The ORM Foundation

In this simplified metamodel for ORM, UUID are used to identify elements in the model.  This contrasts with RDF which uses Universal Resources Identifiers (URI).  This observation led me to the following thought: what if ORM used URI too ? What if type identifiers were prefixed with a web address, implicitly or explicitly ?

It would enable the "interoperability" of ORM models, one ORM model being able to refer to another one.  It would also make ORM more RDF-like, with similar applications.  And possibly make ORM the next generation of RDF ?

 Any thoughts ?

PC

I'm curious on how you derived your assertions:

pcarbonn:

It would enable the "interoperability" of ORM models

 
I don't see how using URI's would help to achieve "interoperability"
Please can you explain your idea here - maybe with an example?

 

pcarbonn:

It would also make ORM more RDF-like

 
Why do you feel a need to make ORM more like RDF?
What do you see as the benefits?

Ken

In principle, we could use IRIs (Internationalized Resource Identifiers, the successor to URIs) to provide global identifiers for model elements, but I currently don't see that as practical, universal substitute for UUIDs. One problem with IRIs is that the same element may be assigned different IRIs by different users, weakening the notion of a preferred identifier, and then you need to explicitly match the IRIs (e.g. using owl:sameAs or owl:equivalentProperty). 

One option I have considered is using IRIs to indicate when predicate readings on different fact types are intended to have the same semantics (and hence conceptually denote the same predicate). For example, the "runs" predicate in "Person runs Marathon" is semantically the same as the predicate in "Horse runs Race", but different from the runs predicate in "Person runs Company". Currently ORM treats these as 3 different predicates. Using IRIs we could treat the first two "runs" as the same predicate (though the fact types still differ). If enough people feel that it would be useful to add such semantics, then we will consider adding this capability into NORMA. Other uses of IRIs could also be considered.

We do intend to extend NORMA with an automated mapping from ORM to OWL. There are some aspects that OWL doesn't cater for yet (e.g. acyclicity constraints) but most of ORM will map fairly easily to OWL (and hence RDF).

Cheers

Terry

Thank you Terry, I understand your points.

Ken, I see that ORM has several advantages over RDF and OWL, and yet, RDF and OWL seem to be more popular.  I thought that some ideas may be copied, and URI could be one of them. But I understand Terry's point.

PC

Terry Halpin:

One problem with IRIs is that the same element may be assigned different IRIs by different users, weakening the notion of a preferred identifier

That is not an issue if IRIs but of usage of IRIs. It is an explicit feature: Everyone is allowed to state anything about anything on the Semantic Web. If someone reuses your IRIs (or UUIDs, it can happen with them the same way) then why bothering? They can do and name what they want in their models. And maybe they really intent to refer to the same concepts. I think not all objects in an ORM model need an IRI, but as soon as you want to match two given models, it would help to refer to its objects via an IRI. Not using IRIs for identifiers is as strange as not using Unicode for character encoding. Luckily there is an IRI representation for UUID, specified in RFC 4122: just add a "urn:uuid:" prefix. A much better solution would be to create an RDF vocabulary for ORM which can then (partly) be mapped to OWL etc.