Re: Fundamental architecture of a proxy implementing keep-alive.

"James A. Donald" wrote:
>
> From: Dancer [mailto:dancer@zeor.simegen.com]
> > I mislike me the use of threads. Having written a couple small http
> > proxies recently, I find that state-machines seem to work better
> > than threaded models.
>
> Are you suggesting thread and one state machine for each client-proxy
> connection,
> or one thread and one state machine for all client-proxy connections?

A single thread of execution for everyone. select() or poll() coupled
with non-blocking connect()s and such. I've written both models
recently, and (bizarrely) the single-thread model is cleaner and
simpler...well...from my rather abstruse sense of aesthetics it is,
anyway. It's certainly faster and less hungry.

D
Received on Wed Aug 25 1999 - 20:15:04 MDT