We've fixed all the known bugs; that just leaves the unknown ones. We've got JRDF, Jena, and our own RDQL implementation. The loading and query speed has also improved considerably - there's none of the problems associated previously with 32 bit systems. Kowari lite is down to 11MB; which considering that Jena and related jars are over 2MB that's pretty good. Querying is now totally disk backed - you can do a completely unconstrained query over arbitrarily large data sets.
It's now at the stage of what to do next. Some of them are fairly obvious like full data type support, scripting (possibly Groovy) and inferencing. But we've also got zeroconf and MAYBE. There's also a large refactoring underway to make Lucene, the triple store, ontology models and others as plugins exposed via a resolver interface. This should allow just about anything to be queried in the FROM clause of iTQL. In the short term, this means extending the querying of file and http RDF sources to include other metadata formats such as MP3 tags, iCal, vCard and XML based RSS. It will allow models to be typed and configured. Using views you could easily create an RSS aggregator. Other, more difficult resolver implementations will be available in TKS.
JRDF has been a little neglected. The next things to be worked on is transactions and a ResultSet like interface.