The question of deployment of Dublin Core came up, I didn't know of any good resources. Stephen Downes helped out with a pointer to OAIster (they have an oyster in their logo, I guess that's the pronunciation): 5,914,431 records from 536 institutions.
The search form provides full-text, title, creator and subject fields, along with resource (media) type. There's a choice of results sort order. They've got a simple HTTP GET interface (using OpenURL). The results come back in a HTML table.
Has anyone done a SPARQL to OpenURL rewriter? Not sure what you'd do about the results, right now it looks like it might need scraping. Hmm, it was on the microformats list that this thing came up, I wonder if a uniform results set representation made with XHTML in the microformats style using the OpenURL fields could be a good way. That could give the same kind of results as they are currently doing, but would only be an XSLT away from SPARQL results sets.
PS. Tantek is still skeptical about the deployment of DC on the Web. He has a point, the OAIster setup only appears to have human-oriented interfaces, there's nothing explicitly DC - "creator" isn't necessarily "dc:creator".[Danny]