Minutes PFWG Mar 31 2014

Minutes from the PFWG teleconference of 31 March are provided below as 
text, and are available as hypertext at:
http://www.w3.org/2014/03/31-aria-minutes.html



- DRAFT -
Protocols and Formats Working Group Teleconference
31 Mar 2014
See also: IRC log
Attendees
Present
Rich_Schwerdtfeger, +1.415.624.aaaa, +1.512.445.aabb, Bryan_Garaventa, 
janina, Michael_Cooper, Susann_Keohane, 
Regrets
Chair
Rich
Scribe
Susann_Keohane
Contents
Topics 
1.      UAIG status
2.      ARIA 1.1 Meaty issues
Summary of Action Items

<trackbot> Date: 31 March 2014
<richardschwerdtfeger> meeting: W3C WAI-PF ARIA Caucus
<dmazzoni> This is Dominic Mazzoni - I just joined
<richardschwerdtfeger> scribe: Susann_Keohane
UAIG status
richard: documents using ReSpec and put out on github
michael cooper: A SVG mapping document is products of both SVG and PFWG. 
SVG folks will have to join pf working group
janina: looking for formal working statement from both groups. Sign off 
from both groups
.... time to publish. we will need formal approval from both groups like 
last call will need svg and pf sign off
richard: we are going to have to create semantics for graphics that will 
show up in svg spec at some point
matt: how many modules - svg, uaig, core user AIG, HTML5
richard: not decided how to deal with modules
modules meaning 2 sub efforts - (1) epub structural semantics (2) graphics
matt: modules are not uaig, but modules that are part of actually ARIA 
specs
.... 6 modules - core spec, core uaig, module epub structiral semantics, 
graphics, spec implementations for svg and html5.1
rich: shane is working on the first reSpec
JamesC: defining what ReSpec and Github is
<clown> http://www.w3.org/respec/
<clown> Mattking^
janina: html accessibility task force charter and it shows two doc as 
joint doc - HTML mapping guide and using aria in html
<jcraig> JC: Git is a type of distributed version control. GitHub is a 
series of tools (review, merge, compare) that work on top of Git.
janina: might want to rename the doc
<jcraig> JC: ReSpec is a JavaScript-based set of transformation tools that 
just make it easier to do the monotonous parts of W3C spec editing.
rich: michael to chair the next two ARIA calls
ARIA 1.1 Meaty issues
DM: web components - elements inside shadow is not accessible from the 
containing document
.... idea is to encapsulate them but there are new tools to jump inside 
components
.... challenge is what is if you need to refernce something inside from 
outside like aria labelled by
<Stefan> +q
DM: two options
<Zakim> jcraig, you wanted to discuss the activedescendant example
james: active descendant - one case where this would be applicable is 
combobox
.... list box that pops up will be a shadow dom control
.... ask DM for other examples
DM: one case provided may not be necessary for this proposal
s/DM for james
DM: activedescendant there might be an alt case for a combo box. listbox 
where every option in the list box is it's own
matt: would a grid be a use case
DM: yes it's certainly possible
stefan: how does AT access the part. Is there already some ref exist? what 
can we expect
DM: web componenet spec doesn't specifically ref a11y, but it makes it 
clear how web component behave. Take all the contents to the html add it 
to the dom where the custom element is found. that is what is presented to 
ATs
.... basic things work as if you added the html right there
stefan: you can hide the code implementation but the screen readers can 
access it
DM: issues with focusability
.... this idea of defining alt way to define id ref and could be a good 
feature outside of web components
<jcraig> Agreed that this feature is not limited to shadow DOM usage
Rich: question - the aria labelled by is inside component but wants to be 
labelled outside
.... question expands into the dom
.... what if you have an a11y that is traversing the dom itself
DM: there are javascript api that let you traverse inside a component
.... a11y dev tools ext for chrome supports web components and find a11y 
issues inside components
rich: which browsers besides chrome
DM: mozilla has provided support to the group, they are all at different 
stages at different browsers
matt: what is the status - what is the context?
<clown> http://w3c.github.io/webcomponents/explainer/ ?
DM: web components could be considered a mini iframe
<Zakim> jcraig, you wanted to suggest we talk about this outside the 
context of shadow DOM. We
DM: shoudl take the discussion outside of web components to progress 
conversation
matt: shadow dom, it is very diff than iframe that is not a seperate doc
<clown> shadow dom: 
http://w3c.github.io/webcomponents/spec/shadow/index.html
DM: it has its own render tree that is how it is seperate
.... it does have its own scripts and style sheets
matt: a11y req around landmark and regions. If there is a main landmark 
inside a web component will there be a problem
<?> everything should work as everything was composed into one page
susann: not sure who is speaking right now
<Zakim> jcraig, you wanted to suggest we talk about this outside the 
context of shadow DOM.
james: if you were to create a dialog and wanted to label it, you would 
have to give it a heading and ref it
.... I am labelled by the first heading inside of me
james; i control the first element inside me matching this classname
james: locked into id ref so it's not possible today
.... it would be easier to ref a particular element inside a control
.... shadow dom is an ex that can't be done currently
.... but it is useful outside the shadow dom
brian: asked a question about ref in mark up
james: open up token parsing and attribute parsing rules to simple 
selectors
.... javascript based use a ccs selector and pass in ref to a dom element 
at page render time
.... any and all selectors or whats avail in javascript
ich: query selector, it could be problematic
rich: you could have a mix that has to be parsed
.... the mix of id and javascript
james: he put in a decent soln to that problem
rich: the idea of a selector is really nice and save code
<jcraig> Dominic proposed on list: aria-foo="myIdRef selector(mySelector) 
mySecondIdRef"
james: this is what dominic posted on list
.... unique syntax as open parm
.... worried about complex selectors
.... two elements or two class, do we allow
matt: any other specs that use ids in this way or is aria the first
james: html and svg do use it in several places
DM: label for = and table cell can spec headers by id ref
.... svg is another classic example
matt: is there any other spec that allows id to be selectors
.... we're breaking new ground
rich: i like the selector indicator and run eval on it
DM: query selector and pass it that string to pass it that element
james: label all would be necessary
.... query selector all picks all
matt: every that has a certain class would make sense
DM: if aria attribute allows for multiple, select all
.... rare if multiple elelments map and only won't the first one
<jcraig> aria-foo="selector([role=heading])"
DM: aria labelled by that takes multiple ids then call all
matt: the author would have to responsible to make sure you have a unique 
match
<jcraig> aria-foo="selector(>[role=heading])" (only the first heading?)
DM: I could go either way. we don't have to specify every permutation
matt: if there is anything implicited like it's all there is agreat 
possible of confusion
rich: 4 heading would give you list of headings
DM: dont need to solve the issue today on query select
rich: logicistic issue who is going to discuss web components group
james: no need to discuss with web components
.... need to discuss in web app and css
.... need a soln before we discuss to each group
rich: who is the person that will go discuss. James could go to web apps
james: is fine with that and is a member of css
.... want to work out all the kinks. Don't see a prob with javascript api
.... there is oddities with putting selectors inside attributes bc there is 
not precedence
DM: agreed there is alot of details to work out
.... query selector can defer computation until AT request it
<jcraig> deferred computation is a good reason for the query approach
DM: potential performance win
.... once when requested inside updated continuously
s/ inside with instead
james: how would you get a reference to self
<jcraig> attr="selector(>[role=heading])"
james: child selector
.... first child heading instead of first heading
DM: document query selector
james: can happen in nested web components
.... even if you added an id to get to self, you will have to ref shadow 
root with other selector and if multiple things match how would you tell
DM: may need to add syntax for self. Need to think about it some more
rich: what are the issues that we need to discuss? (1) attribute selector 
(2) nesting of web component (3) concerned about query selector resolve to 
multiple name
.... other items?
DM: strong feelings about html only route vs javascript only route
<richardschwerdtfeger> q/
DM: does anyone have strong objections to one or other
james: no strong objections but working out kinks
.... this is a shift from typical declarative aria approach
.... there is not other language that use selectors
DM: may won't to choose one to go with first
james: reflective attributes for the rest of stuff
rich: the only thing I worry about is the test tool and how they get at 
information
.... they are use to walking the dom
DM: we would have a javascript interface to inspect
rich: how hard to get features into browsers
.... getting javascript features into browser is always a challenge
.... understand it will be hard to coordinate
DM: that is why he is open to both ideas. reach out to mozilla as soon as 
possible
james: alex and david are both on the list
rich: we would have to get this in svg also
.... if we are going to use the api route
james: element interface instead of html interface and should inherit to 
svg
rich: specific methods to say get role
james: we have specific aria to use in any language, generic mark up 
language
.... language has a possible to modify how aria efects it
rich: if you look at the dom form spec, don't have a get role
james: we will put in aria 1.1 our porposed ext to element
.... keep it into aria spec
rich: start earlier rather than later
.... dont need it to be in dom spec?
james: as long as we extend it in a std way we can keep it in aria spec
.... search html doc for "partial interface Element" and you will see it
rich: what do we do about css attribute selector that are not in dom
james: one of proposal is the role selector, computed role selector
<jcraig> CSS computed role selector :role() selector
james: proposed this last year. render h1, h2 and any element that was 
equivalent to able
s/ able / "above"
<jcraig> h1, h2, *[role="heading"] would be equivalent to *:role(heading)
rich: checkbox?
james: you can use attrib selector for it
<jcraig> reflected attrs would be something like this: el.ariaChecked = 
"true"; results in <el aria-checked="true">
james: I think IE may have had reflected attib in its first implementation
.... beneficial qualities to it
rich: lets make sure we get this covered
.... is trying to reduce the amt of javascript
.... trying to get fast and smaller
<Zakim> jcraig, you wanted to ask why you are trying to reduce JS usage
rich: adding aria attributes after teh fact and adds performance issues. 
structures we use for aria we find in uI across the board
.... we can style the UI using those attrib rather then make modifications 
to the dom
.... trying to reduce footprint and make it faster
james: clarify - you had to change the aria properties and classname on 
element
rich: the viz would change that you set something in the dom
.... alot of component lib don't make use of css capability because of 
running in old broswer
.... css scrollable div didn't exist in IE 6
.... we are seeing performance gains
.... add a11y features in begin and use it to drive style reduces the 
amount of aria code
james: ReSpec makes it easier to add new IDL interfaces
Topic issue 590
<jcraig> issue-590?
<trackbot> issue-590 -- add caption role -- raised
<trackbot> https://www.w3.org/WAI/PF/Group/track/issues/590
rich: it can be closed
<richardschwerdtfeger> Issue 590 Caption Role: Per our last discussion 
this should be closed in
<richardschwerdtfeger> favor of aria-labelledby
<richardschwerdtfeger> - https://www.w3.org/WAI/PF/Group/track/issues/590
<clown> issue-590?
<trackbot> issue-590 -- add caption role -- raised
<trackbot> https://www.w3.org/WAI/PF/Group/track/issues/590
DM: used for tables and figures
.... we dont have 1-1 mapping
.... we got rid of label in aria 1.0 bc we had labelledby
.... bc we now have this planned interface for computed role, move towards 
having a 1-1 mapping with any of the semantic html components
.... may need a generic for the ones we don't have 1-1 mapping
.... not to close issue but reinstating the label role?
rich: any objecting to introducing the label role?
matt: don't understand what the role=label buys you
DM: benefit is to inspect the dom the javascript interface
.... computing role should return heading in examples provided
.... caption has not element role mapping
matt: so what is the advantage?
rich: overtime
.... take it to the list
<clown> IA2_ROLE_PARAGRAPH
rich: response to action item 1398
<jcraig> action-1398?
<trackbot> action-1398 -- Richard Schwerdtfeger to Email list with 
discussion points for localized custom role names on known role types: 
role inheritance hierarchy, roadmap towards custom roles outlined, 
localization implications, potential for subroles etc. -- due 2014-03-24 
-- OPEN
<trackbot> https://www.w3.org/WAI/PF/Group/track/actions/1398
rich: take a look at it and discuss over next week
rrs agent, make minutes
[End of minutes]

Best regards,
Susann Keohane