Meeting minutes
New Issue Triage
spectranaut: aria#1817 about aria-owns
jnurthen: first bit could use work, the rest seems already covered in the spec.
… label clarification
spectranaut: there's a 1.3 blocking issue on aria-owns. maybe related?
… #1161 about owned by
<aaronlev> There are some good notes on now Google avoids problems with aria-owns here
jnurthen: related but not same.
aaron: I added link in IRC on chrome's aria-owns implementation details.
… more about interaction with HTML.
jnurthen: first part would be 1.3ed // clarification.
spectranaut: assign to jnurthen, wait to hear back.
aaron: chrome is pretty complicated about, e.g., changing ids, removing elements with id, multiple elements with same ID
jnurthen: maybe a note "there are limitations, known issues etc"
aaron: the doc is more for other user agents since our implementation seems a good basis.
New PR Triage
spectranaut: core-aam#142
… @jcraig please review.
… core-aam#140, mapping update.
… aria#1816 editorial
jnurthen: question for screenreader users about the notes in the spec.
… headings are just "note"
… in respec, you can add a title attribute to use as part of the heading. could add context
bryan: that would be an improvement
jcraig: respec adds this, so could we automate something like "note for section ..."
jnurthen: if there's more than one, we'd need a bit more but yes, I'll file an issue
jcraig: seems generally useful
spectranaut: aria#1814 draft PR
jnurthen: made one change but: do we really want to do this?
… phrasing content describes a things in HTML that we might not want to include in ARIA
… so maybe we want something completely different after all.
jcraig: I didn't like the name but would be ok if it matched HTML
scotto: really "textyMcTextFace"
Proposal: process Deep Dive next week!
spectranaut: follow up to TPAC discussions.
… come up with some suggestions.
siri: is there info on the results from TPAC?
spectranaut: yes, email was sent to list.
jcraig: can make Oct 6.
bryan: same.
jnurthen: sarah and scotto would be good.
spectranaut: any other topics?
jcraig: had some internal discussion about aria-actions
… heuristic detection seems implementable.
… some author techniques might interfere
… maybe a deep dive might be useful to gather implementation interest.
aaron: I haven't looked in detail yet. We should get James Teh, Matt King to look at it.
jcraig: idea is aria-action=idrefs. then AT will generate synth click events for them.
aaron: sounds like a good deep dive.
spectranaut: aaron, please take a look and say if you want a deep dive
… but matt and jamie are important
<jcraig> Draft PR https://
jcraig: issues are ...
… most of the discussion is in #1440
<jcraig> https://
mattking: at tpac we talked about needing some example web pages, screenreader expectations, experimental implementation.
jcraig: we'd ship it behind a flag to begin with in any case
Placeholder for next meeting discussion: Changing "2F.i" to "2F" in step 2H
bryan: step 2h deals with recursion process. When it recurses, it should check if name-from-content is supported.
… then continues into sub steps of process.
… it was pointed out, it was bypassing the process and going directly.
… bypassing all conditional statements
… no handling of circumstances where you wouldn't want name-from-content
… e.g., heading, button, opens up a menu. Then all of the contents of the menu would be part of the heading name.
… without checking if name-from-content was supported.
… filed it at TPAC
… if someone knows the background to the previous change, that would be very helpful.
spectranaut: change from 2017.
… linked on issue now.
jcraig: I have a related issue I filed a while back. We need to stop using numbered labels for steps.
… this is a good example. It's confusing, adding steps changes numbering.
… if the step was called "aria-labelledBy", then I'd know.
… if we use logical IDs, we might avoid such confusion.
bryan: I like that idea. But for now, are we in agreement that this is how it's supposed to work.
jnurthen: I don't think we can do all the checks. When checking child nodes, e.g., div doesn't have name but you still want the child text.
bryan: I feel like the number of cases where it should be processed is less than those where it shouldn't.
jnurthen: I disagree. spans and divs occure too frequently.
… if you rule those out, it will break a lot of stuff
<jcraig> https://
mattking: it's about number of conditions in the algorithm, right?
bryan: yes. I could put in logic that it's an element that's perceived.
spectranaut: feels like there's something to discuss. Chrome has 2F implemented.
jnurthen: not completely. aaron pointed me to code where for each role, it does different things, depending on the situation
… feels like this is closer to what we want to do
<jamesn> q/
spectranaut: this would be a change affecting implementations
… so we need to clarify what they need to do.
… can we make this a PR?
bryan: I can. But we need implemetors to give input on which conditions make sense.
scotto: is this essentially tyring to fix the problem of, e.g., anchor with article in it but article has no explicit name, so the link technically has no name.
bryan: yes.
scotto: have been working on this with teams a lot recently.
… it shouldn't stop if the article has no accessible name.
bryan: right. it's hard to make a PR without knowing what should change.
<jamesn> is https://
<jamesn> see my comment here - https://
siri: question on anchor containing an article. doesn't the text get used?
scotto: it isn't and that's the problem.
… the way the algorithm is right now, it won't.
bryan: Reversely, say you have multiple options. should they all be part of the name? Probably not but right now they would.
<jamesn> / Get their own name from contents, or contribute to ancestors
<jamesn> examples button, checkbox, heading, link, switch, tab
<jamesn> // ----- No name from contents -------------------------
<jamesn> // These never have or contribute a name from contents, as they are
<jamesn> // containers for many subobjects. Superset of nameFrom:author ARIA roles.
<jamesn> examples alert, comment, grid, table, menu
<jamesn> // ----- Conditional: contribute to ancestor only, unless focusable -------
<jamesn> // Some objects can contribute their contents to ancestor names, but
<jamesn> // only have their own name if they are focusable
<jamesn> examples canvas, caption, details, footer, generic, list, listitem, paragraph
jnurthen: acc-name#128 has the link I mentioned.
… some suggestions in that issue
bryan: sounds good but we'd need an implementor to draft the PR in spec language
spectranaut: maybe start with what Chrome's doing?
jnurthen: I believe firefox is similar.
… then we can get more input
mattking: is there something in the aria spec that separates these roles from another?
jnurthen: maybe. perhaps instead of "named by content" could be "not only from author", getting us closer.
mattking: feels like this logic needs to be a reflection of some characteristic in the aria specification.
jnurthen: agreed. chrome's heuristic should identify _something_
mattking: do we need an aria issue as well?
jnurthen: I think it's an accname. if that feels like there's something missing from ARIA, then we file a cross issue.
spectranaut: I'd suggest to go through #128, go through it, see if there's enough information on what Chrome does, then possibly suggest a spec change to accname.
… then file an issue on aria
bryan: will do.
jcraig: bryan, could you ping me when it's a good time to make the label changes? It has to fit and avoid merge conflicts.
bryan: will do.
Support aria-description
spectranaut: scott PR on html-aam, but also blocking issue from jcraig.
<scotto> correction: html aam, not core aam
spectranaut: jcraig will make a few changes
scotto: will take alignment with html aam, 2 PRs todo
spectranaut: should be a simple check.
scotto: I'll still have to remove the relevant lines
1.3 blocking issues
jnurthen: check if you're assigned. get to work :-)
rss-agent, make minuets
rss-agent, make minutes