Through much of the book, you explicitly set cascade on the particular relationship, usually on the many (or Set) side. You talk about (pp. 136) where this could be a performance hit when setting cascade on the "one" side of a 1:M. However, this is in the context of a recursive relationship.

Now suppose you don't have recursive relationships and you set default-cascade across all your VOs, what would the implication be of this? On the "Set" side, I see no difference but on the <many-to-one>'s I'm assuming those would take a hit, albeit smaller because there's no recursion.

Can you speak to your experiences of using this feature of Hibernate in the context of performance?

Thanks,
Lou

Think about detached objects. As discussed in chapter 4, Hibernate would have to "walk" in directions on reattachment unneccessarily. Cascades are not only relevant for save and delete, update/reattachment is one of the motiviations for fine-grained reachability control.