- From: Sampo Syreeni <decoy@iki.fi>
- Date: Fri, 15 Jun 2001 12:40:22 +0300 (EEST)
- To: <www-rdf-interest@w3.org>
If I remember correctly, there has been some prior discussion here on how to map resources/URIs to their related metadata. Yesterday I ran into a very interesting piece of foresight on behalf of the IETF naming/addressing people, namely URCs (Uniform Resource Characteristics), an idea which I'm not sure has been raised on-list before. RFC2169 proposes a simple, easily implemented HTTP interface to resolve URIs against a database located on an HTTP server, with a complete set of resolution services that can be implemented. The obvious one is URN-to-URL, of course, but there are also ones called N2C and L2C, meant to map from URIs to resource characteristics. URCs are essentially synonymous with metadata. The interface is simple: to resolve <URI>, one does an HTTP GET to /uri-res/<service>?<URI>, where the interesting <service>s for our purposes would be N2C (name to characteristic) and L2C (locator to characteristic). The returned data format is not defined, but is rather communicated via the standard HTTP Content-Type header. All the usual HTTP server responses can be used. In our case, we would probably return redirections to get to dedicated metadata servers (like rdfdb; probably a future thing), straight XML encoded RDF marked with the suitable content type (which begs the question, is application/rdf+xml kosher?), or some newly registered content type for a stream of triples (my personal favorite - you would use the query URI to do a selection on the subject field of your table of triples, and just pump the data back as-is). The RFC even describes a simple MIME content type for lists of URIs, which could be used to list references to multiple RDF files and future dedicated triplestores where metadata for the given document resides. Even if the RFC *is* experimental, one could likely build some nice implementations on it (e.g. extending the idea to HTTP PUT would enable Annotea-like apps to be built without having to configure the triplestore). One might even consider throwing the thing into the standards track, if it proved useful enough. What do you think? Sampo Syreeni, aka decoy, mailto:decoy@iki.fi, gsm: +358-50-5756111 student/math+cs/helsinki university, http://www.iki.fi/~decoy/front
Received on Friday, 15 June 2001 05:40:24 UTC