css-meeting-bot
commented
1 month ago The ARIA-AT Community Group just discussed Issue 1144: Priority of posinset and setsize assertions in test 5 "Open a menu to the last item".
The full IRC log of that discussion
<jugglinmike> subtopic: Issue 1144: Priority of posinset and setsize assertions in test 5 "Open a menu to the last item"<jugglinmike> github: https://github.com/w3c/aria-at/issues/1144
<jugglinmike> Matt_King: Apple does not think the assertion in question should be a "must" or a "should". They believe it should be fully optional
<jugglinmike> Matt_King: I've been giving this a lot of thought, and I actually was just about to make a change to the test plan, but the community group needs to weigh in, first
<jugglinmike> Matt_King: Do folks think screen readers "should" exhibit this behavior? Or just that they "may"?
<jugglinmike> Matt_King: I think the default would be toward consensus among vendors. In this case, the most straightforward path to consensus is "may"
<jugglinmike> Matt_King: In the action menu button, we have a test that opens the menu. Once the menu is open and focus is on the first item, you press the "end" key or the "arrow" key to move to the final item. The question is, when you press one of those keys, should it say "action 4" or "action 4, 4 of 4"?
<jugglinmike> Matt_King: Apple doesn't think that VoiceOver "should" convey the position ("4 of 4" in this case)
<jugglinmike> Michael_Fairchild: Did Apple share their rationale?
<jugglinmike> Matt_King: They shared their rationale in a comment on the issue
<jugglinmike> Matt_King: For context, VoiceOver does do this for tabs, radio buttons in a radio group, and items in a list (which they sometimes call a "table")
<jugglinmike> Matt_King: VoiceOver does not do this in folders or menus. Additionally, in their verbosity settings, they don't offer a way of changing this behavior
<jugglinmike> Matt_King: VoiceOver does allow users to toggle announcement of "name", "status", and "type"
<jugglinmike> Michael_Fairchild: My initial reaction is in line with Matt_King--that it makes more sense to convey it than to not convey it. But I'm still processing...
<jugglinmike> James_Scholes: The implication that the information shouldn't be conveyed because it isn't onscreen makes me uncomfortable. I'm sure we can all imagine a huge number of cases where a screen reader conveys information that isn't immediately present in visual form
<jugglinmike> James_Scholes: A sighted user certainly can see that a given collection has a large number or small number of items
<jugglinmike> James_Scholes: There's also an argument to be made for consistency
<jugglinmike> James_Scholes: And also weighing on my judgment is the fact that they don't let you opt in or opt out of receiving this information
<jugglinmike> James_Scholes: To be honest, I disable this kind of information in my screen reader of choice. I'm partly speaking for the many screen readers who I know rely on this information as essential to their way-finding
<jugglinmike> Michael_Fairchild: the more I'm thinking through it, the more I agree. Visually, you have a sense of the size of a menu and your relative position within the menu
<jugglinmike> Michael_Fairchild: Also, whenever you loop around, there could be some ambiguity as to whether the looping occurred--especially with longer lists
<jugglinmike> Michael_Fairchild: I think it makes more sense to convey it than not
<jugglinmike> Michael_Fairchild: Would it be sufficient to say that there should be a mechanism to convey something like this?
<jugglinmike> Matt_King: We don't have any testing related to what users could do by changing settings. That would be a level of complexity that I don't think we would want to add to the project
<jugglinmike> Matt_King: Stepping back a bit, I've been observing lately some of the conversations we've been having with Vispero about "should" versus "may".
<jugglinmike> Matt_King: It seems more and more common that leaning toward "may" for things that are... If you were to build a brand-new screen reader from the ground-up, and you had to prioritize functionality toward things that are going to enable people to work with your software. That's what "must" is about
<jugglinmike> Matt_King: I think Vispero is starting to regard "should" as a reasonably high bar, as well, because ATs get dinged for not satisfying those
<jugglinmike> Matt_King: So we have to be fairly confident when we use "should" that failing to implement the behavior actually degrades the experience in a meaningful way
<jugglinmike> James_Scholes: I think many people would argue that the high bar is met by wayfinding information. Cena is one
<jugglinmike> James_Scholes: I don't believe Vispero has raised any concerns about this particular information
<jugglinmike> Matt_King: They have not
<jugglinmike> Matt_King: When you're moving through the menu, I think we're aligned on the information being a "should"
<jugglinmike> Matt_King: Not getting this information remains the hardest thing for me as a VoiceOver user on a Mac
<jugglinmike> Matt_King: When I go to the desktop on the Mac, the fact that I have no idea how many icons are on the desktop (especially when there are many)--that's really hard!
<jugglinmike> q
<jugglinmike> James_Scholes: You gave the hypothetical about if you were building a new screen reader. Well, PAC did build one, and we never even considered not including this information in menus
<jugglinmike> ChrisCuellar: Maybe "should" could be conveyed as more of an interop nudge
<jugglinmike> ChrisCuellar: If all other vendors agree that this is a "should" and we can capture sentiment that end-users find this really valuable, then that becomes part of advocacy for interoperability--about one implementation being a bit of an outlier
<jugglinmike> Matt_King: I started drafting a comment to James Craig to request clarification on his words about what appears on the screen, what doesn't appear on the screen, and what ARIA has to say about this
<jugglinmike> Matt_King: I could also ask about the data behind the design of conveying this information in some circumstances but not other circumstances
<jugglinmike> Matt_King: This would also be a good topic to include in the agenda for my next meeting with Vispero
<jugglinmike> James_Scholes: We've had many discussions of this sort over the years, and the CG has historically been willing to compromise. That doesn't appear to be the case for this issue
Per feedback from Apple in #1058, consider making posinset and setsize assertions optional.
Test Setup