On 12/18/06, Ann Robinson <robinsona@us.ibm.com> wrote:

Thanks for the response.

First of all, I have finished the implementation for your concern about the ServiceContext properties and I am running through tests before submitting an updated patch. With this updated patch, when the message context is restored, it will look for existing ServiceContext and ServiceGroupContext objects and use those existing objects instead of unique objects. This means that the restored object graph for the MessageContext is made up of mostly existing objects (of the primary objects, only the MessageContext and OperationContext will ultimately be unique).

What if there is a premature restored version of a ServiceContext and a stored version with quite a lot of state. We will  not be able to  ever retrieve the  state saved in the  SC since all  restored  MCs will be pointing to  the one in the memory.

This is the main concern I have with the proposal you have sent. Things become much complex when it comes to rebuilding of the context hierarchy. Ignoring all other stored versions does not seem to be the right solution here.