This blog can't be viewed on LiveJournal. Instead see

OpenDD: Reinventing RDF?

13th Nov 2008

Today I've stumbled across OpenDD, which claims to be a format for describing social network data.

Having taken a look at their (fortunately quite short) specification, I can't help but think that this looks a lot like RDF. It identifies entities by URLs and describes properties and relationships of those entities, The only place where it deviates slightly from RDF is that the names of properties and relationships are just keywords, not URIs as they are in RDF. You could imagine just prefixing them all with something like and modelling them as RDF, though.

I've got mixed feelings about this specification. While the schema it creates looks like it could be useful for describing social network relationships, I'm not sure why that requires a whole new serialization. What's wrong with application/rdf+xml or application/turtle, for which processing libraries already exist?

The other interesting thing here is that much like RDF a given document isn't about a single subject but rather describes properties and relationships for various arbitrary entities. In order to make use of decentralized metadata like this, we need to be able to verify that the statements made by the metadata resource are authoritative.

With XFN, we know that the relationships are authoritative because they're published in the resource that is the subject of the relationships. With Atom, the <link rel="alternate"> in the resource implies that the Atom feed contains authoritative data about the resource. This could be achieved in a format like OpenDD's by removing all of the "subject" UUID attributes and having the document that links to the OpenDD data be the implied subject for all relationships. Is there an existing RDF serialization with the subject implied in this way? If so, this would seem like a good candidate for publishing social relationship data for those who dislike microformats.