Re: Advancing translational research with the Semantic Web

Pat Hayes wrote:
> Our paper suggests that the URI of the RDF/XML document be used as the 
> name in this case. This works as long as you are consistent about it. 

...and as long as the document is always opened from its official location 
(versa from a local copy), and you keep each item is a separate file (which 
is not practical when you need to exchange a lot of data)?


> Or, you could introduce 'naming RDF' documents whose sole purpose is to 
> have a stable URI as the graph name and which then reference (eg by 
> importing) the documents making up the intended graph.
> 
>> Named graphs ARE supported by most triplestores, but they are mostly 
>> already reserved for other uses, like the representation of provenance 
>> based on the RDF files that the triples were loaded from.
> 
> Reserved? In what sense? a single URi can have multiple properties and 
> hence multiple uses.

Named graphs are great, but I (and others?) use them to keep track of sets 
of statements that are in general retrieved and edited together (and would 
be stored in the same document, but databases don't have that concept).

If I had to use the same mechanism for associating attribution information 
to arbitrary sets of statements, that would complicate things considerably, 
as far as I can see, and wouldn't I need to duplicate a lot of statements?

Received on Wednesday, 16 May 2007 21:04:57 UTC