Raw

This is stuff I'd typed up just after getting back from Bristol, originally meant as a regular blog post but never finsihed. Put it here, it'll get indexed (got RDF!) can find it again faster than the iMac says,. Work commitments not declined, but now I been outed. Won't you take me to. Funky town. Well, I talked about it, talked about it, talk about it...

Next actions

Things I've got on my plate codewise for the next few weeks. The target apps for a couple of these are NDA'd, but I believe I'm ok to talk about the code specifics.

Validation stuff: requirement is a tool that can ensure RDF instance data conforms to a given set of constraints prior to insertion in a data store. I've spent a while (far too long) on this already, mostly looking at trying to plug together existing disparate tools/techniques: RDF/OWL-based reasoning, Eyeball style heuristic checks, Schemarama2 style graph structure checks (mashing up SPARQL query results with rules). Conveniently, a couple of days ago I got to hear Dave Reynolds talk about the different reasoners in Jena, Chris Dollins talk about his recent revision of Eyeball, and a lot of people talking about SPARQL and rules...Presentation stuff: requirement is a good way of declaratively managing query/reports. I've already sketched out an approach that seems to generally make sense and seems to work in a trial implementation. Basically rather than using a SPARQL SELECT and passing on XML results for presentation directly, use a SPARQL CONSTRUCT to build an intermediate model that is presentation oriented (i.e. doesn't feature specific domain knowledge), then do the SPARQL/XSLT thing on that. Conveniently a couple of days ago I got chance to mention this approach to Andy Seaborne, and his not-laughing-out-loud I take as a good sanity check. He'd also got some implementation tips (I'm using his ARQ/Jena query kit). My next step on this one is to turn the trial implementation into something more generally useful.Agent stuff: I was fairly enthused about JADE, the Java Agent Developent Environment, as a means to loosely-coupling various data gathering and processing subsystems. It's straightforward to use, but I've been feeling somewhat less sure about the approach after danbri had pointed out how, for the kind of thing needed for pushing around RDF, it was probably a seriously overengineered approach. So what I'm looking at for this one is extracting some of the ideas that work from JADE and flipping them to a more web-friendly (and considerably simpler) RESTful kind of implementation.Data accumulation stuff: I'd already got Leigh Dodd's Slug scutter toolkit in my sights, and conveniently a couple of days ago I got to see how  ... not sure what came next...I see you baby...shakin' that ass...shakin' that ass...