Re: New Issue: Allowing extension of RDFa Core

Hi Shane,

Sounds like a good idea.

One minor thing though, is that @role is like @typeof, not @rel/@rev.
That doesn't change your suggestion, just the examples and use-cases.

Just in passing, the reason that we've avoided bringing @role into
RDFa processing before is that it applies to the *document* not the
metadata. So whilst RDFa might tell you that 'this person has a phone
number of x', @role tells you that 'this document has a footer of y'.

In other words, you need different rules for getting the subject.

At some point it would be really good to tackle this and have rules
for @role. (But that's separate to your proposal.)

Regards,

Mark

On Fri, Apr 2, 2010 at 12:41 AM, Shane McCarron <shane@aptest.com> wrote:
> (with my PFWG hat on)
>
> Part of the work in the PFWG is the creation of the Role Attribute
> Specification.  This is a simple specification that defines an attribute
> that can be used to clearly identify the (machine interpretable) role of an
> element (e.g., role='banner', role='spinbutton').  A number of roles are
> defined in the XHTML vocabulary, but the role attribute's datatype is ( TERM
> | URIorCURIE )+, and the intention is that this extensible attribute can be
> used with any vocabulary.
>
> Obviously, it would make some sense for @role to be able to generate triples
> that could be used to help find specific roles in resources.  I doubt that
> spinbutton is particularly interesting, but 'definition', 'contentinfo', and
> 'main' are probably low hanging fruit for a semantic web inference engine (
> <http://example.com/somedocument.html> <xhv:main>
> <http://example.com/somedocument.html#fragmentID> ).
>
> However, we have no real way at this point to allow the addition of new
> attributes to our processing model.  There are no 'hooks' in the Sequence
> [1].  If there were, a specification like the Role Attribute could say 'When
> this attribute is used in an RDFa Host Language, its values are interpreted
> as predicates in processing step N.' or whatever.
>
> So, that's the question.  Does it make sense to try to introduce this type
> of hook?  If we had one, would we also need a way of indicating (perhaps in
> a Host Language RDFa Profile document) what attributes hooked in where?
>  Sort of an 'instruction to the RDFa Processor'?
> Thoughts?
>
>
> [1]
> http://www.w3.org/2010/02/rdfa/drafts/2010/ED-rdfa-core-20100401/#sequence
>
>

Received on Friday, 2 April 2010 09:12:50 UTC