Massive queue buildup

Matt Kettler mkettler at
Tue Nov 14 00:35:43 GMT 2006

Brent Addis wrote:
> greylisting is bad and I will never ever run it. I know of one case where greylisting has very nearly killed someone. They probably shouldn't have been using email to do what they were, but thats not the point.

But it is the point. If time is in any way critical, email isn't for you. PERIOD.

Greylisting didn't nearly kill someone, some person used email where time
criticality was a life-or-death issue. That person's bad choice of communication
methods nearly killed someone. Greylisting has nothing to do with it.

Quite frankly, any spam control technology, like say, SpamAssassin, could have
FPed on the message, causing it to possibly be ignored by the recipient. Would
you say SA nearly killed someone? Would you stop using it and insist everyone
else do the same?

What if one of the routers in that email path had crashed or had a hardware
failure, would say that Cisco nearly killed someone? Would you stop using Cisco
products and insist everyone else do the same?

Bad disk on the mailserver? Seagate's a killer?

Loss of power? BGE? Yeah, definitely don't use electricity.. it's unreliable and
could kill someone.

Let's face it, the email user in question is lucky THEY didn't kill someone with
their mistake.

Greylisting is not to blame here. Keep the life-and-death dramatics of someone's
mistakes out of it.

> greylisting adds an unnecessary delay to email, and can quite easilly be beaten if a spammer sets his mind to it.

It doesn't add unnecessary delay to most messages, Not if you do it *right*.

So far this week (today and Sunday) my greylist has:
	handled 16,493 total messages
 	delayed 12,238 messages.
	allowed 4,255 messages to be delivered without delay.
	accepted 330 messages after delay.

Of the 330 delayed messages, only 9 were not tagged as spam by SA. Of these 9, 2
were spams that SA failed to tag, 5 were mass-mailed newsletters (delivery speed
not important), and only 2 were personal messages.

So 2 significant FPs out of 16,493 messages. 0.01% error rate, not too bad.

Any spam control technology has it's downfalls. loss, delay, or depriortization
of mail will be a side effect of any of these systems in some cases.

If you do greylisting right, you can keep the delays down to a sane level and
still hack off a lot of spam. Approximately 72.2% of the inbound mail has been

More information about the MailScanner mailing list