>The hard part in evaluating these technologies is finding independent research. Each OODBMS vendor will skew things to their benefit.

Even worse, every xDBMS vendor says their solution is good enough.

>The reason I pointed you to Ted's blog is because he's well known in both the Java and .Net camps and has attempted to write his own O/R Mapper more than once. He knows his stuff and knows it well.

Thanks, it will be interesting to know his opinion about Cache'.

The hardest thing in understending impedance mistmatch is to understand that there is no problem between OOP and RDBMS. There is a problem between entity-centric model we need to implement advanced logic and document-centric model that still recide on RDB part.
Databases tend to incorporate logic (with stored procedures for example) and became applications. While applications tend to incorporate stored data. I believe distinguishing of database and software is a main error which causes mistmatching. In future we must never separate them.