- From: Dirk-Willem van Gulik <dirkx@webweaving.org>
- Date: Thu, 26 Aug 2004 00:20:09 -0700 (PDT)
- To: Alberto Reggiori <alberto@asemantics.com>
- cc: public-rdf-dawg@w3.org
Just some (personal) background to Alberto his use case. Within various fields we are seeing several ways of using RDF: -> A single 'file' with facts authored by a single entity. OR -> A single file with facts taken from different authorties, or invented locally. -> A URL pointing to that 'file'. -> Depending on the operational requirement; the system using the data may either fetch the file and internalize it OR just keep refering to that URL. Now the transience down the line is the problem; some facts are really authored by the relevant authority - and should be considered simply as a local copy - whereas other triples are created by more than just some smushing; and are originating at some point in the federated processing path. Alberto's example for foaf files form different people beeing merged and augmented with, say, information from IMDB, is a good example. But regardless at some point is it needed to be able to modify, delete or identify the facts from a given 'authority' (in the loosest 'source' sense int he world) on a per 'entity' basis. Even though you've dumped them into one database all together. Pragmatic solutions sofar have been keeping provenance and context at all time when we're working with a copy of the facts -or- using just a URN (or some other URI such as a URL) rather than a copy of the data and rely on fast fetching/caching from the 'origin' when it is really needed. Dw.
Received on Thursday, 26 August 2004 07:31:50 UTC