W3C

– DRAFT –
ARIA WG

30 April 2026

Attendees

Present
CurtBellew, cyns, Daniel, filippo-zorzi, katez, melsumner, np-at
Regrets
-
Chair
-
Scribe
Jacques

Meeting minutes

New PR Triage

<github-bot> I can't comment on that because it doesn't look like a github issue to me.

melsumner: We want to wait on the "add support for namerequired" issue
… this is related to the listbox, want to look at a few other approaches

New Issue Triage

<github-bot> I can't comment on that because it doesn't look like a github issue to me.

spectranaut_: Deep dive added to the menu element proposal

spectranaut_: Question from the web: "What does it mean when a role has no allowed accessibility child roles section"

jamesn: sounds like a good first issue, as long as we know the actual right answer.

spectranaut_: yes, peter answered it.

WPT Open PRs

<github-bot> I can't comment on that because it doesn't look like a github issue to me.

Deep Dive planning

<github-bot> I can't comment on that because it doesn't look like a github issue to me.

dgrogan: Chrome is working on these new attributes, and we have some idea on how to make this screen reader and keyboard friendly.
… hoping to have a discussion on our proposal, thought it would be appropriate for a deep dive

jamesn: sounds like a good idea, even if we had a week with a light agenda we could do it in a standard meeting.
… but a deep dive would be good.

spectranaut_: can we schedule this May?

dgrogan: Next week works for us.

dgrogan: I'd like Scott's input here, apple thoughts would be amazing.

CSS scroll-marker-group modes && WAI-ARIA patterns

spectranaut_: I was looking at this this morning and tabbing wasn't working as expected. I'm wondering if I am on the right version

spectranaut_: When I tab, it goes to the other focusable content in the tabs that shouldn't appear
… and the scroll marker doesn't update

@daniil: sounds like a bug

Daniel: I don't think this was happening to me, this was working right when I tested.

Daniel: this resembles the expected tab behavior, nothing to say at this point.
… looks good to me, it works.

<np-at> I'm getting the same behavior as valerie on macos with the latest canary build

jamesn: I am not sure if I am seeing the feature correctly, I see some bars appearing.

@daniil: that looks as expected, the bar is a scroll bar.
… this is the expected behavior, authors can hide these if they like

Daniel: Now noticing some differences between tab order and screen reader order.
… I would have expected that after the tabs I would go to the headings, but the tab order isn't matching the screen reader order.

Daniel: If you want to get to the tab panel, you have to arrow up, with is counter intuitive

jamesn: I have a lot of questions, but I assume a lot of them are answered in the spec
… such as name calculation for the markers, etc

@daniil: These calculations are documented in the spec link you can find from the explainer.

jamesn: In the tab mode, the scroll marker psuedo-element name calculation is what I am looking for.

@daniil: we only define the role for that.

jamesn: The browser seems to be using the name for either the heading or title of the surrounding div.

jamesn: This seems fine, but I want to understand where it is coming from.

@daniil: This would be folloing the normal rules.

jamesn: I don't think there is a defined rule for this, it needs to be defined.

jamesn: I'm seeing it in devtools.

jamesn: Its coming from the surrounding div title, so this needs to be defined, and shouldn't be entirely reliant on title.

@noah: it looks like this is coming from the surrounding content.

@noah: it picks it up from aria-label

jamesn: the name calculation needs to be defined.

jamesn: Another question: are you setting aria-selected on the selected tab or now? (or expanded)

@daniil: yes it is being set.

@noah: does aria-details use name as a fallback?

melsumner: this seems strange.

@noah: this might be a bug

spectranaut_: Do we need more time now or can we move on?

jamesn: I have other concerns, I have some weird behavior with find on page
… it switches the tab but the controls don't switch

@daniil: sounds like a bug.

@jamesn: I still have some concerns that people may use this wrong
… but that isn't the fault of the feature

@daniil: please review the spec when you can.

How should alt text be exposed?

spectranaut_: there is a mathml alttext that isn't well supported or defined.
… this is asking for a fallback for when mathml is not supported
… the question here is "are there cases where the label should complement rather than fall back"

jamesn: I think I aggree that the label or alt-text could be fallback content for when mathml is not supported.
… couldn't they use description already for when they want it to supplement?

jamesn: I don't care which are the fallback and which are the supplemental, as long as both exist.

np-at: I'm wondering if there is any danger or issue with aria-label's behavior being different here.

<melsumner> strong agree with Noah here

jamesn: That would mean that label should always be an addition.

<spectranaut_> Jacques: This is kind of tangental, one thing I've heard that is hard from the author, if there is a lot of math content, not using mathml, rendering it in html to make it pretty, but they want to use mathml instead, it is hard to provide a good experience for screenreaders that do support it and don't support it

<spectranaut_> Jacques: I don't think the alt text solves that, but it would be nice to have one

<spectranaut_> Jacques: if a screen reader doesn't support mathml at all, rendering in html is better. Possible better than one giant alt text for the whole blob

<spectranaut_> spectranaut_: like canvas fallback

<spectranaut_> melsumner:

<Zakim> melsumner, you wanted to ask if someone needs to use MathML and a SR, wouldn't they use a SR that supports MathML already?

<spectranaut_> Jacques: there is a narrator on windows doesn't support mathml

<spectranaut_> Jacques: there are microsoft product are trying to support other screen readers

<spectranaut_> melsumner: I strongly believe we should not use aria-label as a fall back

np-at: I think mathl content works as inline text, and inline text + aria label is already one of the exceptions
… or at least that is how some screen readers treat it.
… for the paragraph role or other inline content, it may ignore the label sometimes
… I'm not sure what the standard behavior is here.

Minutes manually created (not a transcript), formatted by scribe.perl version 248 (Mon Oct 27 20:04:16 2025 UTC).

Diagnostics

Maybe present: @daniil, @jamesn, @noah, dgrogan, jamesn, spectranaut_

All speakers: @daniil, @jamesn, @noah, Daniel, dgrogan, jamesn, melsumner, np-at, spectranaut_

Active on IRC: CurtBellew, Daniel, filippo-zorzi, Jacques, jamesn, katez, melsumner, np-at, spectranaut_