Re: [Exim] Re: Exim features

On Wed, 10 May 2000, Nigel Metheringham wrote:

> I guess (not knowing tcpserver very well), that this is restricting the
> total number of connections to a port [of course you can almost
> certainly run exim under tcpserver as an altermative].

Yes, basically how it works is it accepts and forks until the limit is
reached and then it stops calling accept until a child exits. New
connections sit on the listen queue until that runs out then they get a
connection refused message.

This is wonderfull for sudden influxes of mail as nothing is defered and
everything is processed as fast as possible.

> I am not sure if you can see the connecting IP address prior to
> accepting the connection (ie I cannot think of a portable way) - and if
> you can't then this behaviour can't be emulated at that level.

I'm pretty sure you cannot, at least not with the BSD socket API.

> Personally I think I prefer the mailer to accept the connection and say
> it can't deal with it now rather than hanging at the connect stage - it
> gives the sender a chance to work out his retry strategy.

It depends, I think. If it hangs at the connect stage for maybe 2 seconds
thats fine by me. But if it is going to hang for say 30s then it should
probably send a message back to free up the sending process on the other
side.

> With regard to the other qmail stuff, it might be easiest to just write
> a qmail director in C and have that as an option for qmail->exim
> transition.

Well, some of them are legitimately usefull features for mailers to have
:>

Jason