« SMS URI Scheme Update | Main | Distribution and the Semantic Web »

Tuesday, August 04, 2009


Feed You can follow this conversation by subscribing to the comment feed for this post.

Patrick Murray-John

Well, not quite. Just as there are piles and piles of XML DTDs and Schemas out there to choose from, each of which tackles a particular domain from a particular perspective, there is also a growing pile of semantic web vocabularies out there for tackling a particular domain from a particular perspective. The point of expressing them all in RDF is to provide easy ways for the data expressed via those different vocabularies to play well together. XHTML, I think, aimed at similar goals, but didn't go over so well. RDF has a similar problem, perhaps, but the goal is same: provide a way to express data in a format that makes it easy to mix data from different domains and vocabularies together. There are different XML schemas for different things, and there are different semantic web vocabs for different things. RDF just provides a way for them to play well together that has, I think, more potential than mixing things up with XML namespaces.

That said, I see your point about effectiveness. Totally. But, I think that in terms of data-mes(s)hing-together, RDF provides an extraordinarily effective mechanism. The real problem is how to deal with all that mes(s)hiness, especially when it comes to user interfaces. From that angle, it really is still a messh out there in the linked data world.

I'll try out an X(HT)ML:CSS :: RDF:???? analogy. CSS has done really, really well at demonstrating how nice, clean separation of content and display can make life easier for everyone, especially web developers. Semantic web folks, I think, don't yet have the last part of that analogy. Maybe we're waiting for a messiah -- more likely we've been concentrating too much on the effectiveness of the data-messhing, and haven't been able to convince the user experience gurus to explore what they could to with RDF. Tom Heath is leading the way there.

When it all comes (finally!) together, I really do think that it'll make everyone more effective at things we hadn't thought about before.



@patrick: thanks for your comment. and i absolutely agree that RDF is better at simply mixing things together than probably anything else out there. which is great and works like a charm as long as your domain model can be appropriately expressed in some RDF schema. however, if you have data that just does not map too well to the highly generalized model of RDF, let's say you have deeply nested data and a lot of sequential data, then you might still be able to easily mash it up, but you're just not able anymore to work effectively with your data. that was my main point, and just to be sure i am not misunderstood: RDF is great for many applications and has very nice properties when it comes to mixing things.

another area where the mixability of RDF has some unintended side-effects is when it comes to provenance: since it is so easy to mix things, it also is easy to lose track of where they came from. this is fine if you just want to amass RDF data, but when it comes to scenarios where you need to later interact with the data, let's say update something and then write it back to where it came from, then this gets actually quite hairy. in this case, RDF's "let's mix things easily", and REST's "let's interact based on self-containing representations of resources" collide, and figuring out what to do here is still a research topic.

The comments to this entry are closed.