RDF for Model, XML for Syntax

Somehow I missed this post from James Tauber on his view of the relationship between XML and RDF back at the beginning of August. What he writes makes a lot of sense and aligns fairly closely with my own view. The question is: how best can the RDF data model mapping be expressed in the syntax schema? At its lowest level, I think the mapping would have to express the triples produced from a particular syntactic construct. Maybe a good place to start would be RSS. There is already a stable interpretation of the core constructs in RSS 1.0 the mappings from the various other versions of RSS are well understood. There might be a way to engineer a RNG schema for RSS 2.0 and map the constructs there into RSS 1.0 and hence into triples.