W3C

– DRAFT –
ARIA and Assistive Technologies Community Group

08 April 2026

Attendees

Present
Joe_Humbert, Matt_King, Siri
Regrets
-
Chair
Matt King
Scribe
cyns

Meeting minutes

Review agenda and next meeting dates

Next meeting will be on April 22 at the usual time

Current interop reporting status

mk: I already moved the plans in CR to google sheet about meetings with AT vendors

(https: //docs.google.com/spreadsheets/d/1pvnIOTpbR5QtFIW7IvXTpIJsCEgeSmhbcefVbg00bpM/edit?usp=sharing)

mk: 5 plans in draft review, 2 active

mk: want to get simple data grid into draft review, blocked by netlify issue (w3c/aria-at#1363))

Nicholas cook: I tried to look at the issue. Needs someone with access to Netlify

mk: will ask Daniel or others from w3c team to look into it

Celebrate Windows bot fix!

mk: windows bots have been down since December, Nicholas fixed, Chris merged, Matt tested. Thank you Nicholas!

Nicholas Cage: Thanks also to Westin from Assistive Labs

Execution of quantity spin button test plan

test plan * Current version: Quantity Spin Button Example Test Plan V25.12.16

mk: joe is signed up to test JAWS, need another JAWS tester

MK asks Hadi, coordinating schedules

Hadi: Can do testing, will not often be able to come to this meeting time

mk: Pls test with JAWS 2025 so it's the same as the bot run

mk: NVDA 2025.3 conflicting result between Joe and Dean.

Joe: did that on purpose so we could talk about it

Joe: When you run with NVDA, 2 assertions convey minium and maximum value, which it does, but only in the error message. If the control didn't have an error, the info wouldn't be conveyed. Is that a failure?

mk: if the error wasn't present, min and mix weren't conveyed. The error gets conveyed, but no the min and max.

mk: priority of htose assertions later in agenda

mk: The test fails those assertions.

Joe: I will respond to Dean that min/max are not conveyed as attributes on controls.

mk: this is why we have humans who know what they're doing

mk: VoiceOver conflicts. Dean and Murray have different output. Neither is present today

mk: we're down to 1 conflict

Joe: I also have info on that one

Joe: Dean asked me to run the conflict test (6) on older Mac and one with MacOS, Safari, VS are 26.3.1 When I used that combo, I got the same result as Murry

mk: Was quicknav off?

Joe: yes

mk: That's a big regression in VO!

Joe: Dean was still getting some of these issue with his version.

Joe: Dean was getting correct info, Murray and I were geting 0%

Joe: with ctrl-option-down-arrow some info was announced, not %

Joe: Dean may change his results

mk: should he?

Joe: Dean may have upgraded

Joe: Murray has test 6 not complete, so conflict doesn't show

mk: NVDA conflict page changed too?

Joe: Dean changed his results to FAIL

(debugging happens)

Joe: it's waiting on my submission to show conflicts

mk: that's a strange behavior of the tool

Joe: Murry needs to save his results to make the conflict show

mk: may be a bug that we need to show the it's not complete in the status table

mk: will try to repro in staging environment

mk: looks like we have a path forward. Hadi testing JAWS, Joe breaking ties

Assertion priority discussion for spinner test 4

<Matt_King> GITHUB: w3c/aria-at#1366

mk: Raised this issue when reviewing NVDA conflicts in quantitiy spin button plan.

mk: you have a spin button that allows input of invalid value by typing. When you do that, an error message appears that tells you how to correct the problem. In that circumstance, there are 3 assertions...

mk: 1 assertion is that screen reader will announce error message, 2 convey min value, 3 convey max value

mk: in the other tests in the test plan, conveying min/max is a P2/should. But in this circumstance, we might have made a mistake when setting the priority to 1

mk: seems problematic that screen-reader conveying both error message and min/max

mk: normally error message that says value is invalid should tell the user how it's invalid

mk: we should consider whether min/max should be SHOULD or MAY for screen reader

mk: NVDA does not convey min/max consistently in any tests

mk: JAWS does not convey min/max in that circumstance, but does in others

mk: will get input from sr developers, but does the group have an option

Hadi: maybe range could be in instructions instead of making min/max required. I like to hear it with the error message, but should it be required? Might get a complaint if it's in isntructions. I'm for MAY, but persnally I think it should be SHOULd

mk: question: example spin button, there isn't a requirement in WCAG or APT that you have to describe the range, because you've specified with aria-value-min and aria-value-max. expectation that screen reader will take advantage of those. But this one also has aria-invalid=true

mk: what does wcag say about case where aria-invalid is true and spin button doesn't have an error, would that be a WCAG failure

Joe: yes

3.3.1

Joe: "...the error is described in text"

mk: is it good enough if the error says "the value is out of range" would you fail it?

Joe: I would probably fail it for input eero suggestions

that is AA SHOULD

3.3.1 Error Identification and 3.3.3 Error Suggestion are the relevant WCAG requirements

<loren> I have to head out a little early, please let know when the engineer bot training sessions is.

mk: If you have a WCAG-compliant spin button. Our SR requirements always assume that the control complies with WCAG. would it be redundant to require SR to do that?

Joe: There's no WCAG requirement that the range be provided. As Hadi said, there are usability needs

mk: we can reduce priority of exposing min/max to 3 where there is an error message present and it's aria-invalid

mk: because the WCAG-compliant error message would give the suggestion

Joe: I think it should still be a SHOULD based on the being a usability need

mk: if the spinner and its error message are WCAG compliant, then you don't need to convey the min/max because the WCAG-compliant error message would say it

mk: We are at time. Please look at the other issues assynchronously

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

Diagnostics

Maybe present: (https, Hadi, Joe, mk

All speakers: (https, Hadi, Joe, mk

Active on IRC: cyns, Joe_Humbert, loren, Matt_King, Siri