I liked Marcelo’s post about “Uploading WikiPedia Dumps to Oracle databases” very much. One reason is that he actually did that, what I was trying to explain / trying to bring across, to my audience during UKOUG. If you are going to work with the XMLDB functionality, think about what you trying to achieve regarding storage and going for the Full Monty. The “think about your logical design, XSD and the wanted storage result with its index possibilities”.
I hope that some people picked up, at least, that message during my UKOUG presentation, because it will avoid a lot of frustration and disappointment regarding XMLDB functionality. I already have seen to many “XMLDB performance is horrible” statements on the OTN XMLDB Forum, without being based on a decent planned and thought through design (method), but then again, we are nowadays database agnostic / J2EE compliant, so who thinks about database design anyway (= “grumpy old men remark”). It is sometimes a small miracle that people like APEX…
Marcelo also went for a an alternative way regarding indexing, using a Lucine Domain Index. You can argue if a method based on Oracle JVM will work in the long run, but I like the “out of the box” thinking trying to solve, in this case as he argued it, maintenance problems.
Two very nice posts on XMLDB in one week, you would almost start to think it is becoming popular.