Re: reliability principles

> If a user makes a mistake and misdirects a message to an incorrect mailbox,

It isn't the user's mistake. It's yours.

> If we have one message failure in the next 500 years

If nobody but you uses exim, and you never deliver very much mail, a
failure rate of 10^{-9} is fine.

However, if exim is used on hundreds of thousands of hosts around the
net, including hundreds of hosts that deliver over 100000 messages every
day, a failure rate of 10^{-9} is terrible.

> You appear to be slagging off exim without pointing out the problems,

In fact, it seems that you've run into reliability problems before.
spool_in.c goes out of its way to detect one of the simplest forms of
corruption, producible only by a system crash. If it finds it, it skips
the message and complains to the sysadmin. Never mind the fact that the
envelope information has already been destroyed.

In short, you reacted to a reliability problem not by fixing it, but by
attempting to limit the damage from one of its symptoms---right?

> and telling people how to run their systems.

I'm not ``telling people''---I'm asking you whether you're going to
comply with RFC 1123.

> To get things in perspective, where does exim fit in in your view of MTAs ?

I haven't learned enough about it to give a thorough assessment.

---Dan