[go: up one dir, main page]
More Web Proxy on the site http://driver.im/

W3C

– DRAFT –
ARIA WG

04 November 2021

Attendees

Present
jamesn, Joanmarie_Diggs, MarkMcCarthy, pkra, sarah_higley, spectranaut, StefanS
Regrets
-
Chair
JamesNurthen
Scribe
spectranaut

Meeting minutes

[New Issue Triage](https://bit.ly/3CH8cfG)

https://github.com/w3c/aria/issues/1636

jamesn: let's leave this to simmer for a bit and discuss next week

jamesn: https://github.com/w3c/aria/issues/1634 move to practices?

jamesn: https://github.com/w3c/accname/issues/146 agenda+ for next week.

https://github.com/w3c/aria/issues/1632

jamesn: why not support on role="group"?

aaron: seems like it get revoked, maybe not intentionally?

aaron: fieldset is a group, and fieldset you can still disable

stefan: original idea declare an entire group, all children get disabled. old discussion

jamesn: milestone 1.3? anyone want to be assigned to determine a proposal?

sarah_higley: scott already has a PR for it

jamesn: is this a deuplicate?

sarah_higley: related to aria-disable 1130 issue...

jamesn: milestone 1.3, lets revist after the other issue is resolved.

https://github.com/w3c/aria/issues/1630

jamesn: spec or practices?

sarah_higley: we just moved an issue over to APG related to overflow tabs, seems related to this and secondary tabs.

sarah_higley: you can assign to me

jamesn: milestone 1.3 not to lose

https://github.com/w3c/aria/issues/1629

scott: you can assign this issue to me, but not a 1.3 issue....

jamesn: does this need normative changes?

scott: no

jamesn: maybe brennan can do it?

jamesn: let's ask him

jamesn: I won't assign it to you yet, but we can come back and assign you later on if need be

[New PR Triage](https://bit.ly/3wcnVRy)

jamesn: https://github.com/w3c/aria/pull/1635 ready to merge

https://github.com/w3c/html-aam/pull/354

https://github.com/w3c/html-aam/pull/350 needs review

scott: will need implementation when lands. no aria-spec impact, just html

jamesn: we should have implementation review

jcraig: I'll review

aaron: I'll review

joanie: I'll review

[Deep Dive planning](https://bit.ly/aria-meaty-topic-candidates)

jamesn: next week I can't do the deep dive

jamesn: anyone else on holiday?

jamesn: I guess we should meet next week!

jamesn: because we have so much to do, but no deep dive. if you have a topic for the following week, please send

[Updating ARIA 1.2 due to IDL implementations (exit and re-enter CR?)](https://github.com/w3c/aria/issues/1598)

<jamesn> https://github.com/w3c/aria/pull/1633

jamesn: I created a pr based on our discussions during the meeting last week

jamesn: I would like more reviews

jamesn: we have no implementation in a shipping spec of IDL

jcraig: only in the browsers

jamesn: can aaron ping dominic?

aaronlev: yes

jamesn: I'll try pining dominic and anna in the thread?

jamesn: this is the most urgent thing we are doing, it is the blocker for 1.2, I would appreciate as much review as possible on this!!!!

[A role for indicating whether a given ruby represents phonetics ](https://github.com/w3c/aria/issues/1620)

jamesn: is there still work to do on this?

jcraig: seems like he is open to us closing it

jamesn: can I leave it to you and aaronlev to finalize these issues and close it out?

jcraig: I'm going to close issues 1619 and 1620, and if we need to reopen we can

[Updated spec text to reflect the processing of hidden elements when c…](https://github.com/w3c/accname/pull/137)

aaronlev: I feel like the innertext was a nice simplification until we realized there were too many details

aaronlev: firefox doesn't implement it quite correctly

jamesn: having inconsistent implementations.... is a problem

jamesn: is your solution for chrome a solution other people could use?

aaronlev: I think so, you can make an accessible object for a dom node even it is not in the final platform tree

aaronlev: it might be harder in other browsers

aaronlev: because of the number of accesisbility APIs we support, with only one tree in blink, and we have to send from render to browser process... we have ways of skipping nodes in output

aaronlev: when we are computing name we can use fuller tree

aaronlev: I think firefox will have to move to that model as well

jamesn: does the fact that we don't have consistent implementations mean we can get to CR, or it's ok because it has always been broken?

aaronlev: chrome, we can do it based on the spec now or based on some different innertext thing

aaronlev: probably firefox and webkit can support current spec, but its probably harder

aaronlev: we are flexible if you want to do something else

<Zakim> joanie, you wanted to ask what exactly what the spec says now means

jamesn: currently everything is broken in a different way

joanie: what do you mean when you say what you implemented what the spec says... I think the spec says don't go down the tree!

joanie: apple does what the spec currently confusingly says

joanie: the spec is in accname, if the div is hidden, it's innertext, but any children that are not directly references by the idea of aria-labelled by, they are not included

jcraig: brian wrote the first version of the spec change, I clarified, pull 137 -- if chrome is doing that specifically, but this PR mentions innerText, chrome is not using innerText?

jcraig: aaronlev can you reply with PR changes to match your implementation...?

<pkra> sorry, have to drop off.

matt_king: joanie, your summary... I thought the whole purpose was for this to be understandable by users, web authors who are trying to make accessible labels. my interpretation of what aaronlev was saying is that chrome is doing what people thought the spec is meant to say. What the spec actually says is bizarre and unexpected.

matt_king: I thought the point was to clear up and include the tree

aaronlev: you might want to have a hidden label or description..

bryan: .... a real world example I couldn't catch!

matt_king: good example

matt_king: is there a concern with the new proposal?

jcraig: if part of the label is displayed, and part not displayed, how do we computer? Once the ancestor is not displayed, we don't care why the sub level elements are not displayed.

jcraig: I don't think we should exposed those substrings willy nilly

jcraig: what is the PR change we can review, the devil is in the details, I can't tell you what I think until we see the PR

aaronlev: I don't think the double hidden is a real case. if people used aria-label... probably because it's in the dom already. if you have double hidden, you can expect predictable behavoir

sarah_higley: we reference tool tips via labelled-by or described by, it's hidden until you have focus, but we want the labelledby or describedby to be consistent. tooltips contain whatever text you want, some of it is display none, the displaynone in the displaynone tool tip I would expect as an author for it to NOT be used

matt_king: I can't imagine a display-none in a display-none

sarah_higley: it might be aria-hidden....

matt_king: I can understand that

jcraig: sometimes password fields have a bunch of rules, but only the unfullfilled criteria will be visible

stefan: is there a general idea of how to handle these nested situations generally???

bryan: in the case where it is area hidden... it doesn't matter what is on the inside

aaronlev: the problem with display none is that there is implementation problems. the browser will do many optimizations.

aaronlev: a lot of styling is not applied, its hard to have a special rule about display-none subtrees

aaronlev: if people want hidden sub trees in display-none... they would have to use aria-hidden, every other way of hiding something would not have an impact

jamesn: sounds feasible

jamesn: while we may have one user agent that works, we don't have consistency, so authors who are using this stuff are already broken. I don't care about breaking authors who are already broken in a slightly different way.

jcraig: aaronlev mentioned the only thing that can hid substrings in hidden elements is "aria-hidden=false", I don't think this is a fair expectation for web authors

jcraig: if you landed on the tool tip, and it is displayed, then you follow the expectations of hidden vs not hidden. if you land on the tool tip and it is NOT displayed for some reason, the aria-describedby is going to follow different rules

jamesn: displayed vs not displayed should not change the value

matt_king: I was going to say that might not be the best example.

jamesn: there might not be much to gain with this discussion. someone needs to come up with another example/fix for us to change our mind on the current change proposed

jcraig: if we make this change, we have three working implementations. aaronlev did point out some edge cases, I would propose that if that is the case, then we should point out that case in the implementation of innerText itself. it if it is the wrong value for accessibility, then maybe it's wrong for the web generally

jamesn: I doubt innertext will change for this

<aaronlev> maybe a deep dive topic?

<jcraig> +1

bryan: I will put example in this PR

<jamesn> https://github.com/w3c/aria/issues/1628

<Zakim> jcraig, you wanted to defend innertext

[Clarify how "otherwise interactive" relates to overriding the none/presentation role](https://github.com/w3c/aria/issues/1628)

jcraig: one of the webkit implementers raised this

jcraig: I think this means the spec is right but could use some clarification, and all the implementations are wrong in this case,

jcraig: if this element is interactive (including if its focusable), then the presentation role should be ignored.

jcraig: onClick should make something interactive and none role should be ignored?

jcraig: what about drag events?

jcraig: should we take delegation into consideration? like click event on body? everything can't be excluded from presentation

jamesn: lets discuss in the issue!

RRSAgent: make minutes

Minutes manually created (not a transcript), formatted by scribe.perl version 136 (Thu May 27 13:50:24 2021 UTC).

Diagnostics

Succeeded: s/the issue/issues 1619 and 1620/

Succeeded: s/matt_kind: good/matt_king: good

Succeeded: s/its self/itself/

Maybe present: aaron, aaronlev, bryan, jcraig, joanie, matt_king, RRSAgent, scott, stefan