Lazy loading differences Toplink vs. Hibernate - plz. explain

Question

I'm in the process of evaluating both Toplink and hibernate as potential ORM framework to use for a project I'm currently involved in.
My question is about how Toplink resolves lazily loaded associations. In hibernate I have to perform a query inside a transactional context boundary, like:
Session s = SessionFactory.getSession();
s.beginTransaction();
... your query logic here
s.getTransaction().commit();When the query involves associations which are declared as lazily loadable, trying to invoke these associations after the transaction boundary has been closed, results in an exception. This differs from Toplink (my JUnit testcase breaks for Hibernate if I set the DAOFactory to return hibernate enabled DAO's) and I'm wondering why?
I'm guessing this has something to do with how Toplink manages its clientsession, but would like to get some confirmation about this. It looks like as-long as the thread of execution is running I can resolve associations using Toplink, but not when I use Hibernate.
This brings me to yet another question - what's considered best practices in Toplink regarding session disposal. Should I do something myself, or let the garbage collector take care of it?

Answer 1

I'm not too sure here, but I think it's because TopLink has a "long running" ServerSession. When you do lazy initialization outside a clientsession it is for read only purposes and it will use the centrally managed ServerSession (and cache). I'm still trying to figure out the exact relationships here, som I'm not too sure. :) Hibernate does not have a centrally shared cache, and will not be able to instantiate objects if the session is closed (for each session, objects are instantiated from it's data cache).
As for handling resources and closing/opening, use the Spring/TopLink integration. It will handle it for you and give you heaps of convenience methods that uses some clever tricks to decide if it should fetch objects with Session or UnitOfWork. It will also do some good Exception handling built into Spring.