Brion VIBBER wrote: Something like this: http://dmoz.org/rdf/content.example.txt Except there are no "topic"s (those are "tree pages" on dmoz). Just a bunch of "ExternalPage" sections, like this: <ExternalPage about="http://www.wikipedia.org/wiki/Oktoberfest"> <d:Title>Oktoberfest - Wikipedia</d:Title> <d:Description>Encyclopedia article about 'Oktoberfest' from Wikipedia, the free encyclopedia</d:Description> </ExternalPage> It'll be up to the end users of RDF data to figure out how to get it into their search engines. They have experience with this, I think. I do, anyway. :-) --Jimbo

(re providing an RDF feed) Stephen Gilbert wrote: What kind of filtering would you suggest? Assuming our intention is to provide an index of encyclopedia articles, it might be reasonable to, for instance, filter out certain kinds of redirects: * to an article that's identical except for case; since searches are usually case-insensitive they're redundant (though harmless) * to a page in a non-article namespace (talk, user, wikipedia, image) * to a non-existent page -- brion vibber (brion @ pobox.com)

Stephen Gilbert wrote: That sounds sensible. In the past when I've done ad hoc data importing from UseMod into Bomis, I filtered out stubs and redirects. It's not really possible to know all the things that a potential user of bulk link data might want, and not advisable to make this too complicated, but a simple "raw" and "filtered" version would be good. "raw" contains everything, "filtered" takes out the stubs (by some conventional definition that we use elsewhere) and the redirects. --Jimbo