[LCP]design issues

[devraj sanyal]

The server architecture i had in mind was of a socket binding on a 
port
and waiting for requests. once a request comes it spawns
a thread/process to handle that request. i believe the original 
question
was whether its better to spawn a thread or fork a process.
My comments were to use processes and use message queues and 
shared
memories to communicate between the original server process and 
the
forked child process.

Dont really know of any method where you can run a fork or a 
separate
process in a remote machine. if message size is unlimited 
message
queue will not be of much help rest i believe is perfectly 
scalable.




On Sun, 31 Aug 2003 Kurt Mueller wrote :
>On Saturday, August 30, 2003, at 10:10 AM, devraj sanyal wrote:
>
>>Go for shared memory and message queues.
>
>This only works if client and server are on the same computer.
>You will never be able to scale the application, and
>distribute it over more than one computer.
>
>
>>But threads are helpful to manage sessions. With IPC managing
>>sessions is tough and a separate module for session/connection
>>management must be written.
>
>Dont understand this point.
>Can you please elaborate on that?
>
>
>Grüessli
>-- Kurt Müller, mu at problemlos.ch
>
>
>_______________________________________________
>This is the Linux C Programming List
>:  http://lists.linux.org.au/listinfo/linuxcprogramming List