Some thoughts (on questions) on RDFa in HTML5

Recently here has been some discussion regarding markup of
domain-specific data and how to use custom vocabularies. Here are some
loose thoughts.

Right now it is fairly straightforward to implement RDFa in an
XML-based markup language [1]. RDFa is part of XHTML2 and there is a
DTD for a modified XHTML 1.1 that includes RDFa (which works in the
W3C validator right now).

To use RDFa you

1. identify one or more custom vocabularies (using the "xmlns"
construct on the html element),
2. mark up data with your custom vocabulary (using various attributes
on an element).

How would this look in HTML5?

1. HTML5 is not an XML-based markup language. So "xmlns" does not
sound right? Should it be called something else? "ns"?

2. How can conformance be tested? In XHTML you apply XSLT on a test
page and the result is compared to the intended result. This is easy
because the parsing rules are defined. Is there a canonical parsing
model for HTML that makes it possible to test conformance in a similar
way?

[1]: http://www.w3.org/TR/xhtml-rdfa-primer/

Regards,

Peter