Semantic Web and RDF

It seems to me a natural consequence of the increasing bridging of traditional data sources into RDF is that the number of "walled gardens" of RDF is increasing. Walled gardens are areas from which there is no escape, and while they certainly have their uses (in particular controlling who can access what), they erode the overall usefulness of the system as a whole, by meaning that you literally can't get to there from here.

Examples of RDF walled gardens include the FOAF generated by such systems as tribe.net and livejournal. While it is useful that these sites are exposing data in a machine readable web 2.0 format such as RDF/FOAF, the inability to link to people, resources and interests outside the walled garden represent a significant barrier to interoperability. I am the same person in my blog RSS feed, my flickr photo feed, my del.icio.us feed , my tribe FOAF and my hand-edited FOAF, but only the last can I link to the others, despite the fact that they are all in the same format.

I've been tinkering with RDF for a while now, and it's great to see so many sites generating RDF of one kind or another, whether it's RDF/RSS news feeds, RDF/FOAF personal information, RDF/EXIF photo information or RDF/Dublin-core information on documents. Unfortunately there's a little fly in the ointment: people aren't validating the RDF they're creating to ensure that it is valid RDF.

The two most common mistakes are:

• unescaped "&" characters, which result in output that isn't even XML, let alone RDF. This mistake seems to be found most frequently in RDF generated from content management systems and blogging platforms. The solution is to escape the & as &.
• missing namespace declarations, which seems most common in hand-edited and small-scale RDF. The most popular namespace to forget appears to be that for the "rdf" prefix, presumably because some tools have a sensible default for this and somehow this default doesn't make it into the final file.

In the last week I've seen literally hundreds of files with each of these errors. 

A friend of mine, Paul Trafford, has observed that the Mathematics Genealogy Project appears to be slowing to a crawl, in more ways than one.

The centrally organised database project collates a research genealogy for mathematics, recording who studied under whom for their Ph.D. Unfortunately there appears to be currently a single grad student maintaining the database and updating takes a week.

This sounds like an ideal use for RDF/FOAF to me. Individuals host and update their own data (which naturally they are the authorities on), with links to the libraries holding copies of the thesis as evidence that the thing exists. They also link (naturally) to their supervisors and colleagues data, and so a crawler can build up a web. People can then build their own interfaces and more importantly, data can be updated in real time.

But, maybe I just have RDF on the brain.