Meeting minutes
New Issue Triage
jamesn: can james craig explain issue 1895
jcraig: needs to be an idl default that never changes and reflects the value
jcraig: this should be almost editorial
jcraig: it should be generated and/or cross referenced
jcraig: no pref regarding 1.3 or 1.4, but let's call it for 1.4 for now and we'll get to it
jamesn: next is 1893
jamesn: this is a dup.
jcraig: I filed this as an editorial change
jamesn: when it is capitalized it is used in a normative way. When it's not capitalized that's where we should reconsider
jcraig: doesn't make sense to me that we have a normative statement for something we might want to do.
jcraig: not certain this is a a dup but maybe it should be incorporated in the original
jamesn: next is 1892
jcraig: we can't have a "MUST" normative statement for ATs. It should be replaced by "SHOULD
jcraig: ATs MUST follow seems incorrect in this context
jamesn: if we say SHOULD it's not a requirement, right?
jamesn: there's duplication, we should remove this repetition
jcraig: feels like there's more than a triage discussion here
jamesn: let's discuss it with Valerie
jamesn: next, 1891
jcraig: seems reasonable
jamesn: I don't like when it feels like we're duplicating work
jamesn: that's it for triage
New PR Triage
jamesn: PR #1894 is completely editorial, right
jcraig: is that change removing that?
jamesn: no it's adding a default
jcraig: looks like the strong got removed
jamesn: ah, yes it removed the default
jcraig: I think this is related to the IDL changes it needs to reflect
jamesn: it doesn't help a developer to look at aria-atomic and see that the default is sometimes false
jamesn: this needs reviewers. James Craig, Valerie too, Scott too and myself
jamesn: next, PR 167 from Valerie
jamesn: enough reviewers assigned so no need to go over it in triage
F2F planning , F2F Survey
jamesn: survey currently has 8 responders, 7 attendees so far. IT would be great to have at least everybody who said in the previous survey they would attend to also confirm in this separate form.
jamesn: so folks can start shopping for flights if needed. 7 is probably enough but we want everybody
aria-hidden error case: disallow aria-hidden on root or document element (e.g. <html> or <body>) agendabot]
<jcraig> F2F Survey results https://
jamesn: aria hidden error case: we just want to confirm what we decided on this. We couldn't remember and the minutes weren't clear enough
jamesn: Scott created a PR to change the authoring requirements to not do this, right Scott?
scotto: right
jcraig: the reason this is an anti pattern is because the authors are doing it anyways
jcraig: I updated the issue very recently and thought it should be a no op
jcraig: it would be a no op that would reflect the value
jamesn: so this would be a core aam then
jcraig: I think so
jcraig: do we need another spec change and then another change to core aam?
jamesn: we could add an authors must not or something like that
scotto: my PR is actually both Authors and UA requirements
jamesn: your PRs doesn't have UA requirements right?
scotto: there's two PRs
scotto: there's an aria PR and jcraig is already reviewing it
jamesn: cool sorry I'm way behind
jamesn: so we just have a bunch of work to review things
scotto: need some help with the language on this PR
soctto: there probably is stuff that should go into one of the AAMs.
scotto: Not sure Core AAM is enough, HTML AAM might be an option too.
scotto: could also consider svg aam or math aam
Discuss how to break up the computedlabel tests in wpt/accname (to verify AccName computation) agendabot]
jcraig: the intial PR to WPT landed, it's the test driver update so these roles are possible
jcraig: it should make roles test much easier, like if you know HTML you should be able to do it
jcraig: right now the accessibility folder is just crash tests. There are other folders for roles and stuff and there's an accname folder and ideally we woudln't want to test it in a giant file but unlike the aria role tests it wasn't clear to me how to break this into files.
jcraig: I wanted to bounce ideas around with this group to find out how to break down these tests like I broke down...
<jcraig> https://
jcraig: if you look at the issues list there's #8 towards the bottom and I broke it down into fallback role tests, abstract roles tests, etc.
jcraig: please do comment on the issue with your ideas
jcraig: this guy from Google wrote a bunch of aria label tests for Chromium
jcraig: we could port those directly to WPT and all the other ones that are relevant for WPT
Need a decision on what the computedrole should return on extension specs
jcraig: right now, one of the things that's fallen out of this is... let me share my screen
jcraig: in PR 38845, if we look at the files changed, there's a difference between the graphics-document/object/symbols
jcraig: in graphics-roles.html we're returning the right role that matches the webkit implementation
jcraig: chrome is not cause they're returning the actual role string
jcraig: Chromium and WebKit have a different behaviors
BenBeaudry: Aaron and I will follow up offline
jamesn: what's the point of having this roles if WebKit doesn't return them
jcraig: the WebKit behavior comes from deep hub (?)
jcraig: either impl is okay if chrome wants to keep what they have
jamesn: what's the point of using these roles if we're exposing the super role anyway?
jamesn: we have people following up on this
Role parity: consider "native-" or "host-" role prefix for host language elements (including audio/video, but not limited to media) where custom ARIA implementations cannot (yet) match the native implementation's accessibility.
jamesn: we touched on this one briefly last week
jamesn: we talked about putting them in Core AAM
jamesn: Valerie was starting to work on a PR for this
<jamesn> w3c/
jamesn: what should it map to if there's no computed role?
scotto: generic is what it goes to. It's either the explicit aria role and the generic one
jcraig: tables don't always map to table
jamesn: I'm not sure if that's not even controversial. It might return a computed role of table
scotto: we should stop having heuristics and actually define how tables map
jcraig: most browser engines use heuristics in this case
jamesn: we're past the hour