I looked into RDF pretty heavily a few years ago. The biggest takeaway is that it was naive in its level of trust.

If I make a new RDF document in a repo, claim it was written by Shakespeare, and then somebody does the search on "Complete works of Shakespeare", I'd show up. And there was no real way to police that.

And unfortunately, we get into web of trust type issues, which in itself is ugly. I really do wish that RDF was usable for search engines, and google, and that they could trust sites. But it's just not feasible, and I don't see any scalable solution to it.