----- Original Message -----
From: "SM" <sm(_at_)resistor(_dot_)net>
To: <ietf-smtp(_at_)imc(_dot_)org>
At 12:38 30-08-2006, Hector Santos wrote:
We have implemented a GreyListing system. The "specs" indicate
to use a 451>response.
This isn't an error in processing. It is policy related and would a
450 is better suited.
Based on your report, they are blocking you for 300 seconds or
5 minutes. That is definitely too long in my opinion and in the
opinion of others in the greylist support list.
Retrying within one minute is too short. Note that the reply code is
not specific to greylisting.
Agree. I should first note that our GreyListing system is still in Field
Testing (about 6-7 months) and while it works really well, I'm still
extremely reluctant to release as part of our total package for obvious
reasons. So we might released it as a non-obligatory separate downloadable
add-on.
There are no official "IETF" draft for greylisting (that I am aware of).
The official guide at:
http://projects.puremagic.com/greylisting/whitepaper.html
recommends a 451. I too was conscious of the specific 451 and first used
450. But currently I have it defaulting to 451. I think most systems just
look at 45x anyway and could care less for the single digit value to change
the temporary negative response retry logic.
In my testing, I found many multi-machine (same sub-net) outbound systems
retrying almost immediately, so I compromise on a 1 minute default block. 5
minutes was too high of a block time for the multi-machine outbound farms.
Our GreyListing system has been on auto-pilot with shocking and amazing
results showing to me that the most systems are GOOD and most BAD systems
don't care. :-) We keep the all payload just in case and I see 0% False
Positives. But I am still reluctant on signing off on it ;-)
No doubt, it needs lots of smarts in auto-whitelisting concepts otherwise
you will get the reports from the layman who is *watching* sending mail and
wonder why it fails which is not so cool in the commercial environments.
--
Hector Santos, Santronics Software, Inc.
http://www.santronics.com