[RDFa] ISSUE-55: CURIEs or QNames in XHTML1.1+RDFa

Last week, we discussed the issue of CURIEs or QNames in XHTML1.1+RDFa.
Mark made a number of good points, in particular that the use of QNames
is actually *wrong*, as evidenced by the fact that SPARQL has chosen to
not use QNames, but instead has defined a grammar that looks more or
less exactly like CURIEs.

Mark proposed that we use the same approach: build into our
specification a grammar for what we expect these values to look like,
without having to reference an outside specification.

After a few days of thinking about this, I agree with this solution. We
are basically doing exactly what SPARQL did. Neither they nor we are
trying to define a reusable concept. It's too bad that we can't get
together and actually *define* this concept, but we need to worry about
RDFa right now, and that's the way to solve it.

Any thoughts, complaints, objections, please send.

-Ben