[mattmcknight]

I have run into serious issues with the document based approach. Imagine a domain where you have several different types of documents. You still end up doing joins. In the example the author gives, doing a query to get you all of the unique tags can be slow. Or, if they are dates, trying to get a max(date) can be tricky. The biggest problem though, is document size. For a complex domain, there are big tradeoffs between keeping the entire graph in the document, which is slow to read, or creating a variety of document types, which brings you back to joining.

A hybrid approach can give you a little of both. For most of the kind of applications I build, I use relational databases, but I dump a denormalized view of the data into Lucene for searching and such and use a caching layer for most reads. This way I get some document type performance on many operations, but still have the normalization and simple transactions of a relational model.