Meeting minutes
New Issue Triage
spectranaut: two issues on ariaMixin from whatwg.
jnurthen: jcraig has been coordinating.
… but I think Anne will move it forward.
spectranaut: which milestone?
jnurthen: let's add it to agenda next week.
spectranaut: accname 175. custom elements to at
… same topic.
… aria 1840 extend rowheader / row
adam: seems like a TH.
jnurthen: can someone respond and ask for more info?
chlane: I can ask for clarification.
spectranaut: there is a proposal in the top card
… core-aam 151.
… assign to me.
… next: aria 1838
user agent MUST statements for 1.2
… 1838 was easy to write a test for. but 1837 is different
jnurthen: wasn't that in core-aam?
… 1.1
point 4.3
spectranaut: yes!
jnurthen: so this was just moved.
spectranaut: perfect.
jnurthen: the other section I think was also in core-aam
spectranaut: but for the future, we should think about tests for this
New PR Triage
spectranaut: aria 1839
… reviewers?
scotto: I was looking into testing
jnurthen: jcraig please. apple's input would be good.
adam: I also did a bit of testing.
… took the example with wilco, with a few more permutations.
… I only looked at macos, chrome, VO. I'll add a link
scotto: I'll share mine to compare notes.
Deep Dive planning
spectranaut: anything beyond accname discussion?
bryan: I didn't see any other responses.
spectranaut: Nov 3 or 10, 7pm Eastern.
aaron: no preference.
scotto: not today.
jnurthen: bryan, tell me when it's ready and I'll book the meeting.
AccName Role Traversal Proposal
[normative][author] clarify widget in trees can only be group and treeitem
pkra: need some guidance on what’s wanted here
… secondary actions proposal seems to imply widgets could exist in tree item
… so maybe we should wait until that concludes
jamesn: once secondary actions happens, we’ll need to update the text in many places, including this
pkra: original issue has to do with what can be contained in a treeitem
… I’ll have another look
scrib: pkra
Support aria-description
spectranaut: where are we with this?
scotto: I've merged the html-aam work.
… there's now just one section.
… this PR just need to link to that.
… unless we want to define all languages in core-aam.
aaron: I'm happy with scotto adjusting the PR
No mention of aria-placeholder in name/description calculation algorithm
bryan: this was never actually added to the spec.
… not sure where it would go.
or should go.
aaron: maybe we need a table where we check the accessible name.
… where the name comes from
jnurthen: wouldn't html-aam be the place for this stuff? Sometimes points to accname, sometimes just HTML.
scotto: we already have placeholder there.
… we could extend it.
… I didn't put it there because it feels odd.
… browsers implement is as a fallback.
jnurthen: I don't think it should be in accname.
chlane: it's not for the name.
scottto: right, it's a suggestion for someone writing input
… it's giant footnote.
siri: I agree, it's a significant problem.
jnurthen: I just don't think it should be
spectranaut: seems like agreement. should we just close the issue?
jnurthen: we could put a note somewhere that aria-placeholder is not here because it should not be.
bryan: but how does it fit into APIs?
aaron: chrome exposes it if there's no other name. otherwise, we expose it through other means of the platform.
<jamesn> aria-placeholder
<jamesn> MSAA + IAccessible2 Object Attribute: placeholder-text:<value>
<jamesn> UIA Property: HelpText: <value>
<jamesn> ATK/AT-SPI Object Attribute: placeholder-text:<value>
<jamesn> AX API Property: AXPlaceholderValue: <value>
spectranaut: core-aam has a mapping.
aaron: e.g., on windows/linux, we use object attribute placeholder (if there's a name).
jnurthen: aria-placeholder oder placeholder?
aaron: either.
jnurthen: ok, maybe needs to be there then.
… do you use placeholder when you're traversing to a child or just at the top level?
aaron: when we're coalescing stuff for the subtree, we'll use the placeholder if it became the accessible name.
… we'll also use the value of the input
scotto: people often misuse placeholder for labeling.
jnurthen: if there's value, nobody sees the placeholder.
aaron: if user knows what it is, it doesn't seem too problematic
chlane: relates to a PR I'm working on. value and name seem so often the same.
spectranaut: what to do next?
scotto: mention it in step of html-aam.
jnurthen: instead of aria-placeholder, could you point to placeholder attribute mapping in html-aam?
scotto: that's what I've done. Could add "see placeholder"
jnurthen: I just don't want to encourage it.
scotto: could add a note to the opposite effect.
bryan: so should it be in accname?
spectranaut: needs to be in core-aam
jnurthen: but aaron said they go down the tree, so it would need to be in accname.
bryan: right.
jnurthen: we're not very good at calling out the different things.
bryan: "in the absence of everything"
jnurthen: "in this order"
aaron: that's how I'd prefer the document.
… instead of prose, have more a table of precedence for the markup on a give note.
… I know it mixes HTML and ARIA but it would be the most useful.
jnurthen: could map back to the different parts of HTML-aam. Half-serious here.
scotto: there's definitely a lot I still want to do.
jnurthen: but we don't link back from accname.
… risk of getting into a loop.
spectranaut: scotto to continue.
spectranaut: can we land this before description changes?
jnurthen: I don't think we should.
scotto: I'll write some notes on the issue
Consider the future of aria-flowto
spectranaut: context: another issue was opened and we had asked to focus on this one.
jnurthen: flowto not working / not being useful has frustrated me, have run into good use cases.
… I would love to fix it.
… but it's unclear how much appetite there is for this.
aaron: getting AT to support aria-details. browsers was easy. AT, even when paying for it, takes forever. it's just very difficult for them.
… so supporting more use cases might be a better idea.
jcraig: I filed this in 2017 because even though we had some implementations, it was underspecified.
… basically, only linear keyboard, only forward.
… but we can have loops.
… go backwards etc.
… many to many relationship is not covered.
… I'd be happy to revisit some APIs related to this concept.
… I'd like the ability "this can flow to this other thing (incl. focus) and these are the relationship"
… controls or owns relationsship could be examples.
… when user hits hotkey, we can display these relationships.
… flowto was just too naive, a single idref pointer.
… devs weren't using it. implementations not thought out.
… but we got busy and brennan raised it earlier this year.
… after 1.5 years of little to no movement.
cyns: I remember there were use cases back then. ANyone know where to find them?
aaron: they were a bit handwavy.
… I think original was java accessibility API, fed into ATK.
… so we copied everything into aria.
jnurthen: diagrams, flowcharts, data visualzations are many
cyns: I think maybe power point wanted it
aaron: organizational chart could work with aria-details.
… that would be something AT are actually working on.
sarah: I wonder if pop-up work, grid/flex order, maybe they are addressing the original use case
… for flowcharts in my experience are not helped by flowto
<aaronlev> aria-details does support one to many, and ATs are working on better support already!
sarah: too broad / general for this use case.
<aaronlev> :) :) :)
sarah: ordering and popup might address the more general use case
jnurthen: flowto allows multiple idrefs
sarah: but how would that work?
jnurthen: wasn't saying it does.
… but you'd probably implement it naively. But piggybacking on details implementations might be a starting point.
cyns: graphics have very different structure. not saying flowto is helpful either.
scotto: use case flow charts, branching into multiple ones. right now we have to implement description text. and hoping you're using a keyboard.
… I'm in support of deprecating the feature. The support isn't there. aria-details is not yet fully implemented. So the counterpoints are not on point.
<Zakim> jcraig, you wanted to mention SVG connectors and to say Would rather drop flowto baggage and instead explore how to make something SVG Connectors accessible. http://
scotto: if there are still use cases left, something new might have a much better chance.
aaron: we could expose flowto as details. we have a hint "details from" on some platforms. we could say "details from flowto". then existing AT could piggypack on that
… they could change their wording to users.
jcraig: still a 1-1 unidirectional relationship
aaron: flowto supports many, details too.
<Zakim> jcraig, you wanted to mention SVG connectors and to say Would rather drop flowto baggage and instead explore how to make something SVG Connectors accessible.http://
jcraig: nothing in there "details for reason X"
aaron: we have an object attribute on ia2, atk to indicate that. could be done on mac
jcraig: but can the author influence?
aaron: e.g., via role
jcraig: I wanted to mention SVG connectors. I don't know the implementation status but it has
… many-to-many relationships
… If we took a pattern like flowcharts that are easy to understand and made that accessible, we may be able to re-use that pattern for other non-flow-chart use cases: For example, all the use cases of flowto, details, etc.
… not just that they're connected but text alternatives for the relationship itself
… e.g., transitions.
… that would work in a much broader sense than flowto.
… outside of flowcharts.
… easy idea to understand, make that accessible.
… could benefit widely.
jnurthen: that sound interesting
jcraig: there's more than flowto and details here.
aaron: details takes the approach: in the forward direction, backwards automatic by the browser.
jcraig: multi-to-multi, how would the user know which reverse relationship to choose when there are multiple? The relationship itself is unlabeled and ambiguous.
… using a pattern like shift+tab, the AT could present the multiple reverse relationship (incoming to the current element) but there is no author-provided label on the relationship to clarify that connection.
spectranaut: we're at the hour.
… can we have a comment on the issue?
<BGaraventa> here is an example of flowto https://
pkra: deep dive?
bryan: pasted link to example for footnotes with flowto.
… works with jaws.
… just an example of what works.
jcraig: If we have a deep dive on the topic, it should include stakeholders on the GitHub issue, like @BrennanYoung and @WeedySeaDragon
RRS agent, make minutes