A have to classes A and B.
They are connected via one to many relation.

I create new object a of type A:
A a = new A(); // currently not persistent

create new object b of type B:
B b = new B(); // currently not persistent

And say:
a.SetOfB.Add(b);

// class A contains
// <set name="SetOfB" inverse="false" lazy="true"
// cascade="save-update">
// <key column="a_id" />
// <one-to-many class="B" />
// </set>


then call: session.Save(a);

WHY the behavior of NH is:
1) insert in B with foreign key to A a_id = NULL
2) insert in A
3) update a_id in B with inserted a_id in A

instead of insert in A then insert in B?

In this situation I cannot make B.a_id not nullable :(((
And when I make it nullable I need more calls of Save(..)

First for a, then for b.
And it doesn't suits the concepts of my application.

I am searching for a good explanation of that behavior to understand the "why" behind it.

That topic is the subject of a couple of unanswered forum messages and I think it deserves a good explanation - I have been looking for one in that forum, in the "best book" about Hibernate, Hibernate reference documentation and the web in general for a couple hours yesterday without success.

I have a couple URLs that I can share:

Those two are to show that the topic is still unanswered:

http://opensource.atlassian.com/projects/hibernate/browse/HHH-2280

http://forum.hibernate.org/viewtopic.php?t=957582&postdays=0&postorder=asc&highlight=foriegn+foreign+key+nullable&start=0

This one provides good insight, but what we need is a black and white sentence somewhere in Hibernate reference and a clear answer on the forum from the Hibernate team to support it:
http://www.cs.bham.ac.uk/~aps/syllabi/2006_2007/issws/h02/hibernate.html#one_to_many_unidir

Thanks in advance.