I have deployed a webcenter spaces application and some custom ADF Taskflows. I have set the backing beans to pageflow scope and get lots of serialization errors.
These are do to lots of variables and method return types declared as abstract classes or interfaces such as List instead of ArrayList for example.
My question is, coding to abstract classes and interfaces is usually a good design practice, not to mention the auto generated JPA code from jdeveloper does this all over. But if I utilize any of the DAOs or Model object in my pageflow scope, or any service class that hasd them, we get serialization errors.
Is the strategy in a clustered environment really to declare all variables and return types as concrete, serializable classes or is there some other way to archictect the application to allow for serialization required in session replication in a weblogic cluster.
Secondary question, implementing a typical java singleton is a problem because each server ends up having its own copy and refreshing data is required per-server. Does weblogic or the framework provide any mechanism to have a context level object with something like operational data available to all users that is useful in a cluster?