Re: Exchange of Named RDF Graphs

Eric Jain wrote:

> Giovanni Tummarello wrote:
>
>> considering statements inside "files which have a name"  seems a 
>> naive way to make nothing else than a "bag of reifications" that is a 
>> way to assign a URI (the name of the graph) to a bunch of statements. 
>> Given that bag of statements can already be done with the existing 
>> specs, this is already a reason to drop the idea.. but there is more.
>
>
> You are right insofar as that in theory everything can be expressed 
> with nothing but triples. But in practice many people, including 
> myself, have run into problems with this approach, especially when 
> dealing with a lot of data.

I've learned about the superman/Venus problem at the swig-meeting in 
Cannes which seems to show that there are theoretical reasons why 
reifications is unsuitable for quoting.

Consider the following base-model:

ex: john ex: claims _1
_1  rdf:subject   ex:morning_star
_1  rdf:predicate ex:not_same_as
_1  rdf:object  ex:venus
ex:venus owl:sameAs ex:morning_star

the completely legal inference
_1 rdf:subject ex:venus

would let us affirm that john claims that the venus is not the same as 
the venus which is clearly not his claim.

Now, I'm not a fan of the idea of named graph, and I think that if it 
can be done by existing means it should be done so. I don't think the 
"practical reasons" for named graphs should lead us to make RDF more 
complicated but rather be solved with e.g. shortcuts in the 
serialization (which look like named graph but in fact resolve to a bag 
of reifications). However I guess there should be a mean to express in 
RDF that different sources assign different meanings to the same URIs so 
that the venus-problem can be solved, named graphs seems not be a 
solution for this, but I'm not sure if its the best one.

reto

Received on Wednesday, 5 January 2005 14:22:07 UTC