W3C

– DRAFT –
ARIA and Assistive Technologies Community Group Weekly Teleconference

13 August 2025

Attendees

Present
Carmen, ChrisCuellar, dean, howard-e, Isa, IsaDC, james, Joe_Humbert, jugglinmike, kelly, louis, Matt_King
Regrets
-
Chair
-
Scribe
jugglinmike

Meeting minutes

Review agenda and next meeting dates

https://github.com/w3c/aria-at/wiki/August-13%2C-2025-Agenda

Matt_King: Requests for changes to agenda?

ChrisCuellar: There's an open issue (#1349) that was filed by james. I wanted to remind folks to check on it

Matt_King: We'll do that first

Matt_King: Next CG meeting: Thursday August 21

Matt_King: Next AT Driver Subgroup meeting: Monday September 8

Lack of screen reader feedback after pressing "Submit results" button

github: w3c/aria-at-app#1349

ChrisCuellar: This behavior is somewhat confusing; we do we do it?

Matt_King: We do it to allow testers to confirm that their submission was correct

James: I would be in favor of a button which saves your results and also moves on to the next test

Matt_King: Another reason for "submit" is that, if there is errors, it prompts you to go back and correct them

Matt_King: We have a "next" button so Testers can advance without losing what they've done

Matt_King: It always saves what you input regardless of which button you press, but it doesn't validate the input if you press "next"

ChrisCuellar: I had some proposals in that issue, so I wanted to remind folks to revisit that discussion

Running vertical temperature slider test plan

Matt_King: I think people simply haven't had time to do the work we assigned last week

Matt_King: Last week, we had conflicts with NVDA and resolved them, and then published the results

Matt_King: For JAWS, we noted that IsaDC was using an older version than Joe_Humbert. She was going to re-run that to resolve the conflicts

IsaDC: I haven't found the time to do that, yet

Matt_King: Okay, that's fine

Matt_King: Then for VoiceOver, we had a bug where if we had already saved results, there were still assertions that were being checked. It didn't automatically unchecked the "yes"/"no" radio buttons when you designated "untestable"

Matt_King: Is that bug resolved?

jugglinmike: It is not resolved

Carmen: But it is ready for release, and we're aiming to release the fix this afternoon

Matt_King: Okay, then that's as far as we can go with the "vertical temperature slider" test plan

Running accordion test plan

Issue 1269: Conflicting JAWS results: "Navigate backwards to an expanded accordion header" (Accordion, Test 2, Command "'Shift Tab' (PC cursor active)"

github: https://github.com/w3c/aria-at/issues/1269#issue-3240186747

Matt_King: Ah, Louis closed this issue last week. This is an agenda mistake

NVDA

Matt_King: Both your browser and NVDA versions are different. Louis is using a newer release of both pieces of software than Joe_Humbert is

IsaDC: I got the same responses as Louis when I tested with the latest release

Joe_Humbert: I'll re-test, then

VoiceOver

Matt_King: All of the tests related to navigating backwards with the quicknav key--those are all going to be affected by this bug in marking as "untestable"

Matt_King: Are there any other problems?

dean: I don't think so

dean: At least, not for the "navigate backwards" test

dean: But as for "navigate forwards", there seem to be some other kinds of discrepancies

dean: In test 3, we got the same AT response, but I marked it as "pass" while the other tester marked it as "fail"

<Joe_Humbert> NVDA conflicts resolved for Accordion

Matt_King: Are you supposed to navigate forward to the personal information? Or to the one after that?

Matt_King: I think that it might be failing because that might be the wrong heading. As in, it didn't actually move..

Matt_King: It's supposed to go to the "billing address"

dean: ...but it's going to the wrong one

dean: Okay, I need to revisit some of my tests. That should resolve some of these conflicts

Matt_King: The command failed which made the assertions untestable

Matt_King: You'll have to wait for the upcoming app fix regarding the "untestable" feature before you can report that, though

Matt_King: We might have a bug in the conflict report, because the conflict report may be showing "fail" instead of "untestable"

Running tabs with automatic activation test plan

Matt_King: There were some issues opened, but I didn't get them added to the agenda

Matt_King: The issues that I looked at weren't related to conflicts; they were related to the test plan and to the app

Matt_King: For JAWS, we have four conflicts. For VoiceOver, we have 3 conflicts. It appears that NVDA is all the way done--awesome!

IsaDC: Hadi posted a question about the "tab list" role announced as a group

louis: I don't think any of the screen readers actually say "tab list", right?

Matt_King: JAWS used to

louis: That's a point of confusion that I'd like to clear up

Joe_Humbert: There are some things that I didn't change despite agreeing with your verdicts because I wanted to discuss them here in this meeting

Matt_King: You did raise some issues, Joe_Humbert, and I did want to get to those.

Joe_Humbert: Let's start with the conflicts because I think those can be resolved

Joe_Humbert: The conflict I had in test 1: we had the same output, but the assertion is that the name of the tab list is conveyed. It is not. The heading before the tab list is spoken

Matt_King: That is a really good question.

Matt_King: You press the "down arrow" once, and it reads the heading. When you press the "down arrow" again, it reads the tabs without indicating that it is in a tab list

Matt_King: I wonder if we should change this test. Should we move the navigate forward from here after the heading?

Matt_King: I think that would make the output a lot more clear

Matt_King: I think if you were reading the report and not looking at the test case, you would be very confused about the failure here

Matt_King: We can fix that by moving the link

Matt_King: There's also a surprising bug in VoiceOver when you are navigating backwards

Joe_Humbert: The other question I have is about the same conflict with unexpected behaviors

Joe_Humbert: Louis reported that reading all the tabs is an unexpected behavior

Joe_Humbert: I agree that it's unexpected behavior for anyone who has not used NVDA before, but I think that anyone who has used NVDA would expect it to dothat

Matt_King: When we did the "radio" test plan, we agreed that because that is how NVDA works (and because they could argue that it's okay even though a lot of us would say it's terrible)

Joe_Humbert: My complaint about it is as follows: when you are navigating virtually, and it's reading out everything, it can be very difficult for a novice user to recognize when inputs are selected

Louis: that's a challenge even for advanced users!

Matt_King: Didn't we agree that we couldn't call that a negative side effect because that's actually us delving into the realm of telling them how to essentially design their screen reader?

Matt_King: We came across this with the "Down arrow" for radio buttons

Joe_Humbert: Any times you have a set of controls which are grouped together, NVDA basically reads the entire group

Matt_King: But only if they're horizontally aligned visually

James: I think we decided that, for radio buttons, we can't report a failure because NVDA would say that the screen reader is functioning as intended (it's just that we find the intended behavior questionable)

Matt_King: We can't call it a negative side effect. They wouldn't call it excess verbosity, and all the information they are reported is accurate

Matt_King: That was our rationale for not calling it a negative side effect (even though it kind of makes our stomach churn)

James: I ran a survey prompted by this meeting, and the responses were overwhelmingly "no" and "I don't know what that is"

Joe_Humbert: It sounds like, for now, we should remove reports of unexpected behavior

Louis: Yup. I can do that

Test content and VO bot Feedback: "Navigate backwards into a tab list where a tab is not selected" (Tabs with Automatic Activation, Test 2)

github: w3c/aria-at#1280

Joe_Humbert: When you run test 1 (navigating forward into a tab list), when you run the test setup, no tabs are selected and no tab panel is visible

Joe_Humbert: For test 2, however, after running the test setup, a tab is selected and a tab panel is visible

Joe_Humbert: I couldn't tell if this difference was intentional

IsaDC: You have to navigate to the first tab

Matt_King: What we could do to make these more equivalent is select the second tab and then place the focus on the "navigate forward" link

Matt_King: What I want feedback on my suggestion to changing the way the test is named

Matt_King: I suggested we change it to "Navigate backwards to a tab that is not selected". Would that make the intent of the test more clear?

Joe_Humbert: Sure. Most of the time you have the "navigate forward" and "navigate backwards" tests, the results are often very similar. Here, though, they were very different

Joe_Humbert: I also think test case 1 is a test case that would never happen. I've never seen a tab panel where a tab isn't selected

Matt_King: In issue 1280, there are two other recommendations related to naming. These will impact multiple tests (the name and content changes). All of tests 1 through 4 would need modifications to the wording. And test 1 would need modifications to the script

Matt_King: I think there are in general three changes to the tab plan. First, for all tests, move the "navigate forward from here" link to after the heading

Matt_King: the second change is for test 1 only: change the script so it selects the tab and sets the focus on the "navigate forward" link

Matt_King: The third is for the first four tests: change the name and the instructions

IsaDC: Got it

Joe_Humbert: I have a question about the implementation. When you navigate forward or backward, the tab panels are present even for the ones that are not selected. Is that intentional?

Matt_King: I believe those tab panels are styled with "display: none" (you can't see them visually), but VoiceOver reads them, anyway

Joe_Humbert: Is that the way it's supposed to be?

Matt_King: No!

Joe_Humbert: Well, I looked at the pattern, and it doesn't say anything about how they should be hidden in the pattern

Matt_King: It doesn't?

Joe_Humbert: No

Matt_King: Perhaps it's implicit

Joe_Humbert: aria-hidden is not mentioned, and it doesn't say that you need to hide them

James: It says that you display one at a time

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

Diagnostics

Succeeded: s/hem/them/

All speakers: Carmen, ChrisCuellar, dean, IsaDC, James, Joe_Humbert, jugglinmike, louis, Matt_King

Active on IRC: howard-e, Joe_Humbert, jugglinmike