On the long, slow death of databases
Read Write Web has just published an interesting summary piece about the semantic web, that predicts the eventual demise of relational databases in favour of "structure on the fly" searching.
The world is clearly too complex to structure upfront... Structure on the fly is done by people adding structure as they use the service and by engines that automatically create structure from unstructured content... Innovation has slowed in the RDBMS world - with open source at one end and Oracle at the other, there is little reason to innovate - just when Semantic Web innovation is accelerating.
This is an interesting idea, but from a technical point of view I'm really not sure if it is relevant to the promotion of the semantic web. Anything that searches through lots of data needs some kind of index (even Google) and many of these indexes are stored in RDBMS platforms because they offer the quickest way of querying the data.
I would have thought that the growth of the semantic web relies on the search spiders gathering the structured data in a more intelligent way (e.g. recognizing microformats, etc.) and recording that in the index. The technical platform used to then query that index is almost irrelevant, as long as it does it quickly and accurately.
Even on an enterprise level, I honestly cannot see a situation in the near future where network and client computer speeds are quick enough to allow on the fly querying of thousands of documents stored on hundreds of machines. I think we'll still need that central index, and that it is probably the search spiders that need to improve the quality of the information they gather before the technical platform of the index requires updating.