- From: Danny Ayers <danny666@virgilio.it>
- Date: Thu, 8 Aug 2002 00:28:05 +0200
- To: "Sampo Syreeni" <decoy@iki.fi>
- Cc: "RDF-interest" <www-rdf-interest@w3.org>
>Actually the assumption is that people's wants are pretty much irrelevant, >here. This is a question in software engineering, not in politics. Software engineering doesn't exist in a void - design decisions such as this have to make sense in the real world. There is a great deal of difference between identifying a few people, a lot of people or all people. The id numbers mentioned so far have national boundaries - will they still work globally? The vast majority of the world's population don't have access to the Internet, let alone email addresses - do we need to id people who will *never* be online? etc. There >are valid applications where one might want to use a unique identifier for >people when they exist, so it's reasonable to expect SW infra to support >those applications. Could you please give me an example that couldn't be equally well served using bNodes? After all, that is in effect what you're doing with the id - the person isn't the id number. I can accept that it may be easier to deal with URIs directly, rather than having to splay back through bNodes, but is easier to deal with a mix of URI-identified and bNode-identified people or bNodes alone? >>In practice, I would suggest that Seth's bNode approach will be >>perfectly adequate - e.g. "give me a list of the books written by the >>bloke who came up with html". Not precise enough? - add another rule, >>which may be his/her surname, email address, national insurance number >>or the title/Amazon URL of another work. > >However, the same goes for everything else, from web pages to abstract >works to whatever else you might name with a URI. By this token, there is >absolutely no need for URI labelled nodes in RDF graphs, since everything >can be taken care of by a suitable arrangement of anonymous nodes and >daml:unAmbiguousProperties. I don't believe that's so - a URI represents a fixed point, without fixed points there's nothing to reason about. >>Locally, we might use a 99-digit id number, but in the wild the >>id-by-association should be plenty (and better matches the conceptual >>frameworks in which it is likely to be used). > >However, *if* NID's are already available, I think it should be possible >to use them as such, just like we use ISBN's for books. For those of us >who don't have/want a NID, an anonymous node is pretty much ideal. This sounds like a good practical approach (assuming politics doesn't get in the way). Cheers, Danny.
Received on Wednesday, 7 August 2002 18:37:04 UTC