- From: Giovanni Tummarello <giovanni@wup.it>
- Date: Tue, 16 Nov 2004 10:00:27 +0100
- To: Stefano Mazzocchi <stefano@apache.org>, www-rdf-interest@w3.org
> > I think it would be a mistake to specify what context is by harcoding > one more metadata field to every statement. What next? what about > licensing restrictions on the time about the change of the context of > the statement? > > the queries will get hairy, agreed, but context is something you know > when you query, what you know where you enter data is just metadata > about statements, and that's exactly what reification is all about. > Yup.. its blank node for existing statements (note, not a single one .. since there can be more, independant, reifications for the same triple) and it's all that's needed. Of course RDF serialization issue of these structures are worth noting (the triple bloating problem) but internally i expect RDF DBs to handle this more efficently so i dont care much about the bloat at this level. AT xml serialization leve, those interested in a more triple efficent (i'd say triple optimal) way to reify, we'll post a short proposal about this probably later today. :-) we're about to start using that since it does in fact save a lot of bandwidth. .. it is just a matter of a scaling factor, of course, but is rather relevant in ACTUAL agent/p2p stuff where a large amount of context information is expected to be shipped around. Question: named graphs as i understand serve as an "efficent" way to handle reification actually. you prived a name for a model (a set of triple) and that is equivalent to a URI in a reified node. (except that you're forced to put that model , the named one, into some other URI like an external file or something) So.. if an acceptably efficent way to serialize reification is available (And its just syntactic sugar, nothing that goes outside the RDF specs as the other do), wouldnt these (and the "wave of incoming quad implementations") loose a bit of their meaning?
Received on Tuesday, 16 November 2004 09:04:47 UTC