I think the really hard problems are almost unsolvable so waiting for them to be fixed before putting effort into this area doesn't seem wise.

In general we need a common mechanism to be able to make simple factual assertions in a web page or when describing a data flow.

The other very significant problem is that learning the technologies around rdf/owl/sparql is a lot of work and to be honest about it, most developers refuse to even learn sql properly.

I'm hopeful google's slow but steady focus on web page/email markup(for google now) will get the momentum behind solving some of the simple problems that it's insane we still have (things like, is there an address that describes this businesses location somewhere on this web page?).