Re: Minutes: [aapi] UAI TF Meeting Tue Jul 23 2013

Michael,

I am reading the minutes. I saw you seemed to not fully understand the tie
in between MSAA and extensions to it like IA2 or UIA Express or at least
not well enough to explain to the director.

Let me clarify for you:

MSAA is the first accessibility API for Windows. It was insufficient to do
Rich Internet Applications. Also, because MSAA has deficiencies different
browser providers changed their implementation of it slightly to work
around shortcomings. Also, when extensions were added to MSAA the
implementation of MSAA varied slightly to accommodate the extensions. So,
to test a "platform implementation" you must test MSAA with its extension
to meet compliance with the ARIA spec.

These are the two implementations:

1. MSAA+IA2 - MSAA based on the Windows COM (Component Object Model) based
broker model that an AT can use ask the MSAA accessibility information
from. IA2 is an extension to it. IA2 provides special Table interfaces,
relationships, extended roles and states and object attributes as an
example. This is implemented on both Firefox and Chrome.

2. MSAA+UIAExpress - This uses a slightly different implementation of MSAA
and something called UIAExpress. UIA is divided into a provider side
(application side) and a client side (AT) API. Bridging technology maps
both MSAA and UIAExpress (provider side) to a client side API. To get the
information COM provides the inter process communication transort layer for
MSAA and the bridge provides the inter process communication on the client
side to get information from the provider side implementation. It should
also be noted that Windows also bridges between MSAA and the UIA Client
side API. Unfortunately, not all information is bridged to the UIA client
side for IE. So, in order to get access to the necessary information you
will see that the mapping tables use a combination of MSAA information and
UIA Client side API interfaces to get the data.

UIA - UIA would require a full client side implementation of IE. Ideally
the provider side and client side would be full implementations of provider
and client side interfaces. Doing this, logistically, is a problem as you
would want to have assistive technologies that are fully dependent on UIA
to truly exercise it. With the exception of Narrator (which is supposed to
use all UIA but given what I have told you that is unlikely for IE), the
legacy AT vendors primarily use the DOM with some smatterings of MSAA, UIA,
and UIA Express to access IE. This stems from the fact that they have had
to support older versions of IE (IE6 for example) which only provided DOM
and MSAA support and gaps in the bridge. Consequently there are gaps in the
UIA implementation. This is why when you look at the MSAA+UIA Express
column you see a combination of MSAA and UIA Client side interfaces being
used as the normative reference to accessing the IE platform. Over time I
expect this to look more like the UIA column bridging gaps are filled.

The combination of MSAA and the extension API should be considered a single
platform implementation. MSAA can not be taken on its own and due to the
variations in its implementation it cannot be treated as a normative
platform implementation by itself.

I hope this helps.

Rich



Rich Schwerdtfeger



From: Joseph Scheuhammer <clown@alum.mit.edu>
To: "wai-xtech@w3.org" <wai-xtech@w3.org>,
Date: 07/23/2013 03:14 PM
Subject: Minutes: [aapi] UAI TF Meeting Tue Jul 23 2013



Link: http://www.w3.org/2013/07/23-aapi-minutes.html


Plain text version follows:

    [1]W3C

       [1] http://www.w3.org/


                                - DRAFT -

            Protocols and Formats Working Group Teleconference
                               23 Jul 2013

    [2]Agenda

       [2] http://lists.w3.org/Archives/Public/wai-xtech/2013Jul/0009.html


    See also: [3]IRC log

       [3] http://www.w3.org/2013/07/23-aapi-irc


Attendees

    Present
           Joseph_Scheuhammer, Michael_Cooper, Cynthia_Shelly

    Regrets
           David_Bolter

    Chair
           Joseph_Scheuhammer

    Scribe
           clown

Contents

      * [4]Topics
          1. [5]Cynthia and RIch's meeting.
          2. [6]aria-checked is undefined.
          3. [7]ISSUE-583
      * [8]Summary of Action Items
      __________________________________________________________

    <trackbot> Date: 23 July 2013

    trackbot, make log public

    <trackbot> Sorry, clown, I don't understand 'trackbot, make log
    public'. Please refer to
    <[9]http://www.w3.org/2005/06/tracker/irc> for help.

       [9] http://www.w3.org/2005/06/tracker/irc%3E


    trackbot, make logs public

    <trackbot> Sorry, clown, I don't understand 'trackbot, make
    logs public'. Please refer to
    <[10]http://www.w3.org/2005/06/tracker/irc> for help.

      [10] http://www.w3.org/2005/06/tracker/irc%3E


    <scribe> scribenick: clown

Cynthia and RIch's meeting.

    CS: Met with Rich
    ... went over tests, got some passing, made recommedations for
    changes.
    ... Joseph has made the changes.

    JS: Yes.

    CS: I need to got back and redo some of the tests.
    ... I"m also setting up a Windows 8 machine in the hope that
    ones that fail in IE10 might pass in IE11.

    <scribe> agenda:
    [11]http://lists.w3.org/Archives/Public/wai-xtech/2013Jul/0009.

    html

      [11] http://lists.w3.org/Archives/Public/wai-xtech/2013Jul/0009.html


aria-checked is undefined.

    [12]http://www.w3.org/WAI/PF/aria-implementation/#mapping_state

    -property_table

      [12]
http://www.w3.org/WAI/PF/aria-implementation/#mapping_state-property_table


    JS: in the states/properites table, I think apple is the only
    one correct.

    CS: if this is a checkbox, it has to be checkable and
    unchecked.

    JS: but this the general case.
    ... What should it say if not a check box?

    CS: the way it's described here is for a checkable object.

    JS: I think it should state for a non-checkable object.

    CS: Then it should be "Not mapped".

    JS: Like apple?
    ... I think we could add to the lef column "treat as false if
    checkbable item", where it lists checkbox, radio, etc.

    CS: There is a list of the ones that require a checkable state.

    JS: this will allow me to finiish a test case for aria-checked.

ISSUE-583

    ISSUE-583?

    <trackbot> ISSUE-583 -- Elements that are descendants of an
    element having aria-activedescendant should not all be
    focusable -- open

    <trackbot> [13]https://www.w3.org/WAI/PF/Group/track/issues/583


      [13] https://www.w3.org/WAI/PF/Group/track/issues/583


    [14]https://lists.w3.org/Archives/Member/w3c-wai-pf/2013JulSep/

    0118.html

      [14]
https://lists.w3.org/Archives/Member/w3c-wai-pf/2013JulSep/0118.html


    CS: I disagree.
    ... We had giant discussion about this, and we finally reached
    a compromise.
    ... I do not want to re-open the topic.

    JS: he says that the ID might not been assigned until just
    before it becomes focussed.

    CS: I would like to see an example of a page that works that
    way.

    JS: another one of his points is that many children have a
    role and an ID for styling, but won't be active descendatns.

    CS: That doesn't mean it needs to become an active descendant.

    [15]http://www.w3.org/WAI/PF/aria-implementation/#keyboard-focu

    s_aria-activedescendant

      [15]
http://www.w3.org/WAI/PF/aria-implementation/#keyboard-focus_aria-activedescendant



    "For any element with an ID attribute, where the element is a
    descendant of an element with the aria-activedescendant
    attribute, apply the following accessibility API states to the
    target to ensure the object is accessible:"

    "Focusable, if the element also has a WAI-ARIA
    [16]http://www.w3.org/WAI/PF/aria-implementation/#def_role"

      [16] http://www.w3.org/WAI/PF/aria-implementation/#def_role


    CS: Right, but that doesn't mean it has to be a target. If the
    author doesn't make it an active descedant, it won't become
    one.

    JS: Consider an AT asking how many focusable elements are there
    in the container, there would be extra elements counted (if
    they had an ID and a role).

    CS: I think this is the only way it would work in Windows.
    ... Again, we had a long discusssion about this, and this is
    what we could agree on.
    ... Even if James is right, I would want to not reopen this
    until ARIA 1.1.

    [17]https://lists.w3.org/Archives/Member/w3c-wai-pf/2013AprJun/

    0476.html

      [17]
https://lists.w3.org/Archives/Member/w3c-wai-pf/2013AprJun/0476.html


    James Craig's first email about the topic ^

    Topic; Testing.

    MC: Like to deal with some issues prior to next Mon's call.

    <MichaelC> UAIG test suite:
    [18]https://www.w3.org/WAI/PF/testharness/testcases?testsuite_i

    d=2

      [18] https://www.w3.org/WAI/PF/testharness/testcases?testsuite_id=2


    <MichaelC> UAIG tests from a year ago:
    [19]http://www.w3.org/WAI/PF/wiki/ARIA/Tests/Testable_Statement

    s#User_Agent_Implementation_Guide_Testable_Statements

      [19]
http://www.w3.org/WAI/PF/wiki/ARIA/Tests/Testable_Statements#User_Agent_Implementation_Guide_Testable_Statements



    JS: 165 more tests.

    MC: are we testing these on the same platforms as ARIA.
    ... I assume so.
    ... Some tests are not platform specific.
    ... Some of the tables don't have IA2.
    ... I'm not sure what that means.

    JS: Can I have an example?
    ... Test case 6 might be a duplicate.

    <MichaelC>
    [20]https://www.w3.org/WAI/PF/testharness/testcases/edit?testsu

    ite_id=2&testcase_id=111

      [20]
https://www.w3.org/WAI/PF/testharness/testcases/edit?testsuite_id=2&testcase_id=111



    MC: this test didn't come from the UAIG, and it only has three
    platforms in the wiki.

    <MichaelC>
    [21]https://www.w3.org/WAI/PF/testharness/testcases/edit?testsu

    ite_id=2&testcase_id=155

      [21]
https://www.w3.org/WAI/PF/testharness/testcases/edit?testsuite_id=2&testcase_id=155



    MC: and this one: there is no column for IA2.

    JS: It's MSAA that's applicable.
    ... IA2 is not really separate from MSAA.
    ... IA2 assumes MSAA roles, states, events, etc. but then adds
    more where information is needed.
    ... IA2 is an extenstion of MSAA.

    MC: We need to inform the director about this in the right way
    so that all bases are covered.
    ... I think that answers my questions on platform.
    ... Next, review these test cases, and make sure they are still
    accurate.
    ... there are places in the testable statements where there
    were ...
    ... I made a note: "note the following need AAPI
    representations of the expected results wrt posinset and
    setsize"
    ... there are things like this peppered about the wiki testable
    statements.
    ... we need to got through and flesh these out.

    JS: there are a quite a few of those.

    MC: some will be simple; some will be more involved.

    JS: when is this due?

    MC: want to good head start on the test suite as we enter last
    call.
    ... want to begin the testing to verify the testing
    methodology.
    ... ideally, enough testing on at least one user agent so we
    are relatively sure of the test suite by the time we enter last
    call.
    ... sometimes this is error correction.
    ... we also want to get to last call failry soon.
    ... the UAIG becomes more of a bottleneck as the ARIA testing
    is completed.
    ... If we want ARIA out by Dec of this year, we can't do last
    call less than three months from that — Sep.

    JS: yes, it is a challenge.

    MC: we've had our chat; we can continue to use this time for
    UAIG test case discussion.
    ... Should do in Mon call as well.
    ... and, we need to assign people to plug these gaps, so that
    we can declare the test suite good.

Summary of Action Items

    [End of minutes]
      __________________________________________________________


     Minutes formatted by David Booth's [22]scribe.perl version
     1.138 ([23]CVS log)
     $Date: 2013-07-23 20:08:29 $
      __________________________________________________________

      [22] http://dev.w3.org/cvsweb/~checkout~/2002/scribe/scribedoc.htm

      [23] http://dev.w3.org/cvsweb/2002/scribe/



--
;;;;joseph.


'A: After all, it isn't rocket science.'
'K: Right. It's merely computer science.'
              - J. D. Klaun -