Milter-greylist

Ugo Bellavance wrote:

> A question that I have with a 1 minute delay, what is the average
> delay that your users have to live with?  I guess most SMTP clients
> won't knock again after 2 min right?

Right.. However most legitimate messages retry in 15 mins. Also, I don't
greylist all my mail. I default to white, and selectively greylist, so my
"average delay" is pretty much 0. Very few greylisted messages ever get
delivered, and nearly all that do are spam.


On a quick survey of delays for this week:

delayed for 00:01:00 (pre-accept first try):
  17802
not delayed and delivered (total)
  13050
not delayed and delivered (not spam tagged)
   6368
not delayed and delivered (spam tagged)
   6682
delivered after being delayed (total)
    298
delivered after being delayed (not spam tagged)
     44
delivered after being delayed (spam tagged)
    254

And of the 298 messages that were delayed, here's the distribution of time:

109 < 15mins
74  15min-1hr
98 1hr-12 hrs
17 > 24hrs

I can't easily do a breakdown of the 44 that were non-spam, I'd have to collect
the SMTP transaction IDs from the SA messages, and then backtrack.


A little SpamAssassin ruleset that I use to track greylisted vs ungreylisted
spam is this:

header INFO_GREYLIST_DELAYED        X-Greylist =~ /Delayed for /
score INFO_GREYLIST_DELAYED                       0.001

header INFO_GREYLIST_NOTDELAYED        X-Greylist =~ / not delayed by /
score INFO_GREYLIST_NOTDELAYED                    -0.001

This way I can grep for those two in my SA logs. However, the timing info comes
from milter-greylist's logging. (I use MailScanner which batch-scans mail, so
the log lines don't always follow each other)