| |||||||||||||||||||||||||||||||
|
[ Charset ISO-8859-1 unsupported, converting... ] > On 5/28/07, Wietse Venema <wietse> wrote: > > Kim Minh Kaplan: > > > Hello, > > > > > > I have written Greyfix? (yet another) greylisting policy server for > > > Postfix. The best current practice in Postfix to defer delivery for > > > greylisting seems to be to return action=defer_if_permit. The problem > > > is that this rejects the mail with an SMTP code 450 whereas it would > > > be better to use an SMTP code 451?. > > > > What evidence do you have that this will change client behavior? > > Evidence is too strong a word. But some experiments I did about a > month ago with Gmail seem to indicate that they use a very large pool > of servers (more than /24 network) and that after a 450 response they > would change mail server across their /24. Now that I changed to a > 451 response they retry using close mail servers. I'll make these > experiments another time to verify if they changed their system during > this one month period. If you are interested in the resulting logs, I > can post them here. > > Please note that the initial idea that it will change client behavior > is not mine. I stole it from > http://cvs.puremagic.com/viewcvs/greylisting/schema/whitelist_ip.txt?rev=HEAD > which is the home of the Relaydelay greylisting software : > > please check that your implementation uses the 451 error code (not 450 > or another 4xx code). Some problems have been reported for sites like > MSN/Hotmail, Prodigy, and various other senders that appear to be > having "weird" retry patterns (sometimes resulting in bounces) when > using code 450 or others. I would like to see some evidence. There is a lot of information on the internet that gets copied from one site to the other without proper verification. Not that I would be surprised that there are broken mail systems, but if they can't properly retry after 4xx then they will likely have lots of problems. Wietse
| ||||||||||||||||||||||||||||||
© 2004-2008 readlist.com