Meeting minutes
New Issue Triage
jamesn: aria/#2124
pkra: deals with how to use aria-owns on elements without children
pkra: spec should be updated to reflect what browsers are doing
pkra: but we need to confirm that
scotto: I linked a related issue for ARIA in HTML
pkra: there’s also an editorial part to it because of very old language
pkra: will bring up on edior’s call
pkra: I’ll take assignment
scotto: me too
jamesn: html-aam/#532
scotto: no triage needed, this is for me
New PR Triage
WPT Open PRs
spectranaut_: we talked about most of these last week
Deep Dive planning
jamesn: we had a tentative plan for value calculation
… but may not need a deep dive after all
spectranaut_: I recall thinking that a little reorganization would be good, but it’s not necessarily complicated
Matt_King: we say how to calculate value in the combo box spec
… but the way it’s written is confusing, and it’s not consistent with similar components, like textbox
… so the idea is to make sure there’s a clear definition in one place
… and then it can be referenced from combobox, textbox, etc.
… the current wording is a bit hacky
sarah: I also have some context and remember when it first came up, so would be happy to help
… I’ll take assignment
jamesn: we need a reliable way to test this
scotto: I still question whether this belongs in accname
… can’t it just be a section in the ARIA spec?
jamesn: the accname spec will need to reference this value calculation algorithm
… why don’t we _start_ in the accname spec, and then decide whether it’d be sensible to relocate it
Matt_King: the original motivation for putting it in accname is because the value calculation itself wants to reference a part of the name calculation
… and until we do this thoroughly, we won’t know how much of the name calculation we’ll end up using
BGaraventa: it comes into play in the recursion
jamesn: okay, any other deep dives?
… scotto, you‘ve got the other two
… but no urgency to schedule a deep dive right away
sarah: I’ve got a potential deep dive topic for aria-maxlength — aria/#1199
… useful both for conveying max length upon focus, but also if you want a UX where assistive tech would “bonk” when a user’s input reaches it
jamesn: can we fill a deep dive with this?
sarah: it’s surprisingly tricky
jamesn: we should have more documented material & homework before we schedule the deep dive
sarah: okay, I’ll add some more information
Final Proposal: AccName Role Traversal Proposal
scotto: this is essentially limited to hyperlinks
… so the basic idea is we add a note: “hey browsers, in this scenario you can use innerText” for accname
… where browsers fail to provide a proper accname in these cases, assistive tech work around it
scotto: because of the hyperlink’s transparent content model, if it’s filled only with things that have name from author, then it stops the accname traversal
cyns: I’ve pinged AaronLev, but this feels icky to me to not have the name be correct in the a11y tree
… need to see how feasible this is, but need to see if we could reinterpret name from author elements inside hyperlink as generic
scotto: we did discuss and abandon that
cyns: the whole point of these mappings was that we would eliminate magic
Matt_King: let’s say we have a large table inside a hyperlink
… and the table has a 3-word name from author
… then doesn’t all the content of the table disappear from the accname?
<cyns> hyperlinks couldn't have block content inside them when we wrote these rules. HTML changed the definition of link, and we haven't fully caught up
scotto: if the hyperlink was forcibly given a name or calculated from one of the name from author, then the contents of the hyperlink would just be squashed and AT would no longer be able to get to it
… so this is a brittle situation and we need to be careful not to introduce a breaking change in the spec
Matt_King: do we expose all the semantics of a hyperlink’s children inside the a11y tree?
scotto: yes, but for the instance where hyperlinks are explicitly named
cyns: it’s more than ickiness, it’s hard to debug, hard to understand what’s going on
… (the absence of the accname in the a11y tree)
<jamesn> i agree
BGaraventa: there is merit to having a careful algorithm
Matt_King: the intent is just to make a whole bunch of content clickable
… but there’s only one href for the whole thing
… and the SR user gets zero benefit from every bit of content being enclosed by the hyperlink element
… it’s uncertain what the ideal SR UX is here
… the intent of the author is completely masked by this markup approach
BGaraventa: I’ve also experienced when a link encloses 5 images, and it presents as 5 links that all go to the same place
… this only matters for inner elements that do not support name from content
… so a spec update should center around that
… and accommodate traversal
Matt_King: we shouldn’t force screen readers to pick up the slack for browser’s responsibility
… a browser should be able to deterministically calculate accname for this situation
… it’s important to prioritize parity for situations like this — how does a sighted user interpret a hyperlink filled with rich semantic content, and what is a comparable AT experience?
jamesn: we’ve had a lot of discussion and no consensus
… let’s move the discussion into aria/#1821
… let’s answer scotto’s proposal and, if needed, make a counterproposal
Matt_King: this feels like a HTML problem
scotto: absolutely, yes
jamesn: the same problem can exist with an ARIA pattern though
jamesn: let’s continue the discussion in aria/#1821
Status Update: Consider providing a way for authors to customize the announcement of state
jamesn: there are additional questions in the issue since we last talked
… where are we with this?
scotto: someone proposed using `aria-valuetext`, and that’s intriguing
… allowing it to be used on more roles
jamesn: seems like the issue is stuck, would anyone like to take assignment?
giacomo-petri: we need to make a decision if we want to proceed with that
… there are instances where the switch button doesn’t mean on/off
… for example, a switch for centimeters vs inches
jamesn: that approach sounds reasonable
… how about a draft PR?
giacomo-petri: yes, I’ll take assignment of that