W3C

– DRAFT –
ARIA and Assistive Technologies Community Group

01 July 2026

Attendees

Present
Hadi, Joe_Humbert, jugglinmike, Matt_King, mmoss, siri
Regrets
-
Chair
Matt King
Scribe
jugglinmike

Meeting minutes

Review agenda and next meeting dates

Matt_King: Requests for changes to agenda?

Matt_King: Hearing none, we'll stick with the agenda as planned

Matt_King: Next CG meeting: Wednesday July 15

Current interop reporting status

Matt_King: I'm now quite familiar with why we put off the rating radio group

Matt_King: So we're still at 3 at draft, 3 at candidate review, and 21 at recommended

Matt_King: We're in a pretty solid place for getting two of those plans into candidate review

Matt_King: I thought I'd have more time for this work, but it's not working out that way

Matt_King: I'm struggling to find the time to have talks with Apple for the purpose of advancement of test plans

Matt_King: This is one of the things that are stuck due to my capacity right now

Update on infra engineering training

Matt_King: We've completed all of our scheduled training

Matt_King: And we have a recording of that

Matt_King: However, that really only covered getting started

Matt_King: We have three starting issues that we're looking for people to sign up for

Matt_King: mmoss, do you have your environment fully set up now?

mmoss: On my personal system (not my work system). I'm still having a weird pathing issue on my work system, though.

mmoss: If you don't set up the VoiceOver driver correctly the first time, then you can get into a bad state from which it's difficult to recover

Matt_King: If you want to sign up to own the VoiceOver Bot upgrade issue, then there's a link in the agenda (issue number 1364) that has a link to each of the "upgrade" issues

w3c/aria-at#1364 (comment)

Matt_King: Once we have an owner for each of those issues, the plan is to schedule a pair-programming meeting with ChrisC and/or jugglinmike as a way to safely get your feet wet

mmoss: That makes sense.

mmoss: Right now, that issue seems to include only placeholder text

Matt_King: Right, we don't have very much detail right now

mmoss: I think we've reached at least version 26.5 by now

mmoss: I'm happy to take that on

Matt_King: Great! I'll start an e-mail thread with jugglinmike and chrisc to help get things scheduled. It will be an open meeting so that others can join if they wish, but we just need to find a time that works for the three of you

Draft review of minimal data grid test plan

Matt_King: So JAWS is 100% complete with 7 conflicts

Matt_King: NVDA is 79% complete with 6 conflicts

Matt_King: VoiceOver has 4 conflicts

Matt_King: We previously talked about VoiceOver

Matt_King: We agreed that we would pass the conflicts about discrepancies in how VoiceOver conveys the role. We intentionally leave room for differentiation there

Matt_King: However, we did identify a problem with the difference between "table" and "grid" since the latter is interactive, and users will have no way of recognizing the availability of that interactivity if the word "table" is used to describe the role

I described this in an issue, here: w3c/aria-at#1375

Matt_King: The strange thing is that we might have to call this a blocking issue for table (unless they want to change their grid support, like to call it "grid" or "interactive table" or whatever they'd like to do)

Matt_King: Before we talk about JAWS, does anyone have any feedback on that?

Joe_Humbert: I'm concerned about passing that, since it gives Apple cause to say, "well, we passed it, so why do we need to change it?"

Joe_Humbert: Tables are very different than interactive data grids

Joe_Humbert: I think conflating them has a very negative user impact

Matt_King: dean and Elizabeth performed the testing here. Dean marked it as passing, and Elizabeth marked it as failing

Matt_King: Together, we agreed to change Elizabeth's results

Matt_King: The point I raised was to force the conflict without failing. I tried to capture our logic in that issue--issue 1375

Matt_King: It would be helpful if you want to add a conflict to that issue stating your stance that a change here is critical

Matt_King: I'm hesitant to change all of those to "failing" because we have said that we aren't going to force screen readers to use specific words

Matt_King: I guess we can say, "you don't have to use a specific role, but you can't use 'table'"

Joe_Humbert: Right. You can't use the same role to represent two different things

Matt_King: You're making me question whether or not we should pass or fail those

Joe_Humbert: This goes back to some of the conflicts that I had with Dean

Joe_Humbert: I passed it because I knew he would fail it so that it would show up as a conflict and we could have this conversation

Matt_King: I don't think the behavior is okay; the question is at which point in the process we highlight the concern

Joe_Humbert: I would vote for failing it. If we say they pass but we want them to change it, that hurts our argument

Joe_Humbert: But like always, I'm going to defer to the group

Joe_Humbert: Dean said last night that he would hold off on finishing until some of these conflicts are resolved

Matt_King: I think your logic is strong, Joe_Humbert. I'm leaning toward rewriting issue 1375, to say that we actually fail these. It's not that we're dictating which word; we're just saying don't use "table" because that's already assigned to a different role in ARIA/HTML

Joe_Humbert: It's almost as if I used the word "text" to describe the role of a button

jugglinmike: I wonder if we want to clarify in the assertion text, e.g. "it distinctly conveys the role 'data grid'"

Matt_King: That would impact a lot of assertions. Maybe we can address this in the project documentation on the interpretation of assertions. That would allow us to keep the assertions very succinct

Matt_King: I think when we have assertions like "role 'button' is conveyed", I don't think we have guidance on what it actually means to convey the role successfully. Like, what does it mean to pass an fail?

Matt_King: Maybe we should get the AT implementers aligned on the parameters for passing and failing those assertions

jugglinmike: Yeah, I think that could go a long way

Matt_King: That conversation has come up with Apple and with Vispero. That is, the interpretation of assertions. I don't think our glossary defines the terms

Matt_King: This is defining the meaning of passing and failing for specific types of assertions, and that is something we have not done. It feels like something that would be beneficial

Matt_King: I feel like I should capture this in an issue...

Matt_King: We need to define the meaning of "passing"/"failing" and "supported"/"unsupported" for each of our types of assertions

Matt_King: Types of assertions isn't something we have defined, either. Now, though, we have enough history and data to write something meaningful about each of them. I think we only have, like, four or five kinds of assertions so far

Matt_King: With a few thousand assertions written, we have a pretty good basis for policy

jugglinmike: And there's also negative assertions

Matt_King: Oh, right. We should include those as well

Matt_King: I'm going to write an issue for this.

Matt_King: I took some notes, and I'll refer to the minutes of this meeting, too

Matt_King: I think Joe_Humbert has made a strong case for this. I'm going to change Dean's results and change the text of the issue I previously posted

Matt_King: I think we should probably talk about the JAWS conflicts, then

Matt_King: So this is mmoss and Joe_Humbert

Matt_King: The first conflict is in test 1!

Matt_King: The output is identical between the two

Matt_King: Joe_Humbert failed it and mmoss passed it

Joe_Humbert: If I remember correctly, this was because I was getting different results depending on which mode it was in.

Joe_Humbert: I guess I'll just change my results...

Matt_King: It says "virtual cursor active"

Matt_King: In this case, and in my experience with JAWS, this behavior is pretty reliable

Joe_Humbert: Okay, I'll change mine to "pass"

Matt_King: Okay, that's the first conflict

Matt_King: Now, for test 5...

Matt_King: I see this happen a lot with JAWS, but I never know the conditions that cause it. It always feels unpredictable. In this case, both of you got the same output

mmoss: This was a weird one. It says "not in table" but it is very clearly in the table

Matt_King: It's a strange thing that happens with JAWS. When it happens, I sometimes do "Insert + Escape" to refresh the virtual buffer. But sometimes I can't get the problem to go away

Joe_Humbert: Why does this report as a conflict?

mmoss: We both failed them...

Joe_Humbert: The impact is also identical. It's just the "details" that differ. I don't think the system compares that field

Joe_Humbert: This is for "control + alt + left arrow"

Matt_King: It didn't have side effects. It just didn't work

Matt_King: I think we can just fail it without a negative side effect

mmoss: I'm happy to change it. We're going to encounter this with another test in this test plan

Matt_King: Okay. Simply failing will be enough for the people at Vispero to recognize the problem

Matt_King: We also have output after "left arrow" with the PC cursor active

Matt_King: "To activate, press enter" is included in Joe_Humbert's output but not mmoss's

mmoss: Mine has "description" and "column three" in my output

Joe_Humbert: And mine does not

mmoss: I encountered what Joe_Humbert did, but only when I was NOT in PC cursor mode

Matt_King: Ah, that is the column heading. I've certainly experienced (with PC cursor) situations where the column heading doesn't get read. I don't remember if I had that problem in this grid when I ran this test, though.

mmoss: I was getting the column heading reliably

Joe_Humbert: I'll double-check a little bit later

Joe_Humbert: Same version of JAWS but not the same browser version. I don't think that will make a difference

mmoss: Me neither

Joe_Humbert: Did you have to change the mode to get to PC cursor? (I want to make sure I'm approaching this in a similar way)

mmoss: I don't remember

Joe_Humbert: I'll double-check and update this one

mmoss: And if you send an e-mail, then we can follow up asynchronously

Joe_Humbert: The other two are the same. We have the same negative side-effect and the same severity. It's only the "details" that differ

Matt_King: The text there shouldn't matter...

Joe_Humbert: I can copy and past the details to verify whether that matters

Matt_King: Going back to test 5 for a moment, the column header is not being read

Joe_Humbert: In test 10, when I copy mmoss's details verbatim, then the conflict goes away

Matt_King: That seems strange

Joe_Humbert: I'm going to copy mmoss's details for the other conflicting results

Joe_Humbert: We've calling this "moderate" because it's not gobs and gobs of extra information. The "bottom of file" part of the output is confusing, though. Adrian was mentioning that one of the most common usages of this is Google Sheets. Maybe that has to do with it, but that's pretty presumptuous

Joe_Humbert: That was test 10 and test 11

Matt_King: I can't wait until we have the latest JAWS bot installed because I think there has been a lot of changes made since version 2025--some positive and some negative

Matt_King: Great, so we reviewed all of the conflicts for JAWS

Matt_King: We don't have Dean present to talk about NVDA

Joe_Humbert: I can change mine because our output is identical. I created conflicts in order to prompt discussion here during this meeting

Matt_King: But you are agreeing with the way that Dean recorded his results, right?

Joe_Humbert: Yes

Joe_Humbert: So I will update to the correct information

Matt_King: Cool. Hopefully that will unblock Dean

Matt_King: Now, Dean isn't present to talk about changing his results for VoiceOver, either

Matt_King: But that would wrap up data grid

Plans for AT Driver specification

Matt_King: I wanted to make sure that everybody here (and everybody reading the minutes) is aware of AT Driver--what it is, what it does, and why it's important for ARIA-AT

Matt_King: We need some help in this space

Matt_King: Even if you are not the right person to work on this, some of you may know the person who is the right kind of person

Matt_King: So please think of this in that way as much as you can. Who do you know?

Matt_King: I included a link in the agenda to the Editor's Draft of the AT Driver specification

https://w3c.github.io/at-driver/

Matt_King: This is a specification for an API that screen readers can implement so that somebody else can "drive" that assistive technology via another program

Matt_King: We want to be able to tell an assistive technology, "do X and give us your response"

jugglinmike: Or in other words, we want to be able to script anything a user can do with an assistive technology

Matt_King: Right

Matt_King: There are parallel technologies for browsers: WebDriver and WebDriver BiDi both allow users to script web browsers in that way

Matt_King: We need this capability to have interoperability in assistive technologies

Matt_King: This is the tech that is behind the ARIA-AT project's "bots"

Matt_King: We want to be able to write a test and then essentially feed them to a screen reader in a format that a screen reader can respond to

Hadi: do we want to be able to use them from the desktop apps, not necessarily from the browser?

Hadi: My understanding is that we want a kind of API that I can embed in my app so that it can selectively do stuff that a screen reader would do

Matt_King: If you have an app that you want to test, and you want to be able to write a script that says, "run a screen reader and perform these actions in my app"

Matt_King: I believe we have written this in such a way that is not constrained to the browser context

Matt_King: All of our implementations are related to browser technologies

Matt_King: I think we've built AT Driver in a way that is not limited to driving apps inside of a browser

jugglinmike: That's right

Matt_King: We're just about out of time

Matt_King: To become a standard, you need the support of a standards body. We're working through the W3C even though it is a web-focused organization because we think that's a good home to do the work

Matt_King: We need to make sure there are enough people to help. These are essentially volunteers; the W3C doesn't pay people to do standards development; it coordinates that work

Matt_King: jugglinmike has been the only editor to work on this so far. We're short on funding from when Meta pulled funding

Matt_King: We have applied for one grant

Matt_King: Even if we have someone who could spend just a few hours a week, that would be better than nothing

Matt_King: Are there other kinds of support that we could seek from across the community to help move this forward?

jugglinmike: I don't have a great answer to that question right now

Matt_King: Okay. We're in a bit of a state of limbo right now because W3C charters are getting renewed. One of the challenges that the people who run the committees are facing is that the spec is at a bit of a standstill at the moment

Matt_King: So if you are (or know people who are) interested, then jugglinmike and I would be very happy to know

Matt_King: This is ground-breaking work. If there are things we can do to help raise the visibility of what the need is, then I'd love to talk about that

Matt_King: We're at time. We can talk more about this in the future

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

Diagnostics

All speakers: Hadi, Joe_Humbert, jugglinmike, Matt_King, mmoss

Active on IRC: jugglinmike, Matt_King, mmoss, siri