I just read a post from Ayende about the sad state of NHibernate code generation. Between the comments, I spotted a few ones from Frans Bouma.
Exactly one year ago (well, minus a couple of days really), after a thoroughly careful evaluation, I decided to purchase LLBLGen Pro's license. I liked the tool since I first used it and in a very short time it let me to become very productive. I think I was a quite advanced user, I wrote new templates, inserted base classes and stuffs like that.
Problem was, a couple of people in the team never really grokked it, and one of them was the manager. I was sincerely not understanding why they were not able to be as productive as me, so I did some research looking for something to help them to become more proficient, and I found a very clear article written by Frans Bouma. I referred that to my team mates, and after reading that, my manager and some other team member found themselves more in camp 3 (the domain model approach) than 2 (the entity approach). At the time, after evaluating what the market was offering, we switched to ActiveRecord + NHibernate + NHibernate.Generics (incidentally, I was totally against using NHibernate without ActiveRecord due some previous hellish experiences with older versions of Hibernate). One of the alternatives we tested was Genome, which I remember to be quite promising, but which at the time was having very sloppy performances (but it was by no means the worst between the ones we evalueted).
I found that personally I feel comfortable with both the approaches (as long as people don't commit repository breaking code, I am comfortable with almost anything), but some people seems not to be so flexible, so I do sincerely believe it could be a good strategy for LLBLGen Pro to start supporting NHibernate templates (and, even better, ActiveRecord and Gentle.Net templates as well).