[re-send.  Originally mis-posted as
http://lists.w3.org/Archives/Public/www-html/2008Jun/0003.html
  -Al ]


Re: http://www.w3.org/TR/2008/WD-xhtml-role-20080407/

This is a personal comment.  It is a concern about a possible problem,
not the assertion of a sure problem.

WAI-ARIA has a dependency on the role attribute module, or at least
has a dependency on the host language affording a suitable @role attribute
for which our roles are attribute values.

The Role Attribute module has been relaxed to allow a variety of ways
that the @role attribute is introduced into a host language or document.

But it appears to require that in any host language the implementation of this
attribute takes as its value

<quote
cite="http://www.w3.org/TR/2008/WD-xhtml-role-20080407/ #s_role_module_attributes">
 one or more whitespace separated CURIEs
</quote>
[CURIE] http://www.w3.org/TR/2008/WD-xhtml-role-20080407/#ref_CURIE

Curies are new and progressing down the Rec track behind the Role Attribute Module. This would seem to be in reverse of the dependency-induced order
as the Role Attribute has a dependency in this way on Curies.

In addition, because of the way that WAI-ARIA roles adjust element semantics, role values carry a connotation of language extension; and Curies expand to URIs which could indicate totally uncoordinated, crowd-sourced distributed
language extension.

The latter could be regarded as undesirable for the maintenance of HTML, the
central format of the One Web.

I am unclear how much saying that @role is a list of Curies means that the processor must process in accordance with the URIs that the Curies abbreviate.

It would seem to me that WAI-ARIA needs host languages to respect the @role values set out in the http://www.w3.org/1999/xhtml/vocab# vocabulary when these values appear un-prefixed as tokens (items in the space-separated list) in the @role value. What WAI-ARIA may need in terms of support for role extension is not known at this time. We have a working decision that role extension will not be addressed in
WAI-ARIA 1.0.

In that sense, ARIA has a dependency on @role but does not need @role to have the dependency on Curies. And there is a chance that cross-host- language support
for @role could be impeded by the dependency on curies.

That's my nervousness. We will be talking about this more as we work to tighten up the WAI-ARIA specification as it deals with host language insertion. But I was told to put it on record as a Last Call comment for the Role Attribute Module because it touches that specification which is already s/in/past it's second
Last Call.

Thanks,

Al
/me

Reply via email to