Configuration data should just be primitives, strings, etc., right? Do you also persist custom objects where class definitions change from deployment to deployment?
Yes we do,
We have a number of small components. These components are deployed on a number of different WebLogic cluster nodes. The components need to be configured at run-time from whitin a central web application. The configuration is stored in a cache whitin the web app and is synchonized to all components running in the cluster.
To simplify the reconfiguration of the component we have chosen to put custom objects in the cache. So we don't use primitive types directly.
Is the use of primitives in the cache the only method to avoid the problem we have?
For something like this, where the class definitions change from deployment to deployment, the best thing I can think of is to use primitives.