Meeting minutes
[New Issue Triage](https://bit.ly/3GrI5uz)
https://
jamesn: I think we should move to 1.4
bryan: aria-current was never meant to be restricted to only interactive elements
<siri> +1
jcraig: I don't think step is fully represented, I don't think there is enough for the AT to know you are on step 14 of 20, unless you have an additional label... we def need more time to discuss this
jcraig: this proposal doesn't need a deepdive
jcraig: maybe there is related issues
jcraig: I'll think about more after call
https://
jamesn: I think we should close, other than the link from accname to html aam
cyns: both directions
peter: there is a note in the html spec...
jamesn: it's not in the references in the bottom of the doc
jamesn: an editor of the doc needs to change hte format so it links correctly via respec. it's a hard coded anchor andit shouldn't be.
bryan: I'll try to look at it and fix
[New PR Triage](https://bit.ly/3lKYId0)
https://
joanie: I didn't merge because james had asked for two reviews, and caroline reviewed it, I'm ready to merge it
https://
jamesn: sarah reviewed, can I have two more reviews?
peter: I'll review
jcraig: the respect prefixes the text with "note". rfc2019 normative/non-normative/informative is clear
jamesn: we can merge when peter reviews
Holiday Meeting Schedule - meet Dec 16. Then next meeting is Jan 6, 2022
<pkra> James was faster than me.
[onboarding](https://github.com/w3c/aria/issues/1401)
jamesn: is this done? jemma, do you need help?
Jemma: this is on going documentation, never "done"
jamesn: can we move from google doc into something we can all see?
Jemma: sure
Jemma: can we just check in everyonce in a while to see if there is anything to add?
jamesn: it can be in the repository or the wiki?
[Should user agents calculate aria-posinset and aria-setsize on treegrid row if there is no author-provided value?](https://github.com/w3c/aria/issues/1602)
mcking: aaron, can you give your summary of the difference between row count and posinset, in practice, what are their differences?
aaronlev: I need to think more before doing that
jamesn: 1303 is the background for this
jamesn: from my perspective, they can be two different things. sometimes they are the same, sometimes not.
jamesn: if you have a virtualized treegrid..... <scribe couldn not catch>
jamesn: just because the properties are available, does not mean the AT should read them at all times when navigating between rows
jamesn: to me they are kind of different, the number of rows may be spoken when you enter the ride yourself. when you navigated around and drop levels.... it does required a lot of screen reader intelligence in order to know what is the relevent data
mcking: I don't think we know if we know what a user might want to know
mcking: they are a useful construct, trees, but when you combine them with the notion of virtualization... you are layering complexity, and I don't know the concept of numbers is the way to help a screen reader user get the same information that a sighted user has
mcking: if you have a virtualized structure on the screen, like scrollbars, and the way things fad at the edges... these are imprecise an subtle, and most users might not notice or care. if you hoist info of the nature on a screen reader user, you are creating cognitive load
jamesn: when you change rows, you hear the row number change
jamesn: despite whether you only have 10 rows, you might move ot the next row and hear 11
mcking: we don't know what kind of end state we want to provide
mcking: but maybe that is not our role, we just provide all information we have to screen reader, and the screen reader decides
jamesn: if you had a treegrid with 100 rows, then that would be too many to go all the way through, if you are row 50 to go to row 51, is that useful? and if so, we need the virtualization
mcking: I never understood if row count applies too...
jamesn: if you don't include them at all, and let the SR calculate them, what does it read?
jcraig: it will count every row that is not a header row
mcking: but is the row count all rows are all levels that are currently visible?
<siri> How will they get sublevel row count? and what about invisible rows
mcking: or the rows that are in the dom .. but not visible because the branch is collapsed
mcking: are those counted in rowcount?
jamesn: they might not be fetched
jamesn: they might or might not be in the dom. they are unlikely to be in accessibility tree
mcking: if you have a tree, but every node is in the dom, but only one is visible (because all collapsed) the accessibility tree only has root node
aaronlev: you do nine accessibility APIs, right?
mcking: if you have a normal tree, with one parent node visible, is that the only tree item in the accessibility tree?
aaronlev: it's possible that is the case
jcraig: I think that is how we do it in mac
aaronlev: on mac, there is no "invisible" objects for the accessibility API
aaronlev: on others there is a concept of that in the accessibility tree
aaronlev: there is no requirement about this. trees are treated as things that when they get focused, it provids the data about what you can do with it
mcking: so we talk about rowcount is for virtualized content, to give you the correct total number of rows....
aaronlev: the setsize is for the level that you are in
mcking so when it comes to treegrids, one question is which row is counted in row count...
mcking: my guess is that it is what person can navigation to without expanding or collapsing.
mcking: I don't know if you want it to be consistent... there are so many views that load more junk, imessage or messanger, if you scroll back in time it loads more stuff. the way I know it loaded more stuff, is you do a three finger tap. at first, you might hear 9-15 of 15, then you scroll back and you hear rows 30-36 of 45, and you scroll back and hear 52-58 of 80
mcking: you get the same idea on the web
jamesn: so knowing that number of how many their are... why put a rowcount on there
mcking: that is just the screen reader counting what is available, not using rowcount
mcking: you can never know what is the total msg history, right?
jcraig: yeah. the screen reader knows "is it rendered in the dom or not"
siri: if the row count is changing based on what is available
siri: how do you know if there is more to reveal?
jcraig: the same way a sited user know? does a sited user know they are going to expand? only when you get closer to the top
mcking: sited users have the scrollbar indicator
mcking: when I'm in imessanger on web, I hit ctrl-home and it will take me to the top of what is loaded, and if I hit page up I know it will load more. I just know, but I don't know how that can be communicated to all users
siri: with sighted users, there is usually some queues to tell you you can load
jamesn: rowcount="-1" means useragents do not calculate rowcount
mcking: does that mean that when the user is in the table they can't know how many rows are currently rendered??
jamesn: when you have a table and you don't know how many rows are in it, then you don't want to give misleading info
mcking: but if there are 20 rows displayed, and you set it to -1, then the screen reader cannot tell me that I'm row on 1 of 20
mcking: there is a big use in knowing there are at least 20 rows available
jamesn: it would be nice to know there is a minimum but probably more...
<siri> +1 james
jamesn: that is what I wanted to be able to provide
mcking: but if the user agent sets it to -1 and the API doesn't have a way to get to the number 20
joanie: this is weedy, and depends on the accessibility API, but you can get from the API how many rows are rendered.. in ATK and IA2 have ways of knowing how many cells are rendered, screen readers need to know that for their own navigation to work. screen readers need a representation for the rendered table. the -1 is the indication that there are more rows or column that are present, but there are more than we know
mcking: I just want to make sure the screen reader can know how many are currently available
jamesn: it says user agents "should not" calculate how many rows exist.
joanie: I would love a straw pull on this because I have a bit rotting change on chromium that aaron is waiting to hear if it should be done
mcking: I don't think they are the same thing posinset/setsize and rowcount/rowindex.
<joanie> Straw poll: Should user agents calculate aria-posinset and aria-setsize on treegrid row if there is no author-provided value?
<Matt_King> prsent+
<Matt_King> +1
<pkra> +1
<aaronlev> +1
<jamesn> +1
<CurtBellew> +1
<joanie> 0
<StefanS> +1
<Jemma> +1 no thinking other wise.
<jcraig> +1 this is what UAs do currently as far as I am aware
<siri> +1
jamesn: maybe we need to make a really big tree grid to see how screen readers handle it
mcking: I've seen some really big tables that cause trouble
jamesn: lets not make spec changes until we do more experimentation
mcking: experimentation would allow us to open a more concrete convo with screen reader developers
<Jemma> I added onboarding doc link to https://
mcking: maybe we can find creative approaches to mimicing virtualization
[Prohibit rowcount and rowindex on rows in treegrid and prohibit posinset and setsize on rows in tables and grids](https://github.com/w3c/aria/issues/1303)
[Spec is unclear on aria-invalid="spelling" | "grammar" uses](https://github.com/w3c/aria/issues/989)
<Jemma> onboarding doc in github is https://
jamesn: aaronlev do we have a path forward, or is there something you need?
aaronlev: the way that we do it in chrome, if aria-invalid it used... it should have never been done in one attribute...
aaronlev: spelling and grammar only makes sense on text in editing context. like on a span for example. it doesn't make sense on a whole form control. aria-invalid=true, it only makes sense on form controls and not spans. I don't think this is clear in the spec, and chrome implements the thing that makes sense
<pkra> Need to go. bye everyone
jamesn: we deprecated aria-invalid on anything other than a form
mcking: when we deprecated it, didn't we discuss adding a new property?
jamesn: we have discussed it sense.
mcking: I'm concerned about letting aria-invalid being something that is not an input, I'm more interested in a new property
aaronlev: I think that would make sense to have a new property for spelling and grammar, there are few implementers of aria-invalid for spelling and grammer, and I think that user agents should support the old deprecated situation for a while
jamesn: I don't imagine this will be removed
RRSAgent: make minutes