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

W3C

– DRAFT –
ARIA WG

16 May 2024

Attendees

Present
Adam_Page, BGaraventa, CoryJoseph, Daniel, Francis_Storr, katez, keithamus, knights, pkra, sarah, Summer, Theo-MSFT
Regrets
-
Chair
-
Scribe
knights

Meeting minutes

I can scribe

New Issue Triage

spectranaut_: The first issue from Scott
… this came out of editor's meeting, triaging old issues and there's an old issue about radio group. This is a broader issue.
… Issue on ARIA #2180 consider loosening requirements for naming certain container elements
… the next issue on html-aam
… Scott already responded
… Does this need further discussion?

keithamus: I can poke at it a little because the reasoning around the other elements is kind of interesting it's something that is regularly authored
… title and meta generally live in the head; I understand that slot is designed to be equivalent of that
… feels weird to prohibit role on ti
… Don't know if this is something this group necessarily needs to champion alleviating that
… or something that HTML should be doing

spectranaut_: Is this something we should talk about in ARIA WG?
… I don't use slot as an element, it probably made more sense in the past if slot was treated as generic instead of nothing space as it's supposed to be

^ scotto

scotto: Need explicit parent and child relationships

scotto: Probably not on us, or could say to HTML to treat this as generic or none
… slot isn't supposed to be used by authors, don't put roles on it, has to be a really good reason to undo that

keithamus: It's totally justified
… what is the use case we would need on slot that couldn't be solved by having the slot going into the appropriate container element
… what problem is this solving to undo intended use of element?

keithamus: I had stickiness around tab list
… and trying to find intersection of API where we have a tab list element, use a slot to provide some of it but otherwise does the sensible thing
… can I customise this and that
… got a bit tricky to manage that
… someone said they wanted to customise this bit and turn it into a slot; a subtle failure carried over and difficult to refactor with all the feature requests
… awkward experience generally; assignment failure so browsers can do more to say don't put roles in slots

spectranaut_: sounds like there's more to discuss
… next item, HTML 545

scotto: This follows from our CSS AM conversation the other day
… we want to spec out the new CSS anchor positioning property to come with defaults
… we don't have spec to put that in until a CSS AIM
… maybe just goes into HTML AIM
… or HTML AIM becomes HTML and CSS AIM
… we need to track it and put it into this spec until there's a better place for it

spectranaut_:

Put an agenda on it^

New PR Triage

spectranaut_:

This is just an editorial for MathML

WPT Open PRs

spectranaut_: There are no open/new ones this week

spectranaut_: About the deep dive, we have one scheduled for next week, 23rd
… aria-notify update
… there's an event in the calendar
… is there any other deep dives to plan?
… James is out for the next month

Deep Dive planning

Thank you!

WAI-ARIA 1.1: Cell and gridcell

spectranaut_: Next issue is an old one, #79 in aria repo
… cell and gridcell issue
… there was a decision to deprecate gridcell once they are treated equivalently
… is there really a need to deprecate, or tell authors to use cell or gridcell within the context of a grid
… how does the group feel about that
… silence
… agreement?

<cyns> seems fine!

scotto: Silence is roaring agreement

scotto: Agree to do that

And make it clear in the spec

If you do use a cell in a gridcell, there are attributes available to you but not if use a cell outside of a table
… be clear in the spec on that

@jamesn Do we have the same attributes to cell and gridcell right now?

scotto: No
… gridcell for instance allows for aria-selected, but cell does not
… that's why it would be quicker to just say in a definition that roles are interchangeable
… make it clear you can't put aria-selected on a cell in a table

jamesn: sounds like some work to do around that

scotto: Interested in reviewing it, not doing the work

thinkbulecount2

spectranaut_: Moving on next item

ARIA should clarify distinction between “contents exposed as descendant text nodes” vs “name from contents” (Was: listitem should include name from Contents or name prohibited) and -> td, th naming doesn't align with ARIA

spectranaut_: jcraig was on a meeting last week and we decided to discuss larger issue
… does anyone want to champion the discussion
… this is #2160 aria clarify distinction

jamesn: I think we should punt it til James is back

Document interop of misspelled aria-labeledby and its conflict resolution

spectranaut_: #2093 aria-labeledby misspelled
… we did ask Rahim, he is not on the call but we can discuss this
… asked for a few more tests of order precedence of the two versions of aria-labelledby
… non-standard spelling tests pass all of browsers except Firefox since it doesn't support this non-standard spelling with the labeledby

jamesn: I think it does support, just that the preference/order is different

spectranaut_: Looks like Firefox fails on div with role group is labelled by ARIA by non-standard spelling

jamesn: I thought all engines support it

spectranaut_: Looks like it doesn't support it tho

keithamus: is it flagged?
… is it within Firefox patched but under feature flag

spectranaut_: I think it would be linked to an issue but I don't see one

spectranaut_: Firefox's implementation is either off or very incomplete
… not supported in Firefox, but other tests important to look at
… link labelled by aria-labelledby supercedes aria-labeledby

jamesn: I'm interested in a test where they reverse the order

scotto: Rahim did comment that he added more tests
… It does seem that there needs to be a decision
… we either treat aria-labeledby the same as the correct spelling
… essentially just a synonym
… these tests show that there is an order of preference where correct spelling takes precedent over the other
… in this case can't say it's a synonym, its first instance should be used, additional instance ignored -- consistently
… we need to make browsers fix it so it's consistent, or this spec needs to be updated to duplicate aria-labelledby and add this additional spelling with order of precedence

spectranaut_: No one on queue?

jamesn: I hate this
… don't we have better things to do than fix author errors?

spectranaut_: What is the easiest way forward in terms of documenting and fixing?
… aria-labeledby is fallback
… and movve on

scotto: doesn't resolve the dual declaration because it happens

spectranaut_: In terms of browser implementation it's easier than keeping track of order by the time you process these attributes

keithamus: the parser will collect attributes as it hits them I think
… I agree with your solution though

spectranaut_: Isn't the case that browsers already implemented what I discussed?
… if aria-labelled is after aria-labeled, it's still used by Chrome and Safari that implement this spelling

scotto: That's a different way as a path forward
… there's an order of precedence where the correct spelling is higher in the algorithm
… where are people going to look for it

cyns: important that it's consistent behaviour across browsers

sarah:

Do we have to do anything about this? ^
… I can misspell aria-labelledby a number of ways
… not out of real of author mistakes
… just let browser handle error as they want to

<Theo-MSFT> Why is this not just caught in linters? Why are we fixing misspellings in the spec?

cyns: Americans hate the way it's spelled

pkra: If people think this should be a synonym
… this is invalid aria syntax, but otherwise, just make it a synonym

<Theo-MSFT> I was saying the same thing as Sarah, Peter above.

sarah: Much less confusing to have one way of doing things and not 2

<cyns> +1 to synonym

keithamus: Totally valid to pick one even if it's not ideal spelling, problem is we have precedence today
… there's inconsistency in browsers
… maybe we need to determine that instead
… discussion to Chrome/Webkit to unchip labelled by one l
… or we should ratify the semantics of what it is
… neither here nor there but that is the conclusion we need to come to in this group
… until someone specifies it, we just have to make a call and tell the browsers

cyns: What if we say there are synonyms

[missed]

jamesn: What about the IDL? Synonym with the IDL as well?

<cyns> synonyms, but if they point to different values THAT's an author error

jamesn: I think this is going to set up a whole world of hurt down the line

keithamus: Having undefined behaviour in my opinion sets the scenario that the browser will defend the behaviour
… we can define not to have this; unchip labeledby
… if we go ahead and say it's undefined, it probably means browsers will copy Chrome
… we can totally undefined things if we declare it as an author error and have Chrome do a warning to say it's misspelled

sarah: I like what keithamus said, have it documented as an author error
… I don't like synonyms
… We already have a number of those things on the spec

<Theo-MSFT> +1 to Sarahs point about synonyms.

sarah: I really want to avoid introducing a misspelling it makes it harder

cyns: I don't know if it's semi-kosher or misspelling
… there are two ways that English spells this word and would be nice not to have this situation
… I don't like it being undefined but don't like requiring the UK spelling

scotto: I think we're stuck with this
… I don't see any browser unshipping this
… for this, I can imagine far more problematic if overnight controls that did have accessible names no longer do

<cyns> +1

scotto: browsers won't unship this

spectranaut_: There are three proposals, one is to reverse this trend
… some advocates against
… other proposals: adding a synonym
… the other is adding a special case
… correct spelling preferred over the other
… anyone against?

scotto: I don't have strong opinions of either, if it's just noted that aria-labelledby is the proper and labeledby is an author error that is respected, I think there needs to be a note in every applicable spec to make people aware of that order of preference
… if aria just wanted to say it's supported but authors are wrong, I'd be fine with that

sarah: I have nothing against defining the browser fallback
… just not the synonym in the spec

CoryJoseph: I kind of like what is being said
… we define what is the correct spelling but can use this other spelling but it is out of spec

keithamus: My preference would be, let's try to unship this but how deep is the problem
… having it as a synonym seems to be giving us a load of work for little gain

jamesn: I think that the general consensu is to put it in author handling errors section as a special case

spectranaut_: I was going to suggest as well
… to reply to keithamus , not sure if we can unship if it's being used

keithamus: We kind of can
… we can see just what the usage is
… provide things like console warnings
… push out to axe to tell you that this is a failure
… to the point where browsers consider removing it as an attribute

cyns: I'm strongly opposed to doing that

scotto: I hear what you're saying keithamus, but I worry
… some data is better than no data

<Theo-MSFT> what do we do for other areas like Color/Colour?

scotto: everytime I see collection of data, it always skews away from accessibility improvements

scotto: You can't get the data from behind logins, for all the heavy apps that have aria spackled on, we don't get those numbers
… it becomes a big issue, data can't indicate how big of a problem it could be, and when it does come, it will take people by surprise

cyns: The work is happening

keithamus: I think that's the takeaway, if we're not interested in unshipping it, we should be interested in specifying it precisely the mechanics for which it operates

spectranaut_: There seems to be a general move forward with author error
… anyone interested in specifying this

keithamus: Yeah, sure

Values table for aria-multiselectable/aria-current are missing "undefined"

spectranaut_: It seems this is also from Rahim
… straightforward solution or?
… should we look at how this is implemented?
… work is being done by Rahim to correct this language
… anyone see reason not to do this change?

sarah: I'm not sure
… False and undefined are equivalent
… can I review it?

spectranaut_: You can take a closer look

Clarify "undefined" assignment for a state/property as the value (not the string)

sarah: I've read this recently
… there are places in the spec where undefined is in quotations, instead of actual value of undefined
… the spec is ambiguous and misleading
… implies that the string undefined is equivalent to the value undefined
… we should fix it if it's not equivalent, or be more clear

sarah: The string undefined is not equivalent, I find
… you can try that test page on different browsers and got the same results
… string undefined is defined as a string
… it's different on Linux
… Linux is doing things differently but Windows and Mac is undefined

spectranaut_: Confused why it would be different on Linux
… so sounds like the string undefined should not be considered undefined

sarah: Everything seems aligned and doing the same thing that string undefined is not the same as the value undefined
… if anyone else thinks we shouldn't go there

spectranaut_: "Passes" in the table -- what does that mean?

sarah: Rahim might have just passed it as a [missed]

spectranaut_: Anyone else want to weigh in on this?

scotto: I agree we need more answers to what is being tested; what sarah said

spectranaut_: We can just record that the string undefined should not be considered undefined
… that's the conclusion we want to to go with

sarah: If anyone wants to test the browser, you can test the fiddle on your machine

Old issue triage!

spectranaut_: That was our agenda

scotto: Might we have the agenda for next week
… keep the meeting light to get to triaging

jamesn: If we have anything urgent coming up

scotto: It's the internet, everything is urgent

jamesn: Trying to find times on the agenda for things like that

Minutes manually created (not a transcript), formatted by scribe.perl version 221 (Fri Jul 21 14:01:30 2023 UTC).

Diagnostics

No scribenick or scribe found. Guessed: knights

Maybe present: cyns, jamesn, scotto, spectranaut_

All speakers: CoryJoseph, cyns, jamesn, keithamus, pkra, sarah, scotto, spectranaut_

Active on IRC: Adam_Page, BGaraventa, CoryJoseph, cyns, dmontalvo, Francis_Storr, jamesn, katez, keithamus, knights, pkra, sarah, scotto, spectranaut_, Summer, Theo-MSFT