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

W3C

– DRAFT –
ARIA WG

23 May 2024

Attendees

Present
aardrian, Adam_Page, CoryJoseph, Francis_Storr, giacomo-petri, jamesn, jocelyntran, katez, keithamus, knights, pkra, Rahim, ray-schwartz, sarah
Regrets
-
Chair
ValerieYoung
Scribe
sarah

Meeting minutes

New Issue Triage

New PR Triage

spectranaut_: I think these are all monorepo PRs or from us

spectranaut_: we'll talk about the monorepo soon

spectranaut_: one aria-hidden PR from scott

scott: this is trying to resolve the expectation around aria-hidden and focusable items. So if you have a container with aria-hidden and a button within. Talking with Aaron on this, and trying to get this closed

spectranaut_: any other reviewers?

scott: this is closing the issue I opened, but also related to work Melanie and Aaron were trying to do with the giant aria-hidden PR. I went in a bit of a different direction, but it's all based off the same ideas

Adam_Page: I'm interested, could also do WPT coverage

Rahim: I believe this is testable, maybe we could create a separate issue? I can follow up

WPT Open PRs

spectranaut_: nothing new, moving on

Deep Dive planning

spectranaut_: this morning, had ariaNotify deep dive. Would be good to share the slides

<giacomo-petri> https://docs.google.com/presentation/d/1dZYiHdf1k0km4U5ZVlGa6QzgLrkcYOoNXzxwyhsln0g/edit#slide=id.g2dfa6629443_2_85

spectranaut_: I don't think we have anything planned upcoming. Anything we want to plan?

(silence)

spectranaut_: moving on

Monorepo status

spectranaut_: monorepo work is in progress. There are a lot of PRs on ARIA, these are PRs that are moving the history of the individual specs into the ARIA repo.

spectranaut_: there are PRs by Peter that are moving the PRs from the old specs to the ARIA branch. We are doing those one by one, I think Peter might do all of them for us.

pkra: I cannot do those where branches are not in the repository. Well, I'd rather not. Those are not a lot, I think the only one on the call is Sarah

spectranaut_: we're planning on finishing this work by next week on tuesday. If everyone can hold off on doing new work till then, it will make it easier to finish. On tuesday we won't be merging the PRs, because they're in various states. The editors of all the specs should decide how they want to handle the description of these individual PRs

spectranaut_: more to come

spectranaut_: any questions, or James and Peter is there anything else to talk about?

pkra: if you want to make a PR to the ARIA spec, that won't cause any issues. If you do one to other ones, we might miss them

keithamus: are there any guidelines to making a PR like the core AAM one?

spectranaut_: the goal is to have one PR for cross-spec changes

pkra: you can do it yourself, but you'd need to wait until the Core AAM spec is moved over

spectranaut_: we will have a PR preview. It's not the same bot so the diff option won't be available. It's using Netlify.

spectranaut_: we have a long list of follow-up tasks that we will convert to issues so other folks can see what needs to be done still. Feel free to provide feedback on the process in the future

Consider loosening 'requirements' for naming certain container elements

scott: this came out of the editors meaning we had, and was brought up in previous ARIA meetings. I filed this issue per the conversation in the WCAG group where giacomo-petri asked if it was a WCAG failure if a table or dialog is not named if using the native HTML elements that don't require names vs. the ARIA ones that do require them.

scott: the line between what is an ARIA requirement and what is a WCAG requirement can confuse people. So that was the setup for this. By making this a requirement, we are saying that even in instances where naming these things could be redundant to other ways of identifying the containers -- a radiogroup could have radios that are fully

understandable by themselves, same for headings before tables and dialogs

scott: e.g. in a dialog that has a heading that is auto-read, having it be the name can be redundant. So we have a failure that doesn't always lead to a negative user experience. My suggestion here is the requirement should be changed. I think it's usually a good idea to name these things, but I don't agree with having a requirement that 100% they

should all be named

scott: it sounded like in the editor's meeting people might've kind of agreed with me

giacomo-petri: that's perfect, and I'm also working on an ACT rule that's checking these. I've removed all the elements with an implicit role from the rule.

giacomo-petri: then, as Scott said, if we set this as a requirement for implicit roles, this means that all tables, forms, need an explicit name regardless of if it's implicit role, which is not always necessary

scott: it's not always a compliance failure if they don't have a name. If it is, it would be caught by one of the existing WCAG rules.

scott: I've been taking that approach and slowly updating ARIA in HTML to be less strict, because not all the rules in there need to be as strict as they are. Some of those were done before conformance checkers being as robust as they are now

scott: one of Steve's goals was that the spec was made to ensure that ARIA and HTML would play well together. some of the rules were about not using ARIA, just use native HTML. Now that things have evolved, that is not really a problem anymore and where issues can pop up, there is already a WCAG requirement for it

CoryJoseph: I agree with everything you're saying, but I can't think of an example where a dialog would be acceptable without a name. I agree we shouldn't be more stringent than wcag, so if wcag doesn't say a dialog has to have a name, then OK. Is there a situation where a dialog could be acceptable without a name?

scott: a standard confirmation dialog in many native apps doesn't have a name. Taking that same concept to the web, a sign-in dialog where the first heading is "sign in", does that dialog also need to be named sign in, when the heading is right there that's going to be the first thing a screen reader reads?

CoryJoseph: I guess from my perspective, I'm thinking if you have the heading there, why wouldn't you say the heading is the name of the dialog? I see your line of thinking, I think there should be name for dialogs

scott: we do also have the effort of trying to implicitly name things from headings, including article and dialog. If that were the case, if they grabbed the first heading and made it the name, then it would be named. THen also, screen readers could put in their logic to not repeat the name.

scott: so depending on how someone gets to the dialog, they might hear the name, and then the heading immediately.

Rahim: quick question on if this has any impact on authoring practices, the APG guide. There's language for a lot of components like radiogroup, dialog where they say if there's a visible title use aria-labelledby and otherwise use aria-label. There's no indication that it's not required, should there be any crossover to updating that guidance?

scott: there should probably be something there to call out that these are ARIA requirements and not WCAG requirements

Rahim: APG has that section on keyboard operability, so some of those are optional, but other places like the ARIA table does not have any indication that they're optional

siri: I think there are cases where the dialog name is requirement, for example if you send any feedback like thumbs up, if there's a dialog that requires more information on why you sent the feedback. In that case, if the dialog doesn't have a good label, then users don't get the same information as visual users

siri: why do we not want a dialog name when we have aria-labelledby and have it map to the heading?

scott: whether it's aria-label or aria-labelledby is not what we're talking about right now. It's whether all dialogs, radiogroups, tables, grids don't always require a name. I'm not arguing that there aren't cases where things do need names, the issue is that this is a pass/fail boolean right now, and we're saying that these always need a name and

that's not true

scott: just want to be absolutely clear that I'm not saying dialogs don't require names, that would be a misunderstanding. I still agree that in most cases, it would be good, and people should provide names to these things. I think it's an overreach to say they're always required

giacomo-petri: I think we can go further in which roles we exclude from the requirement. For example, we have a search form, with a form input inside, with label search, and so with a screen reader, you 'd find search landmark, search form, search input,

giacomo-petri: I'm not saying we never need the name on specific elements. We could even determine which elements might or might need the accname, but I think the example with the search is pretty self-explanatory

spectranaut_: just wanted to say that I agree with this change, I feel like if we want them to be requirements we should agree with WCAG. I think there are some scenarios where the context is enough for the names of some of these elements, and we should allow for that. I just wanted to voice support.

spectranaut_: there's a little bit of pushback from Cory, but is there any other feelings on this?

aardrian: I also support this for the reasons scott outlined. This isn't wcag, I"m good with this

CoryJoseph: just a follow up on my statements, I think aligning with not being more strict than WCAG, I would support it

<siri> +1 as optional

spectranaut_: no other pushback or disagreement? I think we should open a PR and discuss individual elements

scott: jamesn I want to pick on you

scott: you had the strongest arguments against this, if there's an argument to continue to require this, I'm happy to to reconsider

jamesn: I've heard enough, I no longer hold that opinion

spectranaut_: next steps, someone want to volunteer?

scott: I'll do it once the repo's calmed down.

jamesn: PRs on the ARIA spec are fine, since it's the one that isn't moving, right pkra?

pkra: last we decided, yes

Why does <slot> not allow role?

spectranaut_: keithamus, why does slot not require role? This is an issue on HTML AAM

keithamus: so, you can't put role on slot which feels weird because it's often interleaved with other things that have roles. We made a mistake with a package we were workign on because we changed a div to a slot, and then we noticed the issue and had to do a larger refactor around that

keithamus: the only things that don't allow a role are things that go in the head which feels right, but slot feels weird with that

spectranaut_: should we start off with the context from scott from last time? Though you can see from the issue that slot isn't supposed to be exposed to the user at all, so this would change things

sarah: is it currently exposed in the layout tree right now?

keithamus: you can override the display: contents and make it part of the layout tree if you want it to be

scott

scott: for all the opposition when we brought this up, I actually don't care. Just when I brought this up, I was told no. If opinions have changed and HTML editors are supportive of this then fine. I don't see us making this change without getting wider support

scott: to give this a role is kind of going against what the element is defined by. Even per the example that you can add styles to it, you can do that to head content too but that doesn't mean you should. I don't know that that's a strong enough argument there

keithamus: I guess the next step then is to take this to whatwg, which I'm happy to do and prod them a bit until someone tells me a firm no

keithamus: if I do get a no, I can have a good followup action to add words to the spec about how this is a weird case where you can't add a role

scott: right now I have to show a thread from a validator to show the explanation I was given

jamesn: I understand there was a time when mistakenly a role got applied to a slot in some real code at some point, but was this a complex workaround to not do this? I'm struggling to see the benefit of moving that way

keithamus: I think in this situation, we had a div role=tablist which can only have direct children. I don't think slot counts, but we had some difficulty, and I think ultimately we used javascript to apply the role depending on whether or not the element is slotted, so not a wonderful workaround

keithamus: the more I get told probably no, I think I need a definitely no

jamesn: it isn't something I've heard from our web component folks, but maybe they just haven't raised it

keithamus: I'm just in a weird position where I know enough to be dangerous, and here we are

scott: some of the contextual role stuff that's going on -- in Firefox specifically, I know that if slot is exposed in their tree as a generic, then right now, and I think this a bug they need to fix, but right now if a slot is in a <ul> and exposed as generic, then that means all the li elements see that generic as their parent and then think

they're not inside a list and turn into generics. So that's a situation where slot being exposed as a role breaks the accessibility of the list

scott: so that's a place where if slot was role=none, or no role, this wouldn't have happened

sarah: would the new change where you can have generics between tablist and tab, so even if slot had a display it would be OK. Would that have helped your use case?

keithamus: I need to think about it more, maybe

jamesn: I wasn't aware you could change the display, maybe that should change things?

keithamus: scott's argument was that you can also do that for script elements

<scott> script, link, title, head, style... anything that is essentially display none by default

keithamus: dunno if I want to open the floodgates on all of these, my thoughts are that slots are interleaved between other elements, so it can be difficult to massage the nodes

aardrian: was going to parrot what keithamus said, where people do display the code, the script and style block. My thought was that those are edge cases and we don't need to panic about those. I would like to think that we don't need to worry about putting a display on a slot, but I could see a developer doing that. I don't know if that means we

need to allow role on it

keithamus: I think that depending on the conclusion from someone telling me no, I think I'll ask chrome to put a big warning on putting a role on a slot, because we did make this change, and shipped it, and didn't notice until getting an axe warning

spectranaut_: I think that sounds like a good resolution unless anyone doesn't want keith to do that, let's move on

CSS: handing the a11y of anchor positioning

spectranaut_: another one from scott, he left

spectranaut_: anyone else have context on this issue?

pkra: and it's very future looking still, anchor positioning isn't widely available, is it?

spectranaut_: I wouldn't mind ending early, any parting comments or should we just run away?

<elguerrero> Going to drop off -- bye!

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

Diagnostics

Succeeded: s/A+//

Succeeded: s/A+//

No scribenick or scribe found. Guessed: sarah

Maybe present: scott, siri, spectranaut_

All speakers: aardrian, Adam_Page, CoryJoseph, giacomo-petri, jamesn, keithamus, pkra, Rahim, sarah, scott, siri, spectranaut_

Active on IRC: aardrian, Adam_Page, CoryJoseph, elguerrero, Francis_Storr, giacomo-petri, jamesn, jocelyntran, katez, keithamus, knights, pkra, Rahim, ray-schwartz, sarah, scott, siri, spectranaut_