Meeting minutes
New PR Triage
<github-bot> I can't comment on that because it doesn't look like a github issue to me.
spectranaut_: Editorial PR for colspan and rowspan already reviewed by Scott. We need one more
pkra: I can take a look
katez: Scott mentioned thaat having HTML was sufficient
… I wasn't sure what kind of linking we wanted there
spectranaut_: That's probably fine, no strong convention on this
pkra: Agree with Scott, better to use this for something else, I'll take a look
jcraig: First paragraph has "also see ..." for some of those, which was a deliberate choice
pkra: "See related".
Jcraig: If it's relevant to be put in the related concepts table, it may also be relevant to put on that paragraph
spectranaut_: From Sarah, secondary action
Sarah: It's on the agenda
spectranaut_: We have a lot of items, anything to get this unblcked?
Sarah: People who are involved probably are aware laready
Matt_King: I saw this and added my name
Jcraig: Please put me in
Jacques: Me too
Jcraig: It may be worth to breif the group on this
Sarah: Consensus is that focus event buble on the actions but accessibility focus events are not fired, so ATs shouldn't move cursor to the action
… DOM focus bounces but there are no AT signals
Jcraig: Not just accessibility system eventss, it's any system event. All focus bouncing stays in the browser
spectranaut_: Should we add Jammie to this?
Jcraig: And Tyler
Sarah: Good idea
spectranaut_: If he's not in the group he should p robably join
Jcraig: I'll check
spectranaut_: Two PRs against Math-ML. Anyone interested?
spectranaut_: I could take a look
Jcraig: I'll review
WPT Open PRs
<github-bot> I can't comment on that because it doesn't look like a github issue to me.
spectranaut_: No new ones other than Theo's, which I need to still take a look at
Theo: Not waiting for much here. I'm still getting a weird testing failure.
spectranaut_: This is a known bug in Firefox
… We can ask someone to override it
Deep Dive planning
<github-bot> I can't comment on that because it doesn't look like a github issue to me.
spectranaut_: Are you working on the posinset?
Matt_King: I thought you and James were
JamesN: I'll put it back on my todo list. who should be involved?
spectranaut_: There's a list in the issue
Matt_King: I expect to come out of the deep dive with some discreet approaches to solving the problem
Add explicit language and direction metadata to AriaNotificationOptions
spectranaut_: Two issues from i18n HR reviews. Thanks to Sarah and Peter for responding to those
… I wanted to check in. Sarah, is there something we can do to support you Sarah?
Jaques: ARIA Notify handles it by design, it uses the DOM or the documents language. We didn't want to have it as part of the JAvaScript
Daniel: I'll take this offline with the WG Staff Contact
pkra: Is it a concern for us what happens if someone uses an automated translation service and then the service changes the node language but the script doesn't take that into account?
Sarah: I'd have to check, I don't think these overwrite the language of an empty nodde
pkra: But probably to the document
Jacques: Are you saying the translators are editing the lang attribute on the element?
pkra: I was assuming that a translation tool would change both content and language
Jacques: If they change the lang attribute I guess that would be a bit of an issue
Jcraig: The lang attribute is on the html body, the aria-notify string is coming from wherever the javascript takes that, if the restof the page is translated into English then it's not going to be as useful to you to receive this notification in Chinese. Is there a path through the aria-notify implementation that would pick up that process?
… I think this is something we could include in the AAM to clarify that if you are in autotranslate omde this is an autotranslating string
pkra: It's not a new problem, this already happened with a live region
… But it makes it worse now because there's not a way to translate it back later
Jcraig: Good point too. I haven't run tests to see what happens if you run a live region after a translation
pkra: I also don't know how many automatic updates such tools would do
JamesN: Worth an experiment to contrast with dialogs and others which would be translated based on event listeners
spectranaut_: Should we need to clarify the changes we are proposing?
Jcraig: Regardless of the result of Sarah's tests, I think we need an issue in Core-AAM
… There should be an aria-notify section in core-aam
Jacques: How do we know if we are in "translation mode"?
Jcraig: I am only talking about browsers built-in functions.
Translatable Attributes
spectranaut_: Peter, you said you'd work on a draft for this
pkra: I did say that, but I also said that what was requested wouldn't really work
… What they suggest is that an element's language and directionality is different than it's accname and description, which is weird to me. And also the suggestion around using aria-describedby or details is also weird. I think there's a misunderstanding, and would probably cause a WCAG failure
… I do see a note suggesting aria-labelledby for translations, but nothing else really
spectranaut_: Do you waant to make that note?
pkra: I'll make a small note ,but I am not sure if it will be sufficient
Filtering a listbox with a text field (filterable select)
jarhar: I am working on an HTML filterable select element
<HaTheo> people tend to call this wordwheeling lists if that's helpful.
jarhar: I wonder what's the best way to represent this iwth accessibility roles
… Based on previous conversations it seems that the button that opens the picker should be a button if it's always the same or a combo box if it contains values
… the picker should be @@@
… And the text field inside should be a text field that controls the list box
… Curious if anyone has feedback
Matt_King: aGree with Scott about the github example not being a combo box
… This highlights the combo box constraints
… Combo box are single-select, but I don't see anything in the issue that says if it's always single-select or always multi-select
jarhar: It's designed to be multi-select or single select
Matt_King: We'd have to adjust accordingly
Sarah: If you want to build this with existing ARIA roles and patterns you could
… You could use menu and menuitem check box for the pop ups
… That works well for multi-selects
… Keeping the filter as a text box and then putting activedescentand pointing to the list items
… All of this is not quite per spec, but it works
jarhar: I am taking notes on the feedback issue
Sarah: We've used it some times and got great feedback
jarhar: Sounds something worth looking into
Jane: We were looking at reopening the discussion of issues on the ARIA spec that Sarah just highlighted because we've encountered some similar issues with thiss now that this is turning into a pretty common pattern on the web
… In addition to supporting muitl-select and single-select, is the search box only a search box or it can also be an input field?
… Common example would be a tagging interface. Text box at the top as you are typing a tag that shows the matching values but also allows to create a new tag
jarhar: Dynamic loading of options is an important use case
… Adding options based on what you typed, I haven't look in detail at examples where you would want to add something new
Jane: Some patterns to be aware. If you are typing anything the top or bottom item is Add" plus the text you wrote. You select it as another option
… The other thing we've seen is a separate button to add what you just typed
… It is prefferrable to put the add option at the top, especially when there are many items on the list
… Question for the group is if we want to open a separte issue that helps improving the ARIA spec based on common experience
what Matt_King: I'd like to learn the benefits of using menus over multi-select list boxes are
… We should figure out what is stopping them from working
… The GitHub example is cumbersome to use with the keyboard, we could go through some of this on the deep dive
… Could we take an example ane enumerate the problems? And where we should be doing it?
jarhar: Please do in this very issue
spectranaut_: We should wait for scheduling the deep dive until Matt gets the feedback and jarhar processes feedback
… We use to have these one hour before this meeting, or we could schedule them later for Jamie to be able to participate
… What about in two weeks?
Keith: I'll be out
<HaTheo> I am down with any time, having owned a number of these controls I have opinions and research. Though I feel like it might echo others opinions.
Jane: I also won't be available
Matt_King: No problem on my end
spectranaut_: I'll schedule it after this meeting
Text nodes and Allowed Accessibility Child Roles
giacomo-petri: We are allowing the author to include empty structures or generric interveening, but we are not defining the rules for text.
… Borwsers are defining their own
… To me the example I provided is not allowed. Should weclarify that aspect that text nodes are somehow not allowed in those contexts?
Jane: IF we define that in the spec, does it mean that if an author renders them the information wouldn't be presented by AT? I think this could potentially create a worse user experience, that sighted users are going to see that but not screen reader users
spectranaut_: This is about users must, it's an author issue
JAnes: Only validators?
giacomo-petri: Yes
Jane: Yes, then it does make sense, having validators raise it would be good
spectranaut_: Sounds like it's a but, sounds to me like it should be surface
… Maybe we should start talking about plain text
pkra: If I go by the example, we have allowed accessibility child roles, which tell waht is permisible
… Then we talk about children of these elements as objects in the accessibility tree
JAnes: If the text node is not allowed would it also not be allowed in situations where it should be?
pkra: There are situations where we don't limit the accessibility children
giacomo-petri: My problem is with the word "role", which is note defined
… It could be ambiguous for implementers of testing tools
Matt_King: Is the issue that the spec doesn't do a good job of saying that the author made a mistake?
giacomo-petri: Yes
Matt_King: It's obvious to me this example is a mistake, but you're saying it wouldn't be so obvious by just reading the spec
… It's just a wording issue, what children means here
pkra: The simplest one would be to talks about accessibility children
Jane: There should be a note
rrsagent ,draft minutes