Thinking about this common problem I thought one day on making some form of communication between network instances of the applications.

Perhaps the better solution would be to implement some form of PING. This can be a little ugly to security experts, but I was wandering about each instance opening a different TCP/IP port (sequentially assigned upon entry). The shutdown process would clean up the record for this instance/port.

Whenever you want to check who are logged, you have to ping each IP address and port, and that way you are checking whether the app is actually live and running.

Another option (less real time but easier to implement) would involve setting a global timer (with a low ticker, to avoid gobbling up processing and network resources that keeps updating some table or a plain ascii file (one per instance, in a special directory).

When time for an actitity check comes, you have to verify the last update time for each instance, wait for the timer interval (i.e. 1 minute), and then verify again. Any instance that didn't update its log would be considered dead.

Of course in either case any workstation has to have a process to clean up its old logs (being those table record, ascii files, or whatever) to avoid creeping.

The last possibility would involv COM components and some form of inter-instance communication.

I didn't went trough on any of this ideas, but maybe someone could pick somethign up and use it.