This interview with the architect of MySQL looks very cool indeed. Rather than just hoicking data out of files on a big hard drive, the latest version can use a Web site, or multiple Web sites, as a source of data on which database operations can be performed and the results served up to something else. Among other things, it means that if you can do a specific stable URL and catch the get and post requests for that in some interesting fashion, your website can be an element in someone else’s application, which could of course be the database back-end for their website.
It would be interesting to see if some other protocols might fit in there, in which case this could be a very interesting development mechanism.