RDFa specs and discussions usually tell us that you're not limited to using it with HTML, and then they only talk about using it with HTML. I wanted to see how difficult it would be to incorporate the examples from the W3C's RDFa Primer into DocBook and DITA documents, and it wasn't difficult at all. DevX has just published an article that I wrote on how I went about it: Using RDFa with DITA and DocBook.
I used to think that DTDs such as DocBook and DITA didn't need RDFa because they're so customizable, but when you define new metadata elements or attributes for either, you have to write new code in XSLT (or whatever your language is for processing the XML) to go get that new metadata. Once you've added RDFa support modules to these DTDs, though, existing RDFa extractors (that conform to the spec) will pull out any new kinds of metadata that you store in these RDFa attributes, which also reduces your need for future customization of those DTDs.
I hope that people in the RDFa, DocBook, and DITA communities find the article useful.