W3C

– DRAFT –
ARIA and Assistive Technologies Community Group Weekly Teleconference

21 August 2025

Attendees

Present
ChrisCuellar, Hadi, howard-e, IsaDC, james, Joe_Humbert, jugglinmike, Matt_King, mfairchild, mmoss
Regrets
-
Chair
-
Scribe
jugglinmike

Meeting minutes

Review agenda and next meeting dates

https://github.com/w3c/aria-at/wiki/August-21%2C-2025-Agenda

Matt_King: Any requests for changes to the agenda?

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

Matt_King: Next CG meeting: Wednesday August 27

Matt_King: Next AT Driver Subgroup meeting: Monday September 8

Current status

Matt_King: We have seven plans in draft review right now

Matt_King: Five of them are pretty active, and two of them we have to get back to

Matt_King: I was hoping I would have time to look into the "disclosure navigation menu" to see if there is still anything blocking us, but I don't think we'll get to that today

Matt_King: IsaDC is working on three more for draft review

Matt_King: We might get those mostly ready for testing by the end of the month

Matt_King: That will put us within striking range of having 25 plans in Candidate Review

Matt_King: That could happen in September. It will be a pretty big step up!

Running vertical temperature slider test plan

Matt_King: We've completed both JAWS and NVDA on this one, which is awesome

Matt_King: There are only three conflicts for VoiceOver. These are related to two tests that are untestable

Matt_King: Last week, we were waiting on an app fix (where it would automatically uncheck yes/no radios when we checked "untestable")

howard-e: We released a fix for that

IsaDC: Do Testers have to re-submit results?

mmoss: Because we still have conflicts

mmoss: I marked this as "untestable" originally

mmoss: This was one that I re-ran. We didn't have the "untestable" test run for my initial run

Matt_King: This is for tests number 8 and 9

mmoss: Oh, this might be an issue with Dean's results, but not with mine

IsaDC: Yes, Dean passed an assertion incorrectly

Matt_King: Does this submit?

IsaDC: What happens if you submit it?

Matt_King: On yours, mmoss, it says "other"

mmoss: Right, because it doesn't seem like the cursor is actually changing--it's just reading the wrong thing

mmoss: That's why we changed it to "other". I don't think Dean has done that

Matt_King: I think in the description, we would want to say something like "instead of reading the new slider value, it is reading the heading of the slider"

Matt_King: ...so it's a little more clear exactly what the problem is

Matt_King: But instead of saying "seems to have changed", should we just say, "it's announcing the heading"?

mmoss: Yes, I think that's more clear

Matt_King: Okay, so if you could update your test results

Matt_King: Dean couldn't be here today, but he has observed the same behavior, so I think it's fair to update his results for both test 8 and 9

IsaDC: Okay, I will do that

Matt_King: Great--then, this test plan will be done

Running accordion test plan

Matt_King: This is one that I don't know if we can really talk about. We're all done with this one, as well. JAWS and NVDA are both complete

Matt_King: We have 42 conflicting VoiceOver results

Matt_King: Now that we have the app fix, I think we're good to go here

Matt_King: We just need Dean to go through and correct his results. I guess he didn't get to that, yet

IsaDC: I will e-mail Dean and tell him what we need

Matt_King: Just taking care of the "untestable" stuff should cut down that 42 number down by a lot

Matt_King: And I recall that there were some other changes that he was going to make

Running "Tabs with Automatic Activation" test plan

(This topic is not present in the agenda. The group decided to add this topic on-the-fly)

Matt_King: We made changes to tests 10 through 14. That's eight of the 15 tests changed as a result of last week's discussion

Matt_King: There were conflicts for JAWS, but they could go away because they are in tests that changed

Matt_King: In test 3, the outputs look the same between Hadi and Louis

Hadi: I think Louis was complaining about some redundancy--the word "selected" is repeated

Hadi: I didn't think that was a reason to fail the test

Matt_King: It says "Selected" and then it says "One of four selected"

Matt_King: That is a little strange

Matt_King: I don't know if we should report that or not

Matt_King: I kind of feel like Vispero might want to know about it. It's pretty moderate. But it certainly isn't ideal.

IsaDC: I mean, it is reporting it twice

Matt_King: It's not expected. We say there should not be any moderate side-effects

Matt_King: I think there's no harm in reporting it and then seeing what Vispero says. They've been very responsive. I think this is a judgment call

Hadi: I agree. Let's report it. If they are eager to fix it, then that's good; otherwise, we'll live with it

Matt_King: If we look at a strict reading of our criteria, this test doesn't give any incorrect information, so it doesn't quite meet the bar

Matt_King: We could raise an issue with them but not put it in these reports

Matt_King: If we're sticking to our guns about our definition of what "excess verbosity" is, that's probably the more judicious route

Hadi: Redundancy generally raises my suspicion

Hadi: I think we can report it to them and see

IsaDC: I think they'll appreciate the feedback. It's moderate, after all

Matt_King: I'm fine with leaving it in. If you want to match that in your report Hadi, you can

jugglinmike: This reminds me a little bit of alert. If Vispero wants to keep it, would we add a "MAY" assertion for this?

Matt_King: I can't think of an optional assertion that we could add that would be relevant

Matt_King: We already have an assertion related to the selected state, and it's a MUST

jugglinmike: You could write something like "MAY communicate the selected state twice", if only to commemorate the fact that we observed this, talked about it, and decided to accept it

Matt_King: I don't think there's much room for interoperability improvements here

Matt_King: Okay, so this affects multiple tests

Matt_King: I think all of these are related to the same thing

Hadi: I submitted an issue. Do you see the word "none not selected" when using arrow up on a tab list?

Hadi: The question was confusing because it gave the impression that you should be hearing "not selected"

IsaDC: That was a mistake on our side; we've corrected it, but we haven't added the new test plan into the test plan, yet

Matt_King: So you merged it, but it's not in the queue, yet?

IsaDC: No

Matt_King: Okay, let's do that. It's going to change the whole conflict picture

[Matt_King forces the new version of the test plan into the queue]

Matt_King: Now it says that it's 40% complete for JAWS with two testers. And now it says 9 of Hadi's 15 tests are complete (so we essentially lost 6 from you, though that doesn't mean those tests are fully lost). And we have five conflicts

Matt_King: The conflicts are different because you're results for test 3 need to be revisited

Matt_King: The conflicts for test 5 and 7 are still here, and those are the same

Matt_King: Test 5 has the same kind of verbosity conflicts that we were just talking about a minute ago

Matt_King: Test 7 also has that same thing. They're very consistent here

Matt_King: And also test 8

Matt_King: So I think we've covered that conflict. Hadi, can you go back through tests 1 through 4 (and I'm not sure which others)?

Hadi: Okay!

Matt_King: Then we have two conflicts in NVDA, and this was Joe_Humbert and Louis

Matt_King: Joe_Humbert, a lot of the feedback on this test plan came from you, so hopefully we've addressed everything that you thought was a problem

Matt_King: In terms of conflict, we have test 10, "activating the last tab"...

Joe_Humbert: We didn't get to talk about this last time. The bot had repeated information because the speech viewer was repeating information, but the screen reader was not voicing the repeated information

Joe_Humbert: What the screen reader was saying and what was present in the speech viewer was not matching

Matt_King: Oh my!

Joe_Humbert: Louis isn't here, but I suspect he changed the output. I left mine so that I could bring it to the Community Group

Joe_Humbert: I initially thought this was a problem with the bot, but when I went to run the test manually, I observed the same behavior

James: the bot is collecting the information which is sent to the synthesizer

Joe_Humbert: This occurs on a couple of the tests for "tabs with automatic activation" with NVDA. The collected output was wrong because what was present in the speech viewer did not align with what was voiced by the screen reader

mfairchild: I suspect that NVDA is sending the string to the vocalizer twice

mfairchild: And that the voice interrupts itself

Joe_Humbert: I don't know if this is feedback that we need to send to NVDA

James: You can start to hear the first vocalization before the second interrupts it. It sounds almost as if it is stuttering

Matt_King: Is there a chance that there are two "focus change" events?

James: That's possible

Matt_King: I think we might want to mark this as excess verbosity

Matt_King: Does the dupe show up in Braille? Because you don't get interruptions in Braille

James: so you're suggesting that it is render the Braille twice? One after the other? Or that it would "click" twice in rapid succession?

Matt_King: I don't know

mfairchild: I don't think I would report it if it doesn't impact end users

Matt_King: Yeah. I'm having this question in my head...

IsaDC: It doesn't announce it twice

Matt_King: It introduces a stutter

Joe_Humbert: This is why I captured it. It's weird, and I thought we needed to discuss it

Matt_King: The bot reported it twice because the bot gets it so fast

Joe_Humbert: Do I need to change mine?

Matt_King: I'm thinking about how this works with the automation. If we record it once, and then in the future, the bot spits it out twice, again, we're in a situation where it always conflicts

Matt_King: We don't want to say that the "double output" is good output, but we also don't want to say it's a failure

mfairchild: This is a situation where it would be great to be able to document what went wrong but still test it

Matt_King: In the side effects, we have "SEVERE" and "MODERATE". We could add a new value that we have observed something that doesn't matter

Matt_King: If we put in a severity that's equivalent to "NONE" (and maybe rename from "negative side-effect" to just "side-effect"), then we could report it but avoid failing an assertion

Matt_King: In the mean time, the only question is: do we delete the duplication or not?

IsaDC: I think we should keep it because of the bot

ChrisCuellar: I'm curious about the process. This happens in the VO Bot (I'm discussing this with Corey as we speak).

ChrisCuellar: Let's say the bot reports duplicate announcements (for whatever reason). Joe_Humbert, when you get the bot run reassigned to you, and you're determining the verdict--I would assume you are running the test again and confirming whether you observe the same thing

Joe_Humbert: Correct

ChrisCuellar: So whatever the human tester observes on their machine, that should supersede the bot's output

James: We are observing this. I can hear a stutter

IsaDC: If we use the speech history add-on, it would record this twice, as well

Joe_Humbert: It does. That's why I brought this up

IsaDC: I think this a failure

Matt_King: Okay, but we agree that we would report this as a "NONE" side-effect (instead of "MODERATE") if we could

James: It shouldn't say it twice.

Matt_King: We can file an issue to discuss this further and then in the comments for this moderate negative side-effect, we could include a link to that issue

Joe_Humbert: So, besides changing this from "other" to "verbose output", do I need to change the details?

Matt_King: We could say, "Speech is duplicated but only heard once due to automatic interruption"

Joe_Humbert: Sure, I can do that

James: I'm really curious to how this is expressed on a Braille display

Switch and disclosure

Matt_King: The main thing we need right now is testers

IsaDC: And we have Bot results ready

Joe_Humbert: I can help

Joe_Humbert: You can assign me any screen reader you want

Joe_Humbert: Just know that my new test machine is Windows 11, not Windows 10

IsaDC: I'll assign you to all three, then. Thank you!

mmoss: I volunteer for VoiceOver

IsaDC: Thank you!

Matt_King: Thank you, everybody. We're making some really awesome progress here

Minutes manually created (not a transcript), formatted by scribe.perl version 244 (Thu Feb 27 01:23:09 2025 UTC).

Diagnostics

Succeeded: s/Deans/Dean's/

Succeeded: s/betweeh/between/

All speakers: ChrisCuellar, Hadi, howard-e, IsaDC, James, Joe_Humbert, jugglinmike, Matt_King, mfairchild, mmoss

Active on IRC: howard-e, Joe_Humbert, jugglinmike, Matt_King, mmoss