Well, it always depends on your architecture... For instance, you can have session beans with remote views (to be accessed form anywhere) that access entity beans with local views. Session and entity beans run in the same JVM on the same server. That way, the access of entity bean is much quicker as everything goes through their local interfaces. This kind of scenario can be used when Swing applications want to access entity beans on some server.
If you develop an application where the only way to manipulate entity beans is through session beans (Session fa�ade design
pattern) and you now at the beginning that all the beans will be deployed in the same application, it does not make big sense to have entity beans with remote views... Everytime you want to access an entity bean, a RMI call is issued on the network with all the overhead of un/marshalling going on, etc... It always makes sense to think about the architecture of your application before deciding if you go with local or remote (or both) views.
