See also: IRC log
https://www.w3.org/WAI/PF/Group/track/issues/466
CS: example where role doesn't
map; i.e. role="main"
... example of table element with role="main"
... IE puts "main" in AriaRoles property and leaves table role
in AAPI
... one of the other browsers does not expose the role so it
acts like role="presentation"
JC: Safari does not expose the table semantics
DB: possible that FF is similar
to IE - with landmarks, treat them like bookmarks
... there aren't good mappings - probably expose as table role
with object attribute for the landmark
JC: the only way to get
consistent behavior then is tell authors not to put landmark
roles on elements with other semantics
... James Nurthen thought he had a bug - redefined the role but
browser didn't honor it
CS: do we need to specify this in
the UAIG?
... don't know what the authoring guidance currently says
... think of landmark roles as being different from widgets or
other structure
... you can be both "main" and a "table" - should be
allowed
... as we start talking about patterns or other mechanisms,
allows a forward path
JC: think clear thing to do is direct authors not to use landmarks on semantic roles
CS: seems advisory - would like to leave a path forward for other mechanisms
JC: spec says there can be only one role at a time
CS: the core question is - when
we say "not mapped", do we mean "ignore" or treat it as
presentation?
... in HTML, usually unknown things are ignored
... what happens to the children of the table [<tr>s and
<td>s]?
JC: James Nurthen raised this
issue - he was trying to retrofit old software and could not
add <div> elements around the table
... JN wanted the table semantics to disappear but it didn't
work when he added the role of landmark
CS: can you use role="main presentation"?
JC: no, that doesn't work
... it was a layout table in JN's example
CS: but there could be a data table that is the main part of the page, why do you have to wrap that in an extra <div>?
JC: seems like you would want to wrap that in a <div>
DB: for new development, isn't there something in HTML5 you can use?
JC: could use the <main> element but on a retrofit, he wasn't able to do that
CS: 1. what happens when you retrofit and how bad is that problem? 2. what is best path forward?
<jcraig_> So if we decide this is correct, we should tell the authors to never use landmarks on elements with native semantics <table role="main"> <form role="complementary">
DB: workaround - practice has cropped up - datatable="0" - Windows thing
<Zakim> jcraig_, you wanted to say by your logic <main role="table"> should also be both main and table
JC: if do <table role="main"> then should also do <main role="table"> which goes against HTML5
DB: think landmarks should have
been an attribute (aria-landmark)
... to have the landmark override seems like regressing
accessibility, FF chose to expose both and let AT decide
<davidb> JC: given that isn't an option now, think safer to have it explicitly override the role.
<davidb> CS: where there is no mapping we would (lose semantics?)
<davidb> JC: the author has chosen to explicitly override
<jcraig_> JC: if even JamesN didn't understand this, how would you expect a normal author to understand that difference.
<jcraig_> CS: I expect normal authors to have a different understanding than the experts.
DB: IE & FF are leaving this
up to the AT by exposing all of the information
... that is a problem, the spec suggests that we should only be
exposing one role
... want consistency in how things are overridden - if one AT
uses the ARIA role and another uses the semantic role, it's
inconsistent
CS: simplest path forward is to tell users not to do that
JC: what exactly are we telling authors? don't use landmark roles on elements that have native semantics
CS: doesn't HTML5 say something about landmark roles not overriding strong semantics?
<davidb> http://www.w3.org/TR/html5/content-models.html#wai-aria
SN: reading the spec, there's nothing about landmark roles not overriding native semantics and nothing about tables having strong semantics that can't be overridden
CS: one option is to define in
the UAIG how this must be done (either IE/FF model or Safari
model)
... other option is to tell authors not to do this
JC: suggest adding to HTML5 spec and the Authoring Practices Guide
CS: people are making different
assumptions about what "not mapped" means
... could leave it like that or put a "may" or "should" after
the table: "things that are not mapped may override"
JC: could define what "not mapped" means - think it means something different in the OS X column than in the other columns
CS: does it mean ignore or override?
JC: would be okay with a MAY
requirement for not mapped
... MAY leaves the door open for the way IE and FF do it
CS: and we tell authors not to do
it - address it in ARIA 2.0
... UAIG will allow either behavior - need a MAY
statement
... UAIG - UAs MAY use the generic role or use the existing
role
... need authoring guidance telling them not to
<davidb> https://www.w3.org/WAI/PF/Group/track/actions/925
SN: in IE, throw system alert if you change the display status of the element that has role="alert" or "alertdialog" but not if change the display status of the parent
JC: one of hundreds of examples
http://www.w3.org/WAI/PF/aria-implementation/#mapping_role_table
AS: UAIG says the UA should fire a system alert event but doesn't say when
JC: could fire it whenever focus
is inside the dialog
... for alertdialog
... a little more difficult for live regions
DB: performance concern on the side of AT?
JC: both AT and rendering agent
DB: about mutation happening in aria-live="assertive" container
JC: or changes on parent of that container
DB: not sure of current FF behavior - would have to research what FF does and why
JC: if there's an easy set of
hueristics we can use, that's fine - otherwise, we're going to
have inconsistencies - maybe they just get reported as
bugs
... ex: for alertdialog, when focus moves into the dialog, fire
the event
... not sure there are easy heuristics for live regions
DB: need to get all the
information about how IE and FF do alerts in these
scenarios
... reads from ARIA spec about what happens with
aria-live="assertive"
... says user should be notified immediately - should try to
accommodate but can put most of the onus on web developer
JC: fire a system alert event
"when shown" is nebulous because of the multitude of ways you
can do that in a web browser
... if keep it nebulous allows for inconsistencies, may be good
for competitiveness but not necessarily good for the
accessibility community
... authoring practice approach might be the best approach for
now until we fix all the bugs or find the magic heuristics that
will solve all the problems
<davidb> https://www.w3.org/Bugs/Public/show_bug.cgi?id=14503
Andi to add some of this discussion to bug 14503. David will add anything he digs up.
Andi to propose that this action be deferred to post-Last Call
This is scribe.perl Revision: 1.136 of Date: 2011/05/12 12:01:43 Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/ Guessing input format: RRSAgent_Text_Format (score 1.00) Succeeded: s/no mapped/no mapping/ Succeeded: s/topic/topic:?/ Succeeded: s/topic/topic:/ Succeeded: s/onus/most of the onus/ Succeeded: s/topic issue-466/topic: issue-466/ Succeeded: s/leave table role in AAPI/leaves table role in AAPI/ Succeeded: s/Joseph thought he had a bug/James Nurthen thought he had a bug/ Succeeded: s/do we need to specify this in the UAIG/do we need to specify this in the UAIG?/ Succeeded: s/[<tr>s and <td>s]/[<tr>s and <td>s]?/ Succeeded: s/the author has chose to explicitly override/the author has chosen to explicitly override/ Succeeded: s/reading the spec, there's nothing about landmark roles/SN: reading the spec, there's nothing about landmark roles/ Succeeded: s/after the table "things that are not mapped may override"/after the table: "things that are not mapped may override"/ Succeeded: s/:? ACTION-925/: ACTION-925/ Succeeded: s/: ACTION-925/ACTION-925/ Succeeded: s/UAIG says the UA should fire a system alert event but doesn't say when/AS: UAIG says the UA should fire a system alert event but doesn't say when/ Succeeded: s/issue-466/ISSUE-466/ Succeeded: s/JC if keep it nebulous allows for inconsistencies/JC: if keep it nebulous allows for inconsistencies/ No ScribeNick specified. Guessing ScribeNick: Andi Inferring Scribes: Andi Default Present: David_Bolter, Andi_Snow_Weaver, James_Craig, Cynthia_Shelly, Sharon_Newman Present: David_Bolter Andi_Snow_Weaver James_Craig Cynthia_Shelly Sharon_Newman Got date from IRC log name: 06 Dec 2011 Guessing minutes URL: http://www.w3.org/2011/12/06-aapi-minutes.html People with action items: WARNING: Input appears to use implicit continuation lines. You may need the "-implicitContinuations" option.[End of scribe.perl diagnostic output]