W3C

– DRAFT –
(MEETING TITLE)

26 March 2025

Attendees

Present
Carmen, dean, howard-e, Isa, james, Joe_Humbert, jugglinmike, Matt_King, mmoss
Regrets
-
Chair
-
Scribe
jugglinmike

Meeting minutes

ARIA and Assistive Technologies Community Group Weekly Teleconference

Review agenda and next meeting dates

https://github.com/w3c/aria-at/wiki/March-26%2C-2025-Agenda

Matt_King: No meeting Thursday April 3

Matt_King: Next Community Group Meeting: Wednesday April 9

Matt_King: Next AT Driver Subgroup meeting: Monday April 14

Matt_King: We have more on today's agenda than we can probably cover during this meeting

Matt_King: Any requests for changes?

Matt_King: Hearing none, we'll stick with what we've got planned

Current status

Matt_King: The top-level status of what's in each phase hasn't really changed in the past week

Matt_King: I was really trying to get the "vertical temperature slider" over the line so we could have it in the test queue for today's meeting, but I made a mistake which delayed that

Matt_King: I shared some feedback on the test plan

Isa: I saw those. I've been waiting for the build so I could preview the test plan

Matt_King: I think it will be ready to go if all the changes that I made look good to you

Matt_King: I wanted to look at it again also in the actual "preview" mode to make sure that I didn't make any other mistakes

Matt_King: This could end up available for testing as soon as today

Matt_King: ...but it isn't in the test queue right now

Matt_King: I think we should decide now who we will assign, assuming we can land it in the test queue today

Matt_King: because it would be really good to make progress on that before our next meeting on April 9

Matt_King: Once we resolve conflicts on stuff that's in the queue right now, the queue will be empty. That must mean that there are a bunch of people who will have time in the next few weeks

mmoss: I would be happy to work on the "vertical temperature slider"

mmoss: It could be VoiceOver or NVDA. (Hopefully JAWS, as well, by the end of this week)

Matt_King: Great! Is anyone else available?

dean: I am. I prefer VoiceOver, but I can do NVDA because I have Parallels

Matt_King: Okay, we can assign NVDA to mmoss and VoiceOver to dean

Isa: I can do JAWS

Joe_Humbert: I can do whatever I need to do

Joe_Humbert: JAWS has been misbehaving on my system

Isa: Luke is not available for testing

Matt_King: So we have Joe_Humbert for JAWS, dean for VoiceOver, and mmoss for NVDA

Matt_King: Let's try to get this one over the line and into the queue today, Isa

Isa: Okay, I will review it

Radio group with active descendant

Matt_King: I wanted to celebrate because this is all done! Thank you, everybody

Matt_King: It has zero open issues

Matt_King: Pretty soon, the next "radio group" plan will be there

Matt_King: Thanks to everyone who helped get this over the line

Testing Radio group with roving tab index

Matt_King: We have some conflicts. I put a link to the conflicts in the agenda

Matt_King: Isa raised an issue: w3c/aria-at#1221

Matt_King: This is a situation where the conflicts are because Joe_Humbert and Isa got different AT responses on three tests

Joe_Humbert: Most of the conflicts are about the "set size" and "set position"

Joe_Humbert: I noticed when I was testing, and I re-tested three or four times just to be certain that I wasn't making a mistake... So I don't know what's going on with that

Matt_King: The first thing to double-check is that you are both using the same build

Isa: There has been an update since I ran that test plan, but I still get the same results. It doesn't speak the position for me

Matt_King: Joe_Humbert is using the demo, so it's always with the default settings

Joe_Humbert: I installed and immediately started testing. I assumed that this would produce the default settings

Matt_King: It is the default unless you make modifications during the "setup" wizard

Matt_King: If you didn't change anything other than speech rate there, then you have the defaults

Joe_Humbert: I didn't even change the speech rate

Matt_King: Isa are you using the "Insert Z" method?

Isa: Yes. I haven't changed anything because I use NVDA by default

Matt_King: When you do the testing--you press "insert space" followed by "z", and then JAWS says "default user alternate directory mode"

Matt_King: Do you use that method to ensure that the defaults are all in place?

Isa: Yes. I only speed it up a bit

Joe_Humbert: With my output, it's also announcing the group and group label. That information isn't included in Isa's output

Isa: Right, it doesn't announce that for me. I tested again last night

Matt_King: This seems like an usual situation. I think Isa is right--we need a third person

Matt_King: It could be me, but I don't know when I will be available. I still have to install that latest version in addition to running the test itself

Matt_King: ...but we need a tiebreaker, and I don't think we have anybody else on the call right now who could be a tiebreaker

Matt_King: It will be great to get more JAWS testers

Matt_King: I predict that I will get a third, unique set of AT responses

Matt_King: This is going to be kind of stuck until I (or someone else) can run the test plan

Joe_Humbert: I have been going through and checking and hitting "alt delete" and making sure it is in the right cursor mode

Matt_King: Well, pressing "alt delete" won't impact the results

Joe_Humbert: Should I go through and try to reset the defaults just to make sure there is nothing wonky in the way it is freshly installed?

Matt_King: I suppose you could do that. To me, though, what's interesting is that your results make it appear that JAWS has fixed some stuff. Compared to Isa's results, you got more passes

Matt_King: So I'm inclined to believe Joe_Humbert's results are more authentic than Isa's

Joe_Humbert: I just re-did one of the test. When you run one of them, it puts focus on a radio button, and it goes into forms mode. So I have to leave forms mode to run the test

Matt_King: Correct

Joe_Humbert: I'm just leaving forms mode by hitting "escape". Could it be because I'm using the single-key shortcut command to change modes? If I exit forms mode in a different mode, could that impact anything?

Isa: I'm just hitting "escape" as well

Joe_Humbert: Okay, nevermind, then. If we're both doing the same thing, then that's not it

Joe_Humbert: Isa, what operating system are you running?

Isa: Windows 11

Joe_Humbert: I am using Windows 10

Joe_Humbert: I have Windows 11 in a virtual machine. I could test there to see if I get different results

james: I don't have the latest version of JAWS

Joe_Humbert: I hope it's not because I'm using an unregistered version. It would be very, very weird if the licensed version behaves differently than the unlicensed version

Matt_King: Agreed

Joe_Humbert: I can try this on Windows 11 in a virtual machine

Matt_King: It would be weird if that's the problem, but I think it's worth doing

Matt_King: If that's the explanation, then it could mean that Windows 10 and 11 have a difference that JAWS hasn't accounted for. Perhaps even that Windows has regressed...

Testing Media Seek Slider

Matt_King: This should be a lot easier to resolve

Matt_King: We have a ton of conflicts, but we have the same AT responses, so this is just about how testers are interpreting the AT responses

Matt_King: Isa raised an issue w3c/aria-at#1215

Matt_King: The conflicts are summarized here: https://aria-at.w3.org/test-queue/375/conflicts

Matt_King: You have differences in almost every test

Matt_King: In test number one: you both reported the same output

Matt_King: Isa put "pass" and Joe_Humbert put "fail"

Joe_Humbert: I misread the "not" in the assertion. I need to pay more attention now that we have negative assertions!

Joe_Humbert: I'll go through and correct my verdicts accordingly

Issue 1216 - How to handle failures that invalidate assertions

github: w3c/aria-at#1216

Matt_King: There are situations were--and we previously saw this with VoiceOver--where you press a key to move to a certain button, and it navigated to the wrong button

Matt_King: Here, NVDA just totally skipped over the slider

james: It doesn't consider a slider to be a form control

Matt_King: Some of the subsequent assertions might spuriously pass--the role might be correct, for instance, but only coincidentally

Matt_King: We don't have a good way in our testing form to record the fact that the screen reader just did something unexpected

Matt_King: Essentially, this kind of behavior really blocks testing

Matt_King: We do have a checkbox for what happens when you get no output at all

Matt_King: I'm starting to think that we may need something similar for this. I made a specific proposal along these lines

Matt_King: The proposal is in an issue in the app repository--issue 1352: w3c/aria-at-app#1352

Matt_King: [reviews the proposal]

james: Where would the control be to choose the invalidation of the assertions?

Matt_King: I think it would be directly after the output box and directly before the assertions

james: Make sense to me

Isa: It will be a set of radio buttons as well?

Matt_King: I think it would a "yes" or "no". It could not be a check box

james: this is very similar to the "no output" control. I think it should be a check box

james: The typical argument for radio buttons in this case is to guard against people failing to provide input. But in this context, that risk isn't meaningful

Matt_King: We could say, for instance, "the response to 'f' created conditions that invalidated the assertions"

Matt_King: I'll edit this proposal

Matt_King: howard-e and jugglinmike, I think this will create a much more accurate report to say that the assertions were untestable

Matt_King: What we've previously done in these scenarios is a problem

Matt_King: I almost want to say that completing the test for this command with NVDA, we can't record accurate results unless we make this change to the app

Matt_King: We could agree to fail everything and record a negative side-effect as a temporary solution, but then I think we might want to come back to this

Matt_King: Should we agree to do that in this circumstance?

james: Yeah, I think that works--until this new feature is implemented.

Matt_King: The problem is that it creates a non-actionable signal for NVAccess. By failing the assertions, we are saying that there is a bug present which isn't there

Matt_King: You could hypothesize that the intended behavior is correct, but that is a stretch for the tester

james: Once this new feature is implemented (to invalidate assertions), how will that impact the support level?

Matt_King: My proposal addresses that. Because those assertions are marked as "untestable", they wouldn't be counted in the denominator of the assertions. But there IS an assertion in there which would fail

james: Okay

james: Separately, we are asserting that NVDA should consider a slider to be a form control. We're only doing that because other screen readers do

Matt_King: It treats a "range" input as a form control

Matt_King: but not an ARIA slider

james: Oh, okay. That makes sense to me

jugglinmike: To be clear: this is not a state that the automation system could interpret on its own

Media seek slider, revisited

Joe_Humbert: Oh, never mind. I misread the assertion again. I really have to read these negative assertions closely

Joe_Humbert: I was able to install JAWS on Windows 11 in a virtual machine...

Matt_King: That was fast!

Joe_Humbert: ...and I get the same AT responses that I originally reported for JAWS on Windows 10

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

Diagnostics

Succeeded: s/Topic/Matt_King/

Succeeded: s/test/test plan/

Succeeded: s/result/resolve/

All speakers: dean, Isa, james, Joe_Humbert, jugglinmike, Matt_King, mmoss

Active on IRC: Joe_Humbert, jugglinmike, Matt_King