You never rely on finalizers to clean up resources. At best, you check that the resource is properly released and signal an error otherwise, because you should have released it earlier (you have a bug). That may happen when you manually manage your resources, but most of the time, you only need to use a resource inside a delimited block (defer, finally, unwind-protect, with X). If I understand correctly, this is how RAII works too when you allocate objects on the stack.
You're missing the point. The original question (and hence my answer) isn't about whether or not it is a good idea to use finalizers. I'm merely pointing out that both in the case of RAII and a GC, only the lifetime of memory allocated is being managed. Not the resource contained in it. They use different mechanisms to let the developer deal with the resource. In the case of RAII the time at which clean up code is executed is deterministic (destructors), whereas with a GC, it is not (finalizers). Said mechanism, can be implemented, should a language choose to, regardless of whether or not the underlying memory allocation scheme is automatic (GC). (IDisposable is mostly a made up thing that C# has syntactic sugar for, that lets developers eagerly release resources (other than memory) when they're done with it. My point is, neither the language nor the runtime makes any effort to enforce its usage, like it does in the case of memory allocation.)
