It might be good to include a note about when serialization across JVMs is useful, e.g. when using a distributed second level cache in a cluster.