This is an RFE asking for the rmiregistry to be a bit more bullet prove much like rpcbind process. The problem is that if we have a bunch of rmi services
registered and the rmiregistry dies for some reason then all the services must
be re-registered upon startup of the new rmiregistry. This means the services
are restarted.

In the case of rpcbind it has the capability to dump out its state that can
be read again on startup, thus the rpc services need not be restarted again.

This RFE is asking for the rmiregistry to go one step further than the rpcbind
and that is to dump out its state after it has started up and/or when a new
service is registered with it. This means that if the process dies it can
be started again without the need for all the services having to be restarted.

As it stands the rmiregistry is a single point of failure.