1) I can definitely see the startup time for qpsmtpd under tcpserver being an issue, so I'll be using pperl for this in production. Could I get some feeling from people how many prefork processes people are using? I am running a strictly corporate environment with ~300 heavy e-mail users.
I'm hoping to soonish [1] contribute a poll() based frontend to qpsmtpd that will allow it to scale up to many hundreds of simultaneous connections. Unfortunately it may mean some of the internals have to be refactored, but that's often a good thing.
Matt.
[1] But please don't hold your breath - I may get bored of hacking on this soon and shelve it, though that depends on just how loaded my mail server is getting - and it's certainly not improving with about 1000 spams a day [2] :-/
[2] Which is nothing compared to the server I wrote the poll() code for, which is servicing around 3 million spams a day on a single CPU box.
