- From: Janina Sajka <janina@rednote.net>
- Date: Tue, 14 Jun 2011 17:29:46 -0400
- To: wai-xtech@w3.org
Minutes from today's daylong User Agent Task Force teleconference
reviewing the User Agent Implementation Guide follow in text and are
provided as hypertext at:
http://www.w3.org/2011/06/14-pf-minutes.html
W3C
- DRAFT -
Protocols and Formats Working Group Teleconference
14 Jun 2011
See also: IRC log
Attendees
Present
Andi_Snow_Weaver, Rich_Schwerdtfeger, Janina_Sajka, Michael_Cooper, David_Bolter, Cynthia_Shelly,
Joseph_Scheuhammer, ShaneM, jcraig
Regrets
Chair
ad hoc
Scribe
janina, Joseph
Contents
* Topics
1. issue 452
2. issue 440
3. section 4 - supporting keyboard navigation
4. 4.1 controlling focus with tabindex
5. 4.2. Controlling focus with aria-activedescendant
6. 4.3. Handling focus changes from the Assistive Technology
7. 5.1. General rules for exposing WAI-ARIA semantics
8. 5.2. Conflicts between native markup semantics and WAI-ARIA
9. 5.3. Exposing attributes that do not directly map to accessibility API properties
10. 5.4. Role mapping
11. 5.5. State and Property Mapping
12. alt text and role=presentation
* Summary of Action Items
__________________________________________________________________________________________________________________
<trackbot> Date: 14 June 2011
<MichaelC> meeting: PF Daylong Teleconference on ARIA UAIG
<clown> zakim ??P37 is Joseph_Scheuhammer
<MichaelC> http://www.w3.org/WAI/PF/aria-implementation/
<MichaelC> Previous spec review walk-through
<MichaelC> scribe: janina
<scribe> scribe: janina
mc: Suggest walking the doc top to bottom, looking particularly for testable statements.
... Need also to look for discrepencies--confusing statements, etc.
as: Cynthia came up with a list of specifics for us to look at.
... I got most editorial items in.
<clown> http://www.w3.org/WAI/PF/aria-implementation/Overview.html
mc: Starting with the intro section ...
... Note it's informative, but are there thoughts?
... Discusses how ARIA fits in ... ...
... Moving on to A11y APIs, mainly an explanation of what this is
cs: Might be useful to cross-link into the official specs
as: Do we want to say we're willing to incorporate other APIs, if contributed?
cs: This is a normative doc ...
as: Perhaps an appendix after publication?
mc: Several possible approaches for that.
... "These APIs are in common use at time of publication, willing to provide additional documentation "
rs: Not aiming at mobile APIs, correct?
mc: Not yet, but that's a likely addition.
... Key may be to get agreement that without changing the finished doc, no patent implication by adding additional API
documentation
... Moving to Sec 2, DOM tree
cs: Diag would be helpful here
mc: Who could do that?
rs: One that maps DOM node to a11y object
cs: Good starting point.
<MichaelC> ACTION: Cynthia to create a diagram to support 1.2 A11Y vs DOM tree [recorded in
http://www.w3.org/2011/06/14-pf-minutes.html#action01]
<trackbot> Created ACTION-842 - Create a diagram to support 1.2 A11Y vs DOM tree [on Cynthia Shelly - due 2011-06-21].
<richardschwerdtfe> http://www.w3.org/WAI/PF/aria/accessibleelement.png
mc: Moving to Sec 2 Normative UA Reqs
... And Sec 3 Important Terms
as: A few to discuss, probably a copy edit as well
... Accessible Name -- "Who I Am;" role = "what am i"
rs: Notes to include UI Automation in A11y APIs above
as: Take out Java?
[agreement]
rs: Should be AT-SPI
as: Elsewhere we say ATK/AT-SPI
rs: Good
... Need to ask James C what to call Mac API
... Under Accessible Object, do we add visible?
... Let's say User Interface Object
... A user interface object whose basic accessibility is exposed to AT by exposing an accessibility API
cs: Think the existing is OK, don't think we need to get too scientific right here
as: Need a URI for UIA invoke pattern
[pausing to consider what to take up when JC joins call]
mc: Activation behavior? More?
<cyns> how's this
<cyns> Both Microsoft Active Accessibility and UI Automation expose the UI object model as a hierarchical tree, rooted
at the desktop. Microsoft Active Accessibility represents individual UI elements as accessible objects, and UI
Automation represents them as automation elements.
mc: Next AT ... Notes open issue re whether AT should be removed from general definition of User Agent. Now working
this question in WAI-CG.
rs: We had to separate previously, as it significantly complicates
[Notes email with UAWG on this, where they had previously removed, but discussion still in progress]
as: Also in ARIA spec?
cs: OK with encompassing both, but need separate discussion
... If we don't have rules for AT, they have no requirements on them.
as: But not this doc, this doc is not specific to AT.
... Confirmed this definition is in ARIA spec.
cs: Yes, this doc is for browsers.
mc: Checking tracker, noting no issues logged on this
... Will log an issue
<MichaelC> issue: Separation of definition of User Agent vs Assistive Technology, sometimes UA means both but many
requirements targeted just to one
[agreement]
mc: Tie to spec, or UAIG?
... issue-453 tagged to spec
... Else for AT definitions?
... Now attribs
... Class ..
as: Notes typo
<trackbot> Created ISSUE-453 - Separation of definition of User Agent vs Assistive Technology, sometimes UA means both
but many requirements targeted just to one ; please complete additional details at
http://www.w3.org/WAI/PF/Group/track/issues/453/edit .
mc: Event
... Hidden
clown: Q, notes note discrepency with text
mc: only considered from AT perspective? Is that the meaning?
cs: Yes, ARIA-hidden and display=none differences are somewhat confusing
... When does one need both
... Certainly not both at all times
as: Notes spec uses the same definition
<MichaelC> http://www.w3.org/WAI/PF/aria/states_and_properties#aria-hidden
mc: If browsers not exposing display=none to A11y API, does this guide say anything about that? Create an issue to
explore this?
rs: HTML 5 has an element that does the same thing.
cs: All three seems unnecessary
rs: It's for DOM based ATs
cs: At least one AT ignores display=none in the belief that authors misuse it
rs: At some point we could deprecate aria-hidden
clown: but might use with svg
cs: Need to clarify whether this is for this doc, or authoring doc
[discussion]
mc: Does the UA need to do something? Sop, should be in UAIG, but perhaps also tweak to spec and author doc
<cyns> Here's another definition: An accessible object is any UI element that implements the IAccessible interface. To
interact with the accessible object, an Active Accessibility client needs only this IAccessible object.
<cyns> we could change that to "An accessible object is any UI element that implements an accessibility API "
<MichaelC> issue: Does aria-hidden have to be set on any non-displayed element, or does e.g., CSS display:none,
visibility:hidden, or HTML hidden also cause the same behavior in the user agent? If so, UAIG should specify which
should happen, and spec may need clarification.
<trackbot> Created ISSUE-454 - Does aria-hidden have to be set on any non-displayed element, or does e.g., CSS
display:none, visibility:hidden, or HTML hidden also cause the same behavior in the user agent? If so, UAIG should
specify which should happen, and spec may need clarification. ; please complete additional details at
http://www.w3.org/WAI/PF/Group/track/issues/454/edit .
<cyns> accessible object defintion is from http://msdn.microsoft.com/en-us/library/ms971310.aspx#actvaccess_topic2
as: Looking for someone to help untangle this
mc: Moving on ...
as: Live regions -- Need editorial reading on how to use AJAX
cs: A W3C style guide on this?
mc: AJAX not in it, we need to try and get it added
clown: Notes typo in third sentence
as: Already flagged
<MichaelC> ACTION: cooper to raise question of AJAX vs Ajax as W3C style decision and circle back when decision made
[recorded in http://www.w3.org/2011/06/14-pf-minutes.html#action02]
<trackbot> Created ACTION-843 - Raise question of AJAX vs Ajax as W3C style decision and circle back when decision made
[on Michael Cooper - due 2011-06-21].
mc: Moving on ...
cs: Managed State -- Doesn't seem wrong, but doesn't seem to say very much
<MichaelC> ACTION: cynthia to clarify definition of "managed state" [recorded in
http://www.w3.org/2011/06/14-pf-minutes.html#action03]
<trackbot> Created ACTION-844 - Clarify definition of "managed state" [on Cynthia Shelly - due 2011-06-21].
as: Object, seems to need work
mc: Is redundant
as: DOM element is discussed
mc: We wanted definition to distinguish from dom object, or object in programming lang
... Believe we mean something closer to a widget than an object
cs: We do?
mc: We mean conceptual types of object, not lang classes
... This has history from when ARIA was perceived as RDF model
clown: so a landmark is an object?
mc: yes
rs: You want to know it's in the hierarchy as it provides a tontainer
clown: Unit of perception?
<cyns> from webopedia "Generally, any item that can be individually selected and manipulated. This can include shapes
and pictures that appear on a display screen as well as less tangible software entities. In object-oriented
programming, for example, an object is a self-contained entity that consists of both data and procedures to manipulate
the data. "
<davidb> "In the domain of object-oriented programming an object is usually taken to mean an ephemeral compilation of
attributes (object elements) and behaviors (methods or subroutines) encapsulating an entity"
mc: We're getting into philosophical weeds
<MichaelC> issue: We don't know what the definition of "object" means
<trackbot> Created ISSUE-455 - We don't know what the definition of "object" means ; please complete additional details
at http://www.w3.org/WAI/PF/Group/track/issues/455/edit .
<richardschwerdtfe> In http://en.wikipedia.org/wiki/Computer_science, an object is any entity that can be manipulated
by the commands of a http://en.wikipedia.org/wiki/Programming_language, such as a
http://en.wikipedia.org/wiki/Value_(computer_science), http://en.wikipedia.org/wiki/Variable_(computer_science),
http://en.wikipedia.org/wiki/Subroutine, or http://en.wikipedia.org/wiki/Data_structure. (With the later introduction
of
<richardschwerdtfe> http://en.wikipedia.org/wiki/Object-oriented_programming the same word, "object", refers to a
particular instance of a http://en.wikipedia.org/wiki/Class_(computer_science))
rs: manipulable instantiation of an element
<cyns> Issue-455: An accessible object is any UI element that implements an accessibility API. Can we define object as
"UI element"?
<trackbot> ISSUE-455 We don't know what the definition of "object" means notes added
RE Issue-455 -- Webster]
2. Anything which is set, or which may be regarded as set,
before the mind so as to be apprehended or known; that of
which the mind by any of its activities takes cognizance,
whether a thing external in space or a conception formed
by the mind itself; as, an object of knowledge, wonder,
fear, thought, study, etc./me Hi, James. We're on Zakim 92473# -- the usual PF channel
[JC has joined]
[we go to specific questions we wnated to talk with jc about]
as: Sec 5.4 -- case?
mc: Now ARIA tokens, we can say about case
... so matches host lang
[agreement]
<jcraig> link?
<jcraig> http://www.w3.org/WAI/PF/aria-implementation/#mapping_additional_nd_name
as: Notes this note only in UAIG
<clown> scribe: Joseph
<MichaelC> scribeNick: clown
CS: compute the text alternative, and it comes up empty
... look for various label attributes (e.g., aria-label, title)
... it implies that these attributes were empty too.
... reading it again, I understand what it is trying to say.
MC: we don't need to worry about this anymore and move on.
CS: yes.
MC: any issues under the description computation?
ASW: text alt, bullet 2.
... 2A, first unnumbered bullet.
... in the middle, it says, "However, the element's aria-labelledby attribute can reference the element's own IDREF in
order to concatentate a string provided by the element's aria-label attribute or another feature lower in this
preference list."
... cynthia's comment: not sure why that is needed.
JC: discussion a while back to change the precedence of aria-label and aria-labelledby.
... from a specific public comment. need to look it up.
CS: why do we need to point to ourself?
JC: I'll paste in an example.
<jcraig> <img aria-label="Delete" id="me" aria-labelledby="me somethingElse">
<jcraig> <div id="somethingElse"> Item 1</div>
MC: I found the issue. ISSUE-344
<MichaelC> ISSUE-334: Inconsistency between spec and implementation guide
JC: for the example, the label would be "Delete Item 1".
<jcraig> no
<jcraig> "Delete item 1"
CS: can we put this example in?
JC: I think there is a better one in the issue.
... the idea is can repeat the "Delete" part for concatenating with other referenced text.
<jcraig> Item 1 [Edit] [Delete]
<jcraig> Item 2 [Edit] [Delete]
<jcraig> etc
MC: looking for the public comment for any example there.
<jcraig> Where the Delete and Edit buttons included the visible string label of the item
<MichaelC> aria-label & aria-labelledby mutually exclusive?
CS: I'm fine with that if we can put in an example.
<jcraig> combined with, not versus
<MichaelC> ACTION: jcraig to create example for text alternatives computation 2.A first bullet [recorded in
http://www.w3.org/2011/06/14-pf-minutes.html#action04]
<trackbot> Created ACTION-845 - Create example for text alternatives computation 2.A first bullet [on James Craig - due
2011-06-21].
ASW: not changing the text, just adding an example?
MC: yes.
ASW: next bullet: "if this computation is already occurring as the result of a recursive text alternative computation
and the current element is an embedded control as defined in rule 2B, ignore the aria-label attribute and skip directly
to rule 2B."
... cynthia's quesion is why ignore?
JC: gives an example.
CS: so, have an example would make things clearer.
JC: I think that example is in the spec.
ASW: should I copy those examples into the UIAG?
CS: or, link to them.
RS: linking might be easier
... don't have to maintain two copies.
ASW: could do that for the whole section.
RS: yes, why not?
<jcraig> spec example #2 is for the 2B embedded form controls: http://www.w3.org/TR/wai-aria/complete#tac_example2
ASW: 2B, third bullet. "If the embedded control is a select or combobox, use the chosen option."
... cynthia's question: what about aria-labelledby?
... this section says you include the embedded option. it doesn't say you don't use aria-label no aria-labelledby.
JC: It does say to skip the aria-label for embedded controls. But not aria-labelledby.
... I can type in another example.
<richardschwerdtfe> me too
<jcraig> <img tabindex="0" role="checkbox" aria-labelledby="text textfield"> <span id="text">Friend referral</span>
<input type="text" id="textfield" value="John Smith" aria-label="Name of friend">
CS: should we just record an issue and move on?
<jcraig> So if you landed on the radio button, you'd hear "Friend referral John Smith, radio button 2 of 2"
RS: what are we ignoring here?
<jcraig> but if you landed on the text field, you'd hear "Name of friend, John Smith, <editing>"
JC: think of this as a list of "how did you hear about us?"
... with a radio button named "Friend referral", with a text field for you to fill in.
... so the aria-label is redundant with the ?
ASW: is this example in the spec?
JC: we can use this example, but there is an example in the spec. see the link I put in irc earlier.
JS: that's a very similar example.
s/JS: that's a very similar example./JC: that's a very similar example./
ASW: under 2d: should this still be the last resort in HTML5?
CS: I don't know if we can answer this now, but we should keep an issue open.
JC: the text alt computation here is not quite the same as that in the spec.
... Michael, do you have an action for common source?
MC: yes
ASW: bullet 3: should add a bullet for before and after, and add a bullet for CSS.
... we have rules for replacing, but not for adding text.
JC: we could say something about append/prepends for before and after?
CS: for before and after, and list rendering.
JS: what means list rendering?
CS: the numbers in an ordered list, or changing the numbering (different start points).
<cyns> that was list numbering
JC: is this the covered by ?
CS: so, when new text is added and when numbers are added.
... when numbered styles are added.
JC: summarizes the changes.
<jcraig> change "When new text replaces old" to "When new text is added"
CS: can we add list styles without it being a normative change? It is important to know what number of item you are on.
JC: we might just say the content property.
CS: can someone take an action to go figure that out?
<jcraig> :before { content: 'foo'; }
CS, ASW, JC: discussion of css styles for list "bullets".
JC: I thought you were describing that you could redefine the start.
CS: whether it's a letter or number is important too.
JC: I don't think that's well supported by all browsers.
<jcraig> Action should reference http://www.w3.org/TR/css3-content/
<trackbot> Sorry, couldn't find user - should
MC: James N, we are reviewing the text alternative computation.
<MichaelC> ACTION: jcraig to clarify how CSS list numbering fits into text alternative computation in bullet 3
[recorded in http://www.w3.org/2011/06/14-pf-minutes.html#action05]
<trackbot> Created ACTION-846 - Clarify how CSS list numbering fits into text alternative computation in bullet 3 [on
James Craig - due 2011-06-21].
ASW: that's all for the name calculation.
<richardschwerdtfe> woot
ASW: wanted also ask jame about live regions.
<MichaelC> http://www.w3.org/WAI/PF/aria-implementation/#mapping_events_visibility
JC: are you talking about the notifications I sent to you?
ASW: the other a11y APIs don't have a special event.
JC: you think this should go into another table.
ASW: if you look at 5.8.2, I added to the table -- if it's in alive region, the event is "ax live regions changed".
JC: it will also have another ax event if the text node is going away completely.
... it's only if it's removed completely that you get the other event.
... this is fine.
... but let me think about it.
ASW: could you review it, and let us know if it's okay.
... after the second one (when a subtree is hidden), towards the end, it talks about the different APIs.
... I added the mac event there.
JC: I'm not sure about the very last one (required)>
... yeah, it sounds good, but I'll look it over.
ASW: that's all of the questions we needed James C's input.
CS: do you have any comments, James?
JC: not right now. I reviewed a while back, and sent in comments.
ASW: the live region was the last one from that set.
JC: I'll give it one more review.
MC: let's go back to the definitions.
JS: do we have a name for the the Mac a11y API?
JC: give the names.
ASW: we have it right.
CS: is the api on iOS the same?
JC: it's called UIAccessibility.
<jcraig> long name: Mac OS X Accessibility Protocol
<jcraig> short name: AX API
<jcraig> iOS: UIAccessibility
MC: ontology. might be a generator error. doesn't occur in the spec.
... next: owned element.
... owning element.
... perceivable
... property
... relationship
... do we need that? it's an rdf-ish thing.
RS: why do you think that?
MC: let me think -- okay.
ASW: is 'articulable" a word?
... let's just drop "articulable" --" two distinct things"
MC: let's do that.
... role.
... Root WAI-ARIA node
ASW: captilaize first word.
SM: that's not right. In html it's the document.
... make sure it's clear.
MC: maybe: "the element containing non meta-data element".
... which, in html, is the body or frameset.
<janina> Websters says first known use of ARTICULABLE was in 1833
MC: would a framset have aria on it?
<ShaneM> "The element containing non-metadata content. In HTML the body or frameset element. In other languages this
may be the document element."
JN: not normally, but it could.
MC: moving on, "semantics"
... State
JS: where is markup exposed to APIs?
... is that part of semantics? or elsewhere?
MC: not in semantics, but a clarification under describedby.
ASW: are you alluding to where describedby points to markup?
JS: yes.
MC: Sub-document.
ASW: should an iframe be markup up as code?
MC: probably
... understandable
... User Agent. already created an issue -- anything further?
... valid id
RS: not need for "relationship" here.
ASW: where you do specify a relationship with an id, then the def'n applies.
RS: it's an ID that exist on an element on a document.
... does "valid" mean if you have the same ID, does that work?
MC: I think we should change it to ID reference.
<Andi> Valid IDREF: A reference to a target element in the same document that has a matching ID.
RS: what happens if you have more than one element with the same ID.
?
RS: would that be "valid"?
ASW: author error.
RS: we look for duplicate id's for failure testing.
MC: some UAs will choose the first one, some will chose the last one.
JN: its' a wcag error as well.
MC: not sure that we need to worry about that here.
RS: that's sounds reasonable, Andi.
MC: do we need an issue to address multiple non-unique ids?
RS: what happens when you create a relationship between two objects? do you take the first one?
JN: it's undefined. different UAs do different things.
SM: html5 probably defines what happens.
JN: but only for html.
... definitely a failure, but we can't define every different failure.
RS: need to know what the mapping is so you can test it. if it takes the first one, fine.
MC: we need to do some testing of duplicate IDs?
RS: yes, but we need to know what the browsers are supposed to do.
MC: let's create an issue.
<ShaneM> in the access attribute specification, we said "Note: since the id of an element must be unique within a valid
XML document, in such documents, each element group based on targetid values consist of no more than one matching
element."
JN: that's a lot of tests: aria-labelledby, aria-owns, etc.
ASW: yes, that's a lot of tests.
SM: just say, "don't do this".
ASW: notes that text says that UAs only do some testing and error handling.
... the UA wouldn't do anything if there were duplicate IDs.
MC: I'm okay with that.
... moving on. "Value".
... Widget
ASW: cynthia thought we should contrast with Object and Accessible.
MC: we should
<janina> 'solidify' is a good word
MC: element is the thing with angle brackets around it, but difference between widget and object is not clear.
ASW: let's include this on the earlier issue on Object.
MC: issue 455
<MichaelC> issue-455: need to include widget in the consideration of this confusion
<trackbot> ISSUE-455 We don't know what the definition of "object" means notes added
MC: end of glossary.
15 minute break
s/JS: that/JC: that/
<janina> Now waiting on "Baked Ziti with Beef"
<ShaneM> lol - sounds like the name of a sopranos episode
<ShaneM> scribenick: ShaneM
<janina> Michael??
issue 452
need james and rich.
issue 440
Andi: Discussed into the UAI task force and came up with a compromise.
... at the F2F we discussed whether the user agents are required to take the elements out of the tree.
<Andi> http://www.w3.org/WAI/PF/aria-implementation/#include_elements
<cyns> shane, that's Andi speaking
<Andi> http://lists.w3.org/Archives/Member/w3c-wai-pf/2011AprJun/0143.html
clown: What happens if there is a link (aria-owns) to a node that becomes hidden? In agents where the node is removed
what happens to the owned link?
cyns: IE removes the node.
... if the item is no longer in the tree, it would just 'not work'
Firefox keeps the item in the tree and marks it as hidden. So the owns then points to something that is marked hidden.
davidb: the relation is not exposed if the system cannot find the target element.
cyns: Is the issue that it is expensive to reconstruct the tree when the item becomes unhidden?
davidb: not really. There are a lot of consistency issues with relations. With focus going into the tree. And there is
also a "what problem are we trying to solve?" question.
cyns: This comes down to the difference between aria-hidden and display:none. Lots of odd behaviors
... IE takes lots of things out of the tree. not 'visibility:hidden' though. The visual rendering engine ignores
'aria-hidden'
Joseph proposes: if an author aria-hidden to display:none, it is not in the tree. With visibility:hidden it does.
People seem to think this is good guidance / a good hack.
davidb: The use case I remember is a case where you popup a dialog - use aria-hidden to remove the background page from
the accessibility tree while the dialog was 'over' it.
... there might be other ways to do that, but it was a use case from James Craig that he thought was important.
jamesn: another is where there is a complex table where there is a header table. You hide them visually but make them
available.
... a non-interactive table that is a scrollable region. The header cells are not part of the same 'table' to get the
presentation that is desired.
... so you embed the headers in the scrolling area but mark them display:none so they don't show up on the visible
page...
... It could have been done with a 'table' role and divs and spans, but there isn't one.
cyns: it is usually the case that you want the visible content to match the accessibity tree content. This is sort of
an edge case.
jamesn: There are simpler cases too.... like where text is repeated but you want to hide duplicated text from the
screen reader to assist the user.
cyns: Alot of these cases are just bad web site design.
jamesn: but isn't what ARIA trying to do to help ensure that bad design is still accessible?
cyns: I don't think it is the end of the world that these are different.
davidb: aria-owns is just one piece. It is not even the strongest thing that is pushing back from the firefox side.
The spec says that you can leave them in if you want to. It used to say take them out.
richardschwerdtfe: webkit and IE remove it from the tree. Firefox does not.
davidb: I am pretty sure opera does not remove it from the tree.
... I can try to make the change in firefox. It is a team decision. I can try to get it in or the spec is going to have
to offer both (or we will be non-confirming)
richardschwerdtfe: what about the hidden element?
davidb: I will try again. interoperability is a good rationale. If IE and webkit are not going to change that helps my
argument.
... concern about what happens if focus lands in an area that is hidden (and therefore is not in the accessibility
tree)
cyns: can you tell is what those edge cases are just to be sure we are doing the right thing by removing the nodes?
davidb: yes, I will get the cases from the bug and email it to the group.
<richardschwerdtfe> +1 to remove
<Andi> issue 452
closing issue 452 - group agrees that the advisory can be removed.
<Andi> close by removing the advisory from aria-hidden row of state property table
section 4 - supporting keyboard navigation
hook for the tabindex thing in HTML. Also ensures we do what UAAG says we should do.
4.1 controlling focus with tabindex
MichaelC: we are now producing an HTML specific implementation guide. Should we move this to that document before we go
to last call?
... add an editorial note that indicates the section might migrate to the HTML specific implementation guide
4.2. Controlling focus with aria-activedescendant
MichaelC: any special testing requirements?
richardschwerdtfe: we need to ensure an accessbility focus is handled.
... what about when IDs are invalid? We need to be consistent.
MichaelC: If there are test requirements in the implementaiton guide vs. WAI-ARIA should we have separate tests or
should it just be in a single test harness / test case?
There might be some similarity between tests cases, but it would be cleaner to keep them separate.
cyns: subsection 4.c. remove the 'true' before the word DOM.
4.3. Handling focus changes from the Assistive Technology
might want to refer authors to the accessibility API documentation for the platform.
cyns: The 'may statement' add the term 'DOM'.
... add a third bullet?
<cyns> If the current element has an ID and an ancestor with the aria-activedescendant attribute present, the user
agent MUST set the accessibility API focused state and fire an accessibility API desktop FOCUS event on the new active
descendant.
<cyns> Missing from both 4.2 and 4.3 is clearing the focused state from the old focused element (which may be another
activedescendant) in *both* the platform accessibility API and the DOM.
(discussion about how dom focus and cleanup works)
cyns: The concern is that there might be two elements with the focus attribute in the AT implementation.
<clown> document.activeElement
richardschwerdtfe: Not in the DOM though. There is no way to have that happen. There is an activeElement attribute and
there can only be one.
... This is something the accessibilty API might need to do, but it is not something the DOM needs to worry about.
cyns: Then let's add something about ensuring that the accessibility API clears the previous focus so there are never
two.
Andi: The only place we talk about this is in section 4.2
might need a table to show HOW to clear this on each platform.
<richardschwerdtfe> https://developer.mozilla.org/en/Accessibility/AT-APIs/MSAA/States
5.1. General rules for exposing WAI-ARIA semantics
5.2. Conflicts between native markup semantics and WAI-ARIA
There are two sentences that might be removed, but we are not going to decide that here nor today.
5.3. Exposing attributes that do not directly map to accessibility API properties
MichaelC: should we have something in here about Mac?
5.4. Role mapping
cyns: concern about forward compatibility of patterns since some may be abstract roles
<MichaelC> issue: Forward compatibility and overlap between abstract roles and patterns
<trackbot> Created ISSUE-456 - Forward compatibility and overlap between abstract roles and patterns ; please complete
additional details at http://www.w3.org/WAI/PF/Group/track/issues/456/edit .
<Andi> http://www.w3.org/WAI/PF/Group/track/issues/445
<MichaelC> issue-456: see also ISSUE-445
<trackbot> ISSUE-456 Forward compatibility and overlap between abstract roles and patterns notes added
cyns: change 'microsoft UIA control pattern' to 'UIA control pattern'
5.5. State and Property Mapping
Is bullet 3 really needed?
Group agrees it is not needed here as it is already in the previous section.
MichaelC: bullet 4 - the sentence that starts "Where default values..." should be a separate numbered item.
Andi: This is really a clarification of the beginning of item 4, so it should stay associated with it.
bullet 6 - discussion about whether it makes sense to permit dropping of properties / states that are not supported by
the element in question
richardschwerdtfe: do we have to test SHOULDs?
MichaelC: No. Not for CR.
richardschwerdtfe: I am pretty sure you can expose anything.
MichaelC: should we log an issue?
<MichaelC> issue: Do AAPIs support mapping of unsupported states and properties on roles, and do we want UAs to attempt
those mappings or discard them?
<trackbot> Created ISSUE-457 - Do AAPIs support mapping of unsupported states and properties on roles, and do we want
UAs to attempt those mappings or discard them? ; please complete additional details at
http://www.w3.org/WAI/PF/Group/track/issues/457/edit .
Note after bullet 7 should probably be moved after bullet 5.
<MichaelC> ACTION: cooper to create script to hide columns at user choice in UAIG tables to yield less wide tables
[recorded in http://www.w3.org/2011/06/14-pf-minutes.html#action06]
<trackbot> Created ACTION-847 - Create script to hide columns at user choice in UAIG tables to yield less wide tables
[on Michael Cooper - due 2011-06-21].
<janina> scribe: janina
mc: resuming 5.6 -- values
as: we have an open issue on this
... issue-442
<Andi> http://www.w3.org/WAI/PF/Group/track/issues/442
as: need guidance on this
<clown> http://www.w3.org/WAI/PF/aria/states_and_properties#aria-valuetext
<clown> If aria-valuetext is specified, assistive technologies SHOULD render that value instead of the value of
aria-valuenow.
[nix that -- ua vs at]
mc deferred to Monday ARIA Caucus call
mc: moving on ...
... id refs ...
as: multiple ids?
mc: OK on text?
as: Need to ask David
... Andi to find usecase for multiple same elements -- Sec 5.6.3.1
mc: moving on ... 5.6.3.2
as: Need to ask David about the ATK examples. They don't appear correct.
Media proposal at:
http://lists.w3.org/Archives/Public/public-html-a11y/2011May/0466.html
<MichaelC> issue: How should rich text be mapped to AAPIs in descriptions, given that currently AAPIs only support flat
text?
<trackbot> Created ISSUE-458 - How should rich text be mapped to AAPIs in descriptions, given that currently AAPIs only
support flat text? ; please complete additional details at http://www.w3.org/WAI/PF/Group/track/issues/458/edit .
mc: moving on ...
as: how to compute aria level when not provided ...
... this sections says how
mc: had trouble parsing the sentence ...
... moving on ... actions ...
... not author responsibility to set click handlers?
cs: yes, but any action has to call something which needs to hook into something
... so, it's at simulating click
... Noting that UIA doesn't have default action concept
<MichaelC> ACTION: cynthia to define default actions in UAIG 5.7 Actions [recorded in
http://www.w3.org/2011/06/14-pf-minutes.html#action07]
<trackbot> Created ACTION-848 - Define default actions in UAIG 5.7 Actions [on Cynthia Shelly - due 2011-06-21].
mc: Now looking at following table ...
... Seems to need some editorial cleanup?
... Now 5.8 events
mc .1 state and property
<MichaelC> ACTION: cynthia to provide additional background in para on UIA events in UAIG 5.8.2. Changes to document
content or node visibility [recorded in http://www.w3.org/2011/06/14-pf-minutes.html#action08]
<trackbot> Created ACTION-849 - Provide additional background in para on UIA events in UAIG 5.8.2. Changes to document
content or node visibility [on Cynthia Shelly - due 2011-06-21].
mc: moving on ...
jn: Jumping back to table in 5.8 ...
... Some things missing, like mac notification, will these be filled in?
cs: Good point, we should check
mc: Past second table ... re node changes ...
... Do we know the meaning of "appropriate events" ?
cs: events as specified in the table?
... Perhaps change on parent without accessible event, but children have accessible events
... Hmmm, maybe not.
[discussion, we're confused by this section]
as: retained from a doc in 2009 ...
cs: next para is about node deletion
rs: perhaps about deleted nodes?
... so change events for something that's been removed
... first line descendants should be 'immediate descendants'
... Should review with David
... problem is just the two bullets
<MichaelC> ACTION: bolter to clarify meaning of node changes bullets in 5.8.2. Changes to document content or node
visibility [recorded in http://www.w3.org/2011/06/14-pf-minutes.html#action09]
<trackbot> Created ACTION-850 - Clarify meaning of node changes bullets in 5.8.2. Changes to document content or node
visibility [on David Bolter - due 2011-06-21].
mc: moving on ...
as: Just above 5.8.3 ...
... Only applicable to ia2 and at-spi ...
mc: editorial -- smartquotes around 'true'
... moving on ...
clown: strangely worded section -- should? may? must? caps? etc
as: not clear what's required, vs recommended, etc.
mc: yes, needs editorial pass
... content seems ok, though
mc 5.8.4
<clown> proposed: Joseph to look through APG design patterns re: menus for examples that relate to UAIG 5.8.4
<clown> ACTION: Joseph to look through APG design patterns re: menus for examples that relate to UAIG 5.8.4 [recorded
in http://www.w3.org/2011/06/14-pf-minutes.html#action10]
<trackbot> Created ACTION-851 - Look through APG design patterns re: menus for examples that relate to UAIG 5.8.4 [on
Joseph Scheuhammer - due 2011-06-21].
as: Need to skip 6.1 as David and I have a bug to work on it
mc: So selectors ...
rs: Not sure why it's here
... propose deleting the section
mc: suggest an editorial note: "this section may be unnecessary"
clown: it's a 'must' on something they should already be doing
rs: might some ua actually remove these?
cs: concerned about new browsers implementing a11y
... don't think having this section hurts anything
mc: let's leave for now
... 6.3 author errors
clown: notes another 'may' in lower case
mc: Think we've achieved our goals for today. Not many testable statements, though.
<MichaelC> drop item 1
<MichaelC> drop item 2
<MichaelC> drop item 3
rs: our agenda+ for alt text with role=presentation
<MichaelC> drop item 4
<MichaelC> drop item 5
alt text and role=presentation
rs: some statement in this doc, perhaps, to deal with role=presentation, alt='foo'
[discussion of the obvious error]
[suggestion that aria always overides]
jn: not convinced it's necessarily an error condition, e.g. role presentation on image but value specified elsewhere,
perhaps an aria-label; therefore you need the alt for the nonaria ua
[agreement to deal with this in this doc as long as no conflict with html 5 spec docs]
cs: Suggest best is to be safe for the user and honor the alt
[others disagree]
<clown> if the element is not marked as presentational (role="presentation"), check for the presence of an equivalent
host language attribute or element for associating a label, and use those mechanisms to determine a text alternative.
cs: also the responsibility of browsers to catch author erros, where possible, and to protect users from them, where
possible.
jn: disagree it's necessarily an error
cs: my recollection is this is considered an author error because of validation
<jamesn> How about this: <img alt="required" role="presentation"><input aria-required="true">
rs: validators of today becoming almost useless
cs: because they look at static content?
rs: yes
cs: so validator against the dom snapshot
rs: and constantly monitor the page as one operates it
... so not a dtd
cs: could
<ShaneM> just because the html5 people think DTD based validation is useless does not make it so.
Summary of Action Items
[NEW] ACTION: bolter to clarify meaning of node changes bullets in 5.8.2. Changes to document content or node
visibility [recorded in http://www.w3.org/2011/06/14-pf-minutes.html#action09]
[NEW] ACTION: cooper to create script to hide columns at user choice in UAIG tables to yield less wide tables [recorded
in http://www.w3.org/2011/06/14-pf-minutes.html#action06]
[NEW] ACTION: cooper to raise question of AJAX vs Ajax as W3C style decision and circle back when decision made
[recorded in http://www.w3.org/2011/06/14-pf-minutes.html#action02]
[NEW] ACTION: cynthia to clarify definition of "managed state" [recorded in
http://www.w3.org/2011/06/14-pf-minutes.html#action03]
[NEW] ACTION: Cynthia to create a diagram to support 1.2 A11Y vs DOM tree [recorded in
http://www.w3.org/2011/06/14-pf-minutes.html#action01]
[NEW] ACTION: cynthia to define default actions in UAIG 5.7 Actions [recorded in
http://www.w3.org/2011/06/14-pf-minutes.html#action07]
[NEW] ACTION: cynthia to provide additional background in para on UIA events in UAIG 5.8.2. Changes to document content
or node visibility [recorded in http://www.w3.org/2011/06/14-pf-minutes.html#action08]
[NEW] ACTION: jcraig to clarify how CSS list numbering fits into text alternative computation in bullet 3 [recorded in
http://www.w3.org/2011/06/14-pf-minutes.html#action05]
[NEW] ACTION: jcraig to create example for text alternatives computation 2.A first bullet [recorded in
http://www.w3.org/2011/06/14-pf-minutes.html#action04]
[NEW] ACTION: Joseph to look through APG design patterns re: menus for examples that relate to UAIG 5.8.4 [recorded in
http://www.w3.org/2011/06/14-pf-minutes.html#action10]
[End of minutes]
__________________________________________________________________________________________________________________
ScribeNicks: janina, clown, ShaneM
Present: Andi_Snow_Weaver Rich_Schwerdtfeger Janina_Sajka Michael_Cooper David_Bolter Cynthia_Shelly Joseph_Scheuhammer ShaneM j
craig
People with action items: bolter cooper cynthia jcraig joseph
--
Janina Sajka, Phone: +1.443.300.2200
sip:janina@asterisk.rednote.net
Chair, Open Accessibility janina@a11y.org
Linux Foundation http://a11y.org
Chair, Protocols & Formats
Web Accessibility Initiative http://www.w3.org/wai/pf
World Wide Web Consortium (W3C)
Received on Tuesday, 14 June 2011 21:30:30 UTC