Forking server.

[ivanov]

Hi!

I need to have instances of objects requested by client in separate processes. One process with object realization per client.

Can I achieve such architecture with ICE?

Is it possible solution to write an ICE server so as it starts up a new ICE server when receives a request from a new Client and then acts as ICE client of the new server dispatching requests from the particular Client to it?

Thank you in advance!
Sergey.

[marc]

No, this is not possible. You can write user-code that forks a server, but Ice doesn't do it. See also the discussion in this thread:

http://www.zeroc.com/vbulletin/showt...p?threadid=972

[ivanov]

Hello Marc!

Thank you.

But could you please answer my second question:

Is it possible solution to write a server so as it starts up a new ICE server when receives a request from a new Client and then acts as ICE client of the new server dispatching requests from the particular Client to it?

Sergey.

[marc]

Yes, but it's not very straightforward. The best example we have for this is Glacier (not Glacier 2).

Why do you want to do this? It is a rather untypical design choice for distributed object applications.

[ivanov]

I have a dynamic library that has to be instantiated for every client. I can not change the library so I am to follow this requirement.

Am I right that in the same way as GlacierStarter starts new router on demand, starter of mine can run new "server" instance and return new proxy to client?

[marc]

Originally posted by ivanov
Am I right that in the same way as GlacierStarter starts new router on demand, starter of mine can run new "server" instance and return new proxy to client?

Yes, that's correct.