Ontologies and data sharing

So what do y'all think of the idea that we could create one big ontology that describes everything? It's a pretty stereotypically librarian thing to do, but it might be the only way to realize Tim Berners-Lee's idea of the semantic web. Ephram Miles once told my Organizing Information class that he thought it might be overly ambitious and probably not possible to create an ontology that is broad yet descriptive enough to fully realize the semantic web in the way we dream about it today. In any case, the W3C just came out with the second version of its Web Ontology Language (OWL 2), which is a standard way of representing ontologies. The press release about this from the W3C doesn't mention giant ontologies, but it does talk a lot about discipline-specific ontologies. The tag line for the press release is "OWL 2 Connects the Web of Knowledge with the Web of Data." So I became curious about how ontologies might affect our discussions of data sharing.

The W3C lists a lot of semantic web case studies and use cases. I skimmed a few of these, and it appears most people are turning to ontologies to improve search. I'm interested, though, if ontologies might be able to improve data sharing. Tim Berners-Lee (TBL) talks about data sharing in his 2006 IEEE Intelligent Systems article, "The semantic web revisited." In fact, TBL suggests that in order for the larger semantic web to work, scientific communities have to first pioneer the way, much like they did in the early days of the web. There are a lot of ontologies out there. Just do a Google search for "biology ontology." I found many sites with many ontologies listed on them. So aren't we back where we started? There are lots of isolated "data sets" (read: ontologies) that don't really relate to each other. So we have the same problem with ontologies that we do with data: how do we assign authority? how do we create appropriate metadata? how do we make these islands talk to each other? TBL suggests we need more data mining, ways to infer knowledge from patterns, and distributed information systems to "crowd source" the ontologies. But the argument still seems circular to me. It seems like he's suggesting we use computer models to then build ontologies to then build computer models to connect the data.