The problem with cglib is that you can't have a proxy of a proxy. Working with interfaces always solves this problem as Spring doesn't use cglib in that case, just plain old java proxy objects (java.lang.reflect.Proxy from memory). Using interfaces also promotes loose coupling, a nice design goal.
-- Kim On 15/09/2007, Tom McGee <[EMAIL PROTECTED]> wrote: > > I used the factory approach too. See my post from 9/5/2007 "yet > another way to add spring to restlets, any opinions?" > > On 9/14/07, Adam Taft <[EMAIL PROTECTED]> wrote: > > > > > > > Interesting. I'm using lookup-method in production without any > problems. > > > But, there are no transactions and no other proxies so I guess we got > > > lucky. Do the clashes only occur within the class or does it cause > > > problems everywhere? In other words, am I safe if the resource class > is > > > not transactional, but is uses a service class that is? > > > > I believe it might be limited if you're trying to make the methods in > > the Resource class itself transactional, but I can't remember exactly. > > It might have happened in your scenario too. > > > > The problem for me was that transactions which should have rolled back > > didn't. This is really sneaky, because I had to crank up the spring > > logging to 'debug' in order to see the transaction output. Then, I had > > to throw an exception in one of my DAO methods and watch to see if the > > transaction rolled back. > > > > What solved it was by not using lookup-method and instead using a > > different factory approach to creating Resources. This allowed me to > > remove the CGLib dependency out of my configuration, which I personally > > like (the fewer dependencies that rely on "magic" the better, in my > mind). > > > > Adam > > >
