named graphs (was Re: Data URIs)

> I'm finding them more and more appealing. So, supposed I have an axiom:
>     <owl:ObjectProperty rdf:about="#has_direct_part">
>          <rdfs:subPropertyOf rdf:resource="#has_part"/>
>      </owl:ObjectProperty>
> And I want to annotate it. I'm proud of it. Here's a data uri:
> <data:application/rdf+xml,%3C%3Fxml%20version%3D%221.0%22%3F%3E%0A%3C% 
> 21DOCTYPE%20rdf%3ARDF%20%5B%0A%20%20%20%20%3C%21ENTITY%20owl%20%22http 
> %3A// 
> 20owl2%20%22http%3A// 
> 20%3C%21ENTITY%20xsd%20%22http%3A// 
> 3E%0A%20%20%20%20%3C%21ENTITY%20owl2xml%20%22http%3A// 
> 2006/12/owl2-xml%23%22%20%3E%0A%20%20%20%20%3C%21ENTITY%20rdfs%20% 
> 22http%3A// 
> %21ENTITY%20rdf%20%22http%3A// 
> 22%20%3E%0A%20%20%20%20%3C%21ENTITY%20worm-wikipedia%20%22http%3A// 
> %23%22%20%3E%0A%5D%3E%0A%3Crdf%3ARDF%20xmlns%3D%22http%3A// 
> %23%22%0A%20%20%20%20%20xml%3Abase%3D%22http%3A// 
> 7Ebparsia/ontologies/2008/anatomy/worm-wikipedia.owl%22%0A%20%20%20% 
> 20%20xmlns%3Aowl2xml%3D%22http%3A// 
> 0A%20%20%20%20%20xmlns%3Aworm-wikipedia%3D%22http%3A// 
> %23%22%0A%20%20%20%20%20xmlns%3Axsd%3D%22http%3A// 
> XMLSchema%23%22%0A%20%20%20%20%20xmlns%3Aowl2%3D%22http%3A// 
> 3A// 
> 3D%22http%3A// 
> 20%20xmlns%3Aowl%3D%22http%3A// 
> 20%20%20%3Cowl%3AOntology%20rdf%3Aabout%3D%22%22%3E%0A%20%20%20%20%3C/ 
> owl%3AOntology%3E%0A%20%20%0A%20%20%20%20%3Cowl%3AObjectProperty%20rdf 
> %3Aabout%3D%22%23has_direct_part%22%3E%0A%20%20%20%20%20%20%20%20% 
> 3Crdfs%3AsubPropertyOf%20rdf%3Aresource%3D%22%23has_part%22/%3E%0A%20% 
> 20%20%20%3C/owl%3AObjectProperty%3E%0A%3C/rdf%3ARDF%3E>
> Ok, I hear the laughter and the shrieks of outrage :)
> This wasn't the nicest serialization*, but it works. It's reversible.  
> And now we have a URI for an RDF subgraph. data uris are implemented  
> in major browsers and have an ITEF RFC. They clearly are not for hand  
> authoring or reading, but neither is reification. If we chose  
> ntriples or turtle as the base, it would be somewhat terser. If we  
> have a normalization, we could be pretty sure of getting the same URI  
> for the same syntactic expression (trickiness there!). You can use  
> that anywhere you'd use an URI, so the subject or object of a triple.  
> It would nest (double encoding doesn't cause a blowup.
> Using literals is a bit easier to write by hand.

I don't have a strong opinion about these different approaches, but I
think we should be clear that these are primarily interim designs for
use until there's a standard format for RDF with named graphs.  And we
should make sure whatever approach we pick can be compatible with a
future named-graphs standard, so OWL 2 wont have to change: there will
be the RDF/XML serialization (using one of these interim hacks) and the
RDF/XML+NamedGraphs serialization that doesn't need such hacks.

As far as I can tell, there is consensus in the RDF community that there
should be a named-graph standard.  Making such a standard is,
unfortunately, an awkwardly-sized peice of work.  It doesn't really fit
inside any other work, and it's rather small for a whole new Working
group.  So I don't know exactly when/how it will happen.

   -- Sandro

Received on Wednesday, 2 July 2008 19:11:50 UTC