 |
 |
On 4/15/15 5:45 AM, Corey A Harper wrote: > Hi Holger, > > I'm assuming by Dataset, you're talking about "a distinguished RDF > graph and zero or more named graphs." [1] > > If I'm correct, I think the concern that both I and Tom raised > remains. If I'm consuming data from an external source, and applying > my own validation constraints, it seems overly restricting to say that > these must be in the same Dataset. Or are you saying that, in my > scenario, I would recast the consumed data, give it a new named graph > ID, and consider it part of the same dataset for te purpose of my > validation application? In SPARQL implementations such as Jena, a dataset is primarily an interface that maps names to graphs. A Graph is another interface that can have any number of implementations. Many Graph implementations are in fact wrappers of remote graphs. So there would not be any need to copy data from a remote source into a local copy (although this may speed things up). And yes, datasets may also include new graph names for external graphs or virtual graphs such as MultiUnion graphs (which we use a lot in practice, e.g. to validate an imports closure graph). And finally, there is the SPARQL SERVICE keyword, for specific SPARQL end points. Holger