W3C

– DRAFT –
ARIA and Assistive Technologies Community Group

17 June 2026

Attendees

Present
dean, Joe_Humbert, jugglinmike, Matt_King
Regrets
-
Chair
Matt King
Scribe
jugglinmike

Meeting minutes

Review agenda and next meeting dates

Matt_King: Any requests for changes to the agenda?

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

Matt_King: Next CG meeting: Wednesday July 1

Current interop reporting status

Matt_King: There aren't any meaningful updates in the overall status of test plans since the previous meeting

Matt_King: I'm working on figuring out what we will test after the data grid

Matt_King: I do want to get the rating radio button out of the queue, and I'm looking at the issues associated with that

Matt_King: I'm trying to figure out how readily-resolvable those are--by just me, or with input from others (and who)

Matt_King: This is partly due to a bandwidth issue right now. I'm not sure how quickly I can keep test plans coming. I would like to have a minimum of one test plan for people to be working on at any given time

Matt_King: If anyone would like to become an expert in writing test plans, it's quite an investment, but it's not super-technical for people who know all three screen readers. The editing is done in CSV files. Scripts use those files to generate JSON (so you don't need to edit JSON directly)

Matt_King: If anyone wants to advance to another level in this work, I'm open to training people on writing test plans!

Matt_King: They don't all have to be for complicated APG patterns, by the way. We could also use simpler test plans for basic HTML and ARIA features

Draft review of minimal data grid test plan

Matt_King: For the data grid, Joe_Humbert has completed JAWS and NVDA

Joe_Humbert: That's right

Matt_King: I saw you raised some issues in both, and I've responded to those

Matt_King: The issue that is related to JAWS forms mode versus PC cursor. The output in the response edit fields--is that the output from the bot?

Joe_Humbert: I didn't have to edit any output at all

Matt_King: Got it. So if Vispero runs it with the bot, they will get the same results

Joe_Humbert: Right

Joe_Humbert: Is this representative of what users would expect? I don't know many users who will exit out of forms mode into PC mode within a data grid

Matt_King: I think it happens quite a lot for people who have to use data grids

Joe_Humbert: I understand if its happening automatically, but I don't think users would manually exit out of forms mode

Matt_King: A surprising number of users are. But a lot of these grids are getting wrapping in the "application" role, which does the same thing. That's how Google Sheets works.

Joe_Humbert: Ah, so if it's wrapped in "application" mode, then it would be in PC cursor mode and not forms mode

Matt_King: Right. It's a weird thing

Matt_King: Interestingly, JAWS 2026 is even more representative of user experience because even forms mode and PC cursor mode exhibit the same behavior

Matt_King: So this seems like a bit of a regression in JAWS behavior

Joe_Humbert: I closed the other issue because it was something I was noticing, but testing in full-screen mode would make things much harder. So I generally don't run in full-screen mode

Matt_King: I think even if a popup window is maximized, it's still not full-screen

Matt_King: A screen reader user presses Alt+Space and then X in Windows. That maximizes a popup window, but I don't think it actually goes into full-screen mode because that's popup

Matt_King: IsaDC and I noticed this: the script can behave differently if the window is small. This doesn't make sense if you review what's happening in the script, but we definitely observed flaky focus behavior. I think that may be a Chrome problem

Joe_Humbert: Got it

RRSAgent make minutes

Joe_Humbert: I edited my test run for NVDA as soon as you told me that it was available

Matt_King: Cool! So that worked

Matt_King: Now, for VoiceOver

dean: I received your e-mail. No problem, I just re-did it all this morning

Matt_King: Oh, already! I see that, now. It's all there. Awesome!

Matt_King: We have four conflicts between you and IsaDC

dean: I have a question about that

dean: This is one of those cases of vernacular. Two differing definitions

Matt_King: Navigate backwards into a grid, test 2...

dean: They are wrong. It says "role grid is conveyed". They said "no", but it says "table"

Matt_King: It's kind-of/sort-of okay that Apple is calling a grid a table

Matt_King: I just don't know how people should recognize that they're navigating within a static table versus an interactive table/grid...

Matt_King: If you look at the "help" message that VoiceOver gives, it tells you to use the VoiceOver commands Ctrl+Option+{Up,Down,Left,Right}

Matt_King: I don't know what happens with VoiceOver when you reach the end of a grid that loads dynamically

Matt_King: It might just work, but it might depend on how the JavaScript is written

Matt_King: I don't know if this is beyond the scope of ARIA-AT. I almost feel like raising an issue with Apple, but I don't know if we can really claim that they failed to convey "role=grid" because tables and roles are the same thing

Matt_King: But one might argue that they are conveying the wrong thing because table is a different role than grid

Matt_King: I think that's essentially what IsaDC is saying here

dean: I'll leave that to you

dean: But there are three conflicts due to this behavior

Matt_King: I'm surprised there are only three!

dean: The fourth one is once again a bit of a subjective call

dean: Assertion is "number of columns, 5, is conveyed". The output says "column 5 of 5". To me, the number of columns is conveyed

Matt_King: Clearly, yes

Matt_King: I think IsaDC is mistaken on that

Matt_King: Thinking about our progress, I think what we should do for the table-vs-grid thing, is to raise an issue to bring Apple in and get some input even while we're in draft review

Matt_King: Because the only way to start a conversation during candidate review is to fail the test, and I'm not comfortable assigning a failing status to this test because it is so subjective

Matt_King: We could reach out to IsaDC via e-mail or maybe on GitHub for that specific issue. She's not present today, but I don't want to wait until the next meeting if we don't have to

dean: I'm ready to move on to NVDA

dean: While we're here, though, I have another question. There are two tests where it records "no output". We have the same results on that, so the app says "no conflict". But I wanted to report a side-effect because the cursor does not move. Am I correct to believe that is a severe negative side-effect?

Matt_King: Which test was that?

dean: "shift+y" and "y"

Matt_King: Ah, okay. That command. So I actually am not certain if that is a try default. The question I have (and I don't have a Mac right now) is when VoiceOver is in its "default settings" state, are "y" and "shift+y" actually available as quick-nav commands?

Matt_King: When I last had a Mac, I think on the version I had, those commands were not available, but on IsaDC's version, they were

Matt_King: One way to test that on your computer is to turn on letter quick nav (with ctrl+option+q) and then turn off VoiceOver's "help" mode (with ctrl+option+k), and then press the letter "y". If VoiceOver doesn't say anything (like "moves to next column in a table"), then we just need to remove the command from the test plan

Matt_King: So when you press the "h" key in "help" mode, it should say something like ("moves to next heading")

Matt_King: If it accurately describes any quick nav command but it is silent for "y", then we know we shouldn't be testing it.

Matt_King: If it's not defined as a command on your computer, could you please raise an issue for that test and request that we remove that command? Please make sure to include your version of MacOS

Matt_King: I'm so glad you asked that! There are some VoiceOver documents out there that claim that "y" is a command for moving to the next column, and other docs that don't claim that. So I'm confused

Matt_King: The official Apple documentation is odd that it doesn't have a definitive list of quick-nav commands. Not one that I can find, anyway

dean: I'll get NVDA done before the next meeting

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

Diagnostics

Succeeded: s/agendum 3 -- Update on infra engineering training/agendum 3 -- Draft review of minimal data grid test plan/

All speakers: dean, Joe_Humbert, Matt_King

Active on IRC: Joe_Humbert, jugglinmike, Matt_King