For the iCite net prototype, I have already committed to working in Java, but am still considering the data format and methods of query processing. iCites aren't meant to be wedded to a particular data format, but the assumption is that a large amount of interfacing with iCites will be in some form of XML, especially XHTML and RSS.

So, I am evaluating different options about how I will implement iCites using XML as the underlying data format, and also what options there are for querying that data. For transforming the data between say XML and XHTML or RSS, it is pretty clear that I can use XSLT. However, I am not confident that XSLT and XPath are right for querying in general.

So, I have been reading up on XQuery, which looks interesting. The big corporate database players have recently submitted a Java specification request (JSR-225) for the XQuery API for Java (XQJ).

The open source Qexo - The GNU Kawa implementation of XQuery looks interesting. Does anyone have experience with this? What do you think? What do you think about XQuery?

Prior to this iCite project, I have implemented a more constrained form of the iCite model using RDBMS and SQL. This could work for iCites, but the issue that I don't think I want to get into is that each iCite would best be done as a distinct table (because each iCite can have a different data model).

And, for example, if each blog post of a blog or each web page of a website is an iCite and has its own table in a database, that is a lot of tables being created. Seems much less of a stretch to think about these as XML documents, which suggest more extensibility, both in structure and in quantity. This goes well with the web services and XHTML / RSS interfaces to the iCites—in other words, an iCite can be stored as XML, be transferred as XML, and be displayed as XML.

So, I am researching the best way to do things like RDBMS joins in XML. Is XQuery it? What do you think?

(Note: another option I am considering is doing all the data storage in Java objects, and using Pervalyer to persist them. But, I think using objects, I would have to dynamically generate class definitions for new data structures, or use classes that internally store hashtables of attribute / value pairs for the data structures. Plus, then I need to create or find an object query language. This all doesn't impress on me that Java objects are a good way to do this.)