybe we say for these elements that kind of inheritance happens.
pkra: In the world of SVG, it makes even less sense.
pkra: So maybe this whole concept isn't working anymore.
Matt_King: I think you're on to something.
Matt_King: When you were talking about spec is wrong where element is complete, and also content model...
Matt_King: It seems we're trying to talk about something not really expressed in the specs.
Matt_King: Because they're written as technical explanation of how browsers parse code.
Matt_King: But we're trying to talk about semantics as rendered to end users.
Matt_King: A table without cells is meaningless to an AT user. Allowed, but semantically meaningless.
Matt_King: Or confusing.
Matt_King: But specs don't define these semantic dependencies.
<Zakim> jcraig, you wanted to say this predates not just the child/descendant terminology change, but also the rejection of "orphaned" role mappings, which is probably also relevant here.
jcraig: Thanks for digging into this, Peter.
jcraig: You mentioned term change from "child" to "descendant".
jcraig: We also had orphan roles.
jcraig: It's possible we might need to remove language.
jcraig: Jamie Teh and I were looking at orphaned role changes in specs.
jcraig: Saw no good user-driven reason to enforce mapping those to none roles.
jcraig: Adding them slowed things, so we removed those requirements.
jcraig: More digging needed.
spectranaut_: Interesting, James. So whatever you do with orphan roles might resolve this?
jcraig: Not that direct, but it might be the inverse.
spectranaut_: Yeah, once you add ARIA none some things become orphan roles.
jcraig: But potential user harm there. Like putting form tags outside cell in table layouts of yore.
jcraig: Which could be harmful to users.
jcraig: But I don't see value in forcing browsers to remove role semantics that do no harm.
spectranaut_: We have special roles in ARIA, and when they overwrite native semantics is "It Depends."
spectranaut_: This is another scenario where some things in one sub-tree get role none but sometimes hinges on other things.
spectranaut_: Wary of middle ground Peter suggested, but James comment maybe fits.
spectranaut_: Maybe we defer to host languages to decide which children get invalidated.
pkra: For Matt, there might be an argument for an empty table or list while they are getting fetched / populated.
pkra: For James, great point, but it contravenes putting into HTML AAM land.
pkra: For something like tables I think it's more complicated. And a dev can slap a table role to satisfy a layout or test.
pkra: If this is just a way to repair things, that would be fine too.
Stefan: Should we prepare ARIA for orphan cases?
Stefan: If people put role none on figure, what should happen with children?
Stefan: In early days of ARIA we discussed having author errors.
Stefan: Maybe we should think about that again.
Stefan: If you don't, you open justifications for errors.
front-endian-jane: Is there a good place to read up on past work with orphan roles?
front-endian-jane: Are there situations where this can cause weirdness with certain nesting?
front-endian-jane: Like nested lists with role none on a ul inside another ul, where they get promoted in the list depth.
front-endian-jane: How would that behave?
<jcraig> Reply to front-endian-jane's question about where the Orphaned Role discussion happened, it was mostly in GitHub: WPT Interop Accessibllity, ARIA, and maybe the AAMs... I'll try to find one or two.
Matt_King: I thought this started as a discussion on how to make spec say what it means.
Matt_King: Then turned into conversation about removing or reducing browser requirements.
Matt_King: Feels like two quite different discussions.
Matt_King: We're in this middle ground where role none works well with inheritance.
Matt_King: So we need to be careful we don't break those.
Matt_King: I think we should be careful about merging ideas that spec is clear versus what it means.
pkra: Matt, there were two stages for me.
pkra: First is that lang doesn't make a lot of sense.
pkra: Second is that role presentation inheritance doesn't work broadly.
<front-endian-jane> Thank you @jcraig! Some links would be greatly appreciated.
pkra: Then the realization that it is good it doesn't work for all of them.
Steven_lambert: Axe-core would love a definition of when things inherit presentation.
<Zakim> jcraig, you wanted to reply to pkra's comment "we need to keep what's being done" but there was also acknowledgement that the implementations aren't doing anything with this... if the req is both invalid and ignored by implementations, why not just remove it?
Steven_lambert: We've had to do a lot of guesswork.
<jcraig> web-platform-tests/
<jcraig> web-platform-tests/
jcraig: Peter mentioned we need to keep what's being done but also broader acknowledgement implementations aren't doing anything with this requirement.
jcraig: Because it would break web.
jcraig: Original intention was vague.
jcraig: Possibility we could fix this, do what we mean.
jcraig: But hard to figure out what we mean here is there's no user harm.
<front-endian-jane> +1 to pulling it out for now. That would make things a lot clearer for site author's looking at the spec too.
jcraig: Easiest short-term fix is to pull out invalid / ignored requirement, put it back in later if it was needed.
spectranaut_: C P K M P M K
spectranaut_: If there really are just some HTML elements whose specific children inherit none, then listing that is fine.
spectranaut_: Do we know that handful?
spectranaut_: I felt you knew what they were.
Matt_King: To parrot Peter at the beginning, we know the ones where this works and provides benefit.
Matt_King: I think that list is short? Maybe?
Matt_King: I don't know how complete is WPT in this space.
Matt_King: That might be best place to start.
jcraig: That's part of what the orphaned role tests were.
jcraig: Again, inverse.
jcraig: We could generate some of these tests to figure some out.
jcraig: Maybe a blanket case, or specific roles.
spectranaut_: I was imagining different tests. Not random roles and other roles.
spectranaut_: I want to know what people think is supported now.
spectranaut_: If you add none to a table, it eliminates the children.
jcraig: We have tests for that.
Matt_King: Do they account for nested tables.
jcraig: We have that.
jcraig: I thought you meant odd phrasing of all descendants. So even child contents of table with none. Or from invalid nesting.
spectranaut_: I want to know if there are ideas for next steps, PEter.
<jcraig> I will try to post a link to some of those table and list tests in WPT...
pkra: I sense we want to talk about this. Happy to draft a PR.
pkra: To remove inheritance, point to host language options.
pkra: Clear we need to talk more about this.
pkra: Figure out which elements trigger inheritance.
pkra: Maybe I got lucky by picking elements that behave differently.
spectranaut_: I think test writing is next. Who's interested in helping?
jcraig: Happy to take time separately to walk whomever how this works.
jcraig: Especially with Valerie's AAM stuff or additional computer properties.
<jcraig> Here's one WPT test called "el-li-orphaned-within-list-role-none" in https://
<jcraig> I didn't immediately see a <table role=none> test, but that might be in there as well. Happy to dig further, and coach anyone on the WPT PR, if there is momentum to write more tests.