Meeting minutes
[New Issue Triage](https://bit.ly/2WqjXHi)
jamesn: "treeitem does not list aria-level as a supported property" seems like something we should fix, anyone disagree? anyone want to fix that?
scott: assign it to me
jamesn: I'm just going to assign these accName tickets to the latest milestone
[New PR Triage](https://bit.ly/3Dl07Oo)
jamesn: should we fix our test cases to have a doctype in them?
val: I can make that fix
[Meaty topic for next week](https://bit.ly/aria-meaty-topic-candidates)
jamesn: last meeting we agreed to have another meeting on grids and layout grids. That would be september 9th
jamesn: next week we have modal/non-modal dialog
jamesn: any objections to grid on september 9th? Anyone have other topics they'd like covered?
<Jemma> no objection
jamesn: Joanie, where are we on this?
cyns: you are still blocked by me
jamesn: it seems relatively trivial
cyns: if someone else has time, it'd be great to have someone do the IDL
jamesn: is it essentially rolling back the previous change?
cyns: yes
jamesn: is there anything else we want to sneak into the new draft?
jamesn: we have to remove aria-description from the IDL...is there anything else we want to put into this new draft? Nothing that requires additional tests (unless something is broken)
jamesn: I'd like to keep it as small as possible.
jamesn: do we need to do a CFC, Michael?
MichaelC: I think we should, yes
Jamesn: we can maybe even get a CFC out next week
jamesn: Joanie found it unclear whether user agents should calculate aria-posinset and aria-setsize on treegrid row if there is no author-provided value
joanie: I was initially questioning it because I was implementing chrome support. I subsequently did some more digging...one of the things I've found is that matt filed an issue that we should deprecate these on row, and that they are redundant...then I dug more and, no, they are totally different
<jamesn> https://
<Jemma> https://
<Jemma> https://
matt: I vaguely remember this. When you have a set of rows that are hidden because the parent row is collapsed in a tree grid. So those rows that aren't visible, could be captured in the row count. Row count is technically the total number of rows in the table
jamesn: if you have rows in a treegrid that is virtualized so that not all rows are in the DOM, which is why you'd want to explicitly declare the row count
matt: I guess I was trying to compare tree and grid when it comes to row count and posinset
jamesn: I agree with Aaron that they're not the same
jamesn: we had this conversation back when we put this feature in
<Jemma> https://
matt: This is starting to come back to me! On row index, you are supposed to "number" the rows, right?
jamesn: if they're not all in the DOM, yea
<Jemma> matt's comment in apg "I believe you should be safe using aria-posinset and aria-setsize in a treegrid and not using aria-rowindex and aria-rowcount. However, I will raise this for group discussion in an upcoming authoring practices meeting."
matt: let's say you have 2 collapsed rows and each of them hides 5 so you have 12 rows
matt: are we supposed to say that row 2 comes row 6 when expanded?
jamesn: I wouldn't want anyone to use posinset there. Its only when you're at a child-level that posinset and setsize are relevant
matt: you are going to have 2 different sets of numbering that are disconnected. What should authors do?
jamesn: there are in fact 2 different sets though. I don't know how to simplify that
jamesn: has anyone come across a treegrid that works well?
matt: we do have problems trying to communicate trees well with screen readers
matt: maybe we do want all this information
Jemma: can we hear what joanie said at the beginning of this topic?
Jemma: issue 1303's last comment
https://
jamesn: you only need to use rowcount and row index if you are virtualizing your tree. Otherwise it'll be calculated by the browser
matt: in the case where rows are hidden though...is there an expectation that in a tree grid, when you go from the first visible row to the second visible row, that you would hear row 1 and row 6?
jamesn: no, you should hear row 1 and row 2 until you expand it
matt: but in the spec, it says something different
jamesn: Then I'd agree that there is a problem there
jamesn: we need to refine the spec for treegrids specifically
matt: if you have a virtualized table with 100 rows and 20 of them are visible, are we supposed to do anything with this? I'd say let it be a 100 row table. If you are an AT user, you don't care which ones are visible
jamesn: but that is confusing
SarahH: so far for treegrids, I've only used posinset, setsize and level
ScottO: is there a way that it would be consistently known that you've left the nested rows and gone back to the previous row where the numbering has now changed?
Matt: screen reader tells you when you've changed level
Matt: if I pressed the down arrow once and went from row 1 to row 52, that creates a real problem for the developer, you have to know what's hidden before you build this thing?
Matt: you never know how many rows are going to be shown
SarahH: I'm starting to realize why treegrid with virtual cursor is used
<Jemma> mck: matt's suggestion is "We should consider prohibiting aria-rowcount and aria-rowindex on rows in a treegrid. Using them instead of or in addition to aria-posinset and aria-setsize could create problems." from issue 1303
jamesn: we can discuss this on the 16th deep dive
<Jemma> https://
<Jemma> apg tree grid example
jamesn: I think we need either a new issue, or work with it in this issue, in order to clarify the language in the spec to make it clear that rowcount and rowindex are not required when you have intermediate rows that are not expanded
<Jemma> +1 to Jamesn
<Jemma> +1 for clarification
jamesn: do we have consensus that this needs to be cleared up?
jamesn: anyone want to take that on?
jamesn: Joanie, does core aam specify how row count and index are calculated?
joanie: if you have something that implements the platform accessible table interface, it can't have a table that doesn't calculated things like row count.
joanie: even if core aam doesn't say it, there is an implicit rule that you have to calculate it
<Jemma> https://
SarahH: the APG example doesn't use row groups
jamesn: you either need an ownership or something like that in order to calculate it relatively easy
jamesn: what does core aam say about these calculations
<jamesn> Otherwise, if the role supports aria-posinset and aria-setsize, process the parent (DOM parent or parent defined by aria-owns), counting items that have the same role.