On 2010-01-29 10:59, Adam Thompson wrote:
On 2010-Jan-29 09:13, John Lange wrote:
A 29 minute grey list!? Are the people using this server really ok with a minimum 30 minute delay in getting their mail?
I had my personal mail server set to 2hrs in the very recent past, but with reasonably extensive whitelisting.
I for one welcome a return to delayed e-mail delivery!... ;)
Given that mail zombie's usually only try once for each email address, I have mine set to 30 seconds.
Granted, the bots are getting more sophisticated and sometimes try more often but it's still effective.
Switched to postgrey, which had a default of 5 minutes, and noticed that such a short delay was permitting an awful lot of spam. Boosted it to 10 minutes, which cut things back down again (in combination with policyd-weighted, anyway).
Considering that e-mail was never meant to be near-real-time in the first place... it wasn't so long ago that sendmail's default operation mode was queued-only on a 30-minute (or longer!) interval from cron.
I'm not sure what's typical these days, but I would assume 15 minutes or so would be a fairly common retry interval. That's why I set my greylisting delay on the CompSci mail server to 14 minutes. (And that's with a default whitelist policy, so it only affects particular addresses which are more likely to be spam senders or recipients.)
On the MUUG server, I caved to pressure from some mailing list subscribers, and reduced the delay to 4 minutes (i.e. just under the 5 minute retry interval of some more aggressive mail servers). This seems to be working reasonably well.
I've found that some spambots don't retry at all, in which case even a small delay is sufficient; some seem to retry quite persistently, in which case even a larger delay won't help. However, there is a 3rd type, which will retry for a short while before giving up. (We see these in our logs with a retry interval of a minute or less sometimes.) It's for that 3rd type that the retry delay will be more critical. In any case, I wouldn't see any need for a delay of more than 14 minutes.
Heck, it wasn't even all that long ago that I got mail whenever I connected to my ISP and did an ETRN, twice a day.
So a 30-minute delay in receiving mail seems perfectly fine to me. Plus, I would MUCH RATHER people believe that e-mail does NOT reach me immediately; the expectation of immediate delivery - implying immediate action, immediate comprehension and immediate reply - has been shown in multiple studies to destroy worker productivity.
A valid point, but I don't think I'd want to use greylisting delays to try and enforce this sort of productivity enhancement. :)