<MichaelC> present_

<scribe> scribe: MarkMcCarthy

TPAC Agenda - https://github.com/w3c/aria/issues/806

jn: reminder, TPAC agenda is needing done, coming up soon!
... planning to send confirmations by end of next week

Charter Status

jn: not sure of many updates for charter status, people working on i

mc: had a phone conversation with commentors, resulting in not huge edits
... hopefully gets resolved in some way soon

jn: current one expires oct 31, no extensions at the moment, work on it if need be

Reminder: Cutoff date for Spec Changes for next Working Draft – tomorrow Oct 5 2018

jn: cutoff date for aria 1.2 changes (current working draft ) due tomorrow

cm: so many things depend on generic, i wonder if it's a good idea to add right now

jn: until it's in, we won't have done the work to get wider comment
... let's get something in to be as sure as we can

cm: ok, we probably should make sure of whether we're doing generic or none

jn: we'll discuss that shortly
... might be that we think we're ready, but may not have everything 100%, so this will help

Accname issues – final steps.

joanie: good news, i propose that we issue a CfC this afternoon to transition accname. will continue working on implementation reports and bugs
... if CfC passes, we should be able to move on
... those following along know about the case at hand, turns out to be an edge case
... we have a problem to fix in 1.2 to make it clearer what the answer to that case should be
... only gecko does what bryan says is the correct answer
... unless built into edge, chromium or webkit, we can't move forward
... this in mind, I plan to skip that test unless there are objections

mck: support skipping that particular test
... if anybody puts this code out thinking a good label will result, they're mistaken
... better spec if there's cleaner language and code, but might be difficult right now

bg: part of this comes down to whether we can immediately find solution

mck: i agree, might need more discussion

joanie: in this case, we defer to HTML, which would expect display text

mck: discussion we'll want to have, in 1.2, on where CSS text comes into play
... better to leave out for now

joanie: i can skip this test?

jn: yes

joanie: all in favor of sending out the CfC?

<carmacleod> +1

<HarrisSchneiderman> +1

<jamesn> +1

+1

<melanierichards> +1

mck: this should help with clarity and consistency

joanie: ok cool, looking pretty good if we count the work being done with google
... thanks bryan for all the hard work with this

bg: going to be gone between Oct 13-Nov 1, so i might be hard to reach

joanie: well hopefully your spec will be in PR by then

<jamesn> https://www.w3.org/wiki/TPAC/2019

jn: dates are known for next year's TPAC in south of Japan

Role Parity – Generic role and aria-textseperation – try to resolve issues so can merge tomorrow

<jamesn> https://github.com/w3c/aria/issues/699

jn: best to catch up on the issue

<carmacleod> joanie can you verify: https://github.com/w3c/aria/issues/699#issuecomment-427026128

jn: over the last week, melanie and matt have been working on a lot of comments, carolyn working on giving some of them answers
... matt's question on role of none struck me

<jongund> I like the idea of using role=none

cm: joanie, i listed what AAM specs say in this comment ... could you confirm that I'm understanding AAM tables right?
... with current mappings, does it make sense to say div and span default to none?

mck: if we don't rationalize mappings, possible to have one role for both?

joanie: two issues, if it's focusable, it's included in tree as a text object
... that said, a div is like a block of text, a section, whereas span is inline
... if we have generic role and spacing property, what would happen in mapping?
... presence of properties combined with roles may have dedicated mapping tables

cm: discussing giving div and span role of none, not even generic

mck: still looking at this, at what the actual difference is
... a div and span with role none and block styling, is that the same as a paragraph with role presentationa nd block styling?

cm: i'm not sure

joanie: what role presentation says is if presentation applied to object, it's not in tree but exposes contents
... a div or span, object is there, text included in parent
... if there is a reason for separate dedicated object, we should have a generic role
... do we want a dedicated accessible object?

mck: primary puprose of generic was to map div and span

joanie: a div with role none wouldn't be in the tree, parent element gets the content. without role none would be in the accessibiilty tree

cm: this isn't how i read coreAAM for presentation
... says if it has required owned descendents, map it as section, etc.

joanie: this is the case if you have a say a table, which has cells, can't have a table cell w/o table
... cells get role section

cm: ah i have written this incorrectly

joanie: yes, children/content have to be there, accessible object does not

mck: so you're saying if you have a div with text, button, more text...
... in a paragraph

joanie: depends on platform
... two text strings would be separated

mck: so now, if paragraph has a div with role none, and same things inside, how does it behave?

joanie: in theory, text of div gets sucked into paragraph, paragraph has embedded object (button)

mck: paragraph contains text with button... how does tree represent that?

joanie: where the button is would have an embedded obj character

mck: what is the value of that difference, between div with role none, and without?

joanie: if you have a large block of text, like a paragraph, and want different navigation
... or divs in a tree, because that's the way it is because that's what gecko and ibm came up with

mck: if you have text-separation on it, and a bunch of divs in a row, do they have to have separate objects in the tree?
... if each div with role none and had text-separation=paragraph, how would they behave?

joanie: in theory, parent container gets text content. becuase of text-sep, when text is folded into parent, content would be separated by two newlines. hypothetically
... user still has content but no longer an object
... is there a case where you'd still want an object?

mck: argument came down to semantics of div and span, especially regarding text in those items
... spans by default would have no separation, divs would, depending on screen readers etc.
... seems like all semantics wrapped up in how text is separated has nothing to do with how to detect edges of container
... possible for in one screen reader to navigate divs and spans

joanie: elaborate?

mck: in jaws, can set key to navigate to next div and span. dunno why you'd do that
... couldn't have single role for both if we don't go with something like role=none
... to navigate by div or span isn't navigating semantic content. no obligation t osupport that
... just my opinion!

<joanie> https://github.com/w3c/aria/wiki/Plans-regarding-role-parity#generic-role

joanie: on the one hand, let's discuss div and span. what all gets the generic role?
... do header/footer get generic role? maybe you want to jump to that
... if we go the none route, things will start getting removed from a11y tree

<carmacleod> Summary of elements we said would have generic role: https://github.com/w3c/aria/issues/699#issuecomment-426681491

mck: i think we have to decide something about purpose
... if divs and spans dont have semantic roles, they should be transparent containers
... similar to attributes, bold italic, it's not an object in and of itself

joanie: unless it were focusable
... could be that an author uses a b tag and a tab index

hs: could be useful for a scrollable region

mck: could see where someone might create a mode in a reader, tab to highlighted words that have actions
... don't want them tabbable because they're in the middle of a text and could disrupt flow
... maybe has a popup
... can imagine a situation where text isn't read as text, comments, misspelling etc.

joanie: if item is focusable, it needs an accessible object. see wiki on bold/inline (when exposed)
... if it's exposed, has to be as a generic something

<jongund> +q

janina: wondering if we're gonna have a similar conversation when discussing multilingual embedded content

cm: right, leaning towards keeping generic role? scary to wipe all context of many things

mck: differences in platforms making me question what are attriibutes of object, vs...
... what's the difference between a paragraph and a span of text in a paragraph with bold attribute
... only thing that happens is im in the bold text then out of it

jn: need to create same semantics as exist in HTML
... with roles
... having a couple extra roles is less harmful than not adding things we do need

mck: some might be concerned with that

jn: working to create a role that could be used in several areas
... none is something we expect general authors to use

mck: is the case for a role based on whether we need an object in a11y tree?
... just because there's a role doesn't mean screen readers have a unique navigation
... is the idea that header and footer are meaningless semantics, make it generic? don't distinguish between header, footer, div, span?

joanie: sometimes might want to announce contents of header and footer

jg: thinking about authoring, sometimes div and span might be interactive. might be a factor in determining if they have a role
... difference of b or i tag and span with styles applied?

joanie: none
... if something is mapped to tree, use generic role

mck: if the HTML AAM says generic with text-sep, when get to the API level, does that end up creating an object for that span?

joanie: no, beause what would go in HTML AAM would be something like not mapped
... if there's a need to map it, use aria role=generic

mck: a key aspect is having generic being an element that doesn't have a role
... a screen reader doesn't tell me that it's generic
... what is the effect on AT? for AT devs?
... the way we have generic worded, it's not clear if it's an invisible container or not on the end user perspective
... this is what led me to the "none" thinking, 'oh it's invisible, it's not there'

jn: i appreciate your thoughts Matt, don't see the harm in doing generic
... the fact that divs and spans are exposed means we need something other than none?

mck: they're exposed at an API level, not to AT users - we don't want them exposed to AT users

jn: while we don't have conclusions, my gut is that we need generic
... and we can explore none more

cm: seems that way to me too

mck: i'm kind of okay with going with the gut feeling, but don't have a clear understanding of the issue
... feels weird to put an object in the a11y tree if we don't want it there
... only in the tree in some circumstances, like if they're focusable for instance

joanie: historically speaking, how do we deal with children in objects, then it became documented, and set into practice

jn: overtime, need a conclusion: move this to 1.2

cm: that's fine, can probably get some words in

jn: doesn't mean stop working, but that this is important enough and should go through procedure once we have something, and we can publish after
... reasonable conclusion?

mck: +1

cm: +1

+1

<HarrisSchneiderman> +1

jn: we stick with generic, don't explore role=none right now

mck: i love the rationalization!

<HarrisSchneiderman> +1 too scary ;)

cm: because it's too scary!

jn: carolyn, if you can wordsmith?

cm: yes, matt, thanks for comments on PR

jn: let's discuss more at TPAC

mck: didn't get into topic of precidence

cm: can discuss in issue

jn: thanks everyone!

This is scribe.perl Revision: 1.154  of Date: 2018/09/25 16:35:56  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: Irssi_ISO8601_Log_Text_Format (score 1.00)

Default Present: jamesn, MarkMcCarthy, Joanmarie_Diggs, MichaelC, HarrisSchneiderman, melanierichards, CurtBellew, jongund, matt_king
Present: jamesn MarkMcCarthy Joanmarie_Diggs MichaelC HarrisSchneiderman melanierichards CurtBellew jongund matt_king carmacleod
Found Scribe: MarkMcCarthy
Inferring ScribeNick: MarkMcCarthy
Found Date: 04 Oct 2018
People with action items: 

WARNING: IRC log location not specified!  (You can ignore this 
warning if you do not want the generated minutes to contain 
a link to the original IRC log.)