- From: Dan Brickley <danbri@w3.org>
- Date: Mon, 22 Apr 2002 18:42:02 -0400 (EDT)
- To: Joshua Allen <joshuaa@microsoft.com>
- cc: "R.V.Guha" <guha@guha.com>, Margaret Green <mgreen@nextance.com>, <www-rdf-interest@w3.org>
On Mon, 22 Apr 2002, Joshua Allen wrote: > > I think you are right --- we have a situation where a bunch of > different > > scientists are each publishing data, meant to be used by others. XML > > schema forces them into a model where they all have to agree to a > schema > > and use that to publish their data. If one of them has a new kind of > data > > that hasn't been thought off before, then it could be tough luck. > > Well, if you want your RDF to interoperate, it needs to have a > consistent schema or a mapping between your RDF schema and the other > guy's. This is exactly the same situation with XML Schema. Not quite. We get benefit from RDF when we merge data from multiple sources if those sources are making different claims about the same things. We don't need to map classes and properties to each other, only figure out what things we're talking about. As noted here frequently, this only benefits us in a world where we aren't _too_ confused about the meaning of URIs. I made a quick writeup, complete with testcase, of this last year for rdfweb. It degrades into waffle (thinking out loud, badly) towards the end, but the princple should be clear. Since last updating that page there's a 3rd implementation that succesfully answers the test query (SQL stuff in Ruby, pretty trivial to implement). Any full DAML+OIL system should also be quite capable of answering the sample query. Anyway, http://rdfweb.org/2001/01/design/smush.html Hopefully it makes the point that schema mapping, in general, isn't necessary to get some benefits from RDF. Dan -- mailto:danbri@w3.org http://www.w3.org/People/DanBri/
Received on Monday, 22 April 2002 18:42:21 UTC