I was aware of the data dumps that freebase offers. However, I do not think that is the right model for the web moving forward. I believe that the web will organize as a collection of federated systems, for the simple reason that data sets size will make them beholden to some location in the cloud. As such, I am looking for a typical federated style of consuming/collaborating mashup data mining using the Freebase data.
In the other thread with Steffen, the idea of semantic web programming, which in one form is the consumption of RDF data streams, was brought up, and I have been diving in that direction. In that context, the important of ontologies appears to jump out.
If you're looking for the data as RDF you can find the complete Freebase data set published as linked open data at http://rdf.freebase.com
As far as data federation goes, each Freebase type publishes its own RSS feed of updates. For example, updates to Protein. I imagine that it should be possible to synchronize changes between federated databases using these feeds although I don't know of anyone who is doing that right now.