I just came across the Wikidata project and was excited to see that we
share pretty much the same goals.
After seeing the Wikimania video and reading the technical proposal, I
think choosing any other data model than RDF would be a mistake.
DBPedia has done a lot in this area (as I'm sure you must known) and
failing to build on this work would be a bad start for Wikidata. There
is no need to invent new models and APIs, RDF and SPARQL are standard
and proven tools.
Also, implementation-wise I would suggest scrapping any wiki-based
software (Semantic)MediaWiki and build a fresh generic RDF-based
platform.
Hereby I want to present to you Graphity which is an open-source
generic Linked Data management platform, found at http://graphity.org.
We just released v1.0.0 for PHP which supports JAX-RS annotations for
RESTful APIs and lightweight RDF object API very similar to Jena's.
I hope you can find some of our code useful or even contribute to it.
If you find it interesting, we can provide more documentation and/or
support.
Java version is on the way, as well as the most important component -
generic Linked Data browser. It will implement most of the technical
requirements specified in the Wikidata proposal. A prototype for O3.6
extension can be seen here: http://semanticreports.com