I can remember looking at rspamd years ago. To me it looks like a choas of applied rules. I can remember asking these devs some questions about rule processing order, and they could not even answer them. Do they still make own graphs? That already tells you their whole approach is old-fashioned.
If you want full control and efficiency you should take this mailfromd. This maintainer Sergey is also an ace in his area of expertise. > > Replace SA by rspamd ? SA is unbearably slow. You have 14 mx > hosts, get 14 rspamd vhosts and I doubt you'll have any issues. > > Scott > > Hi Scott, > > On Sat, Nov 30, 2024 at 2:10 AM Scott Q. wrote: > > > I see this thread is getting quite personal, but if I may comment on > > something you wrote. > > > > >> I more or less said that, you are a mail administrator yes? then > you > > know bloody damn well that 99% of spam checks are carried out after > > reception of the mail > > > > depends what you define by reception. We perform spam checks on the > DATA > > portion of the mail reception and can refuse a message in that > segment. > > Therefore, spam check is done on the message before accepting it. > > > > > TLDR: reasonableness for testing in DATA depends on your traffic > levels > > LV: > I understand some small players may do this, however most don't, In my > "part-time contractor role" outside of my day-job, nearly everyone > uses > postfix/amavisd/dovecot in the years I've seen nearly none of them use > milters to test they accept it then hand it off to spamassassin via > amavisd, even of the couple of exim installs I've worked on, only one > used > a weird integration that tested in DATA, but as a small law firm they > weren't high traffic so can do that. > > In my regular day-job role as network administrator for a ISP, and 2 > other > ISP's over past 16 years, none have done this, these ISP's are in > Asia and > U.S., if a message is accepted it then passed on to anti spam > measures, > including current employer who operates 14 inbound MX's alone, they > are no > gmail or outlook but are high traffic where each inbound machine > processes > around 360 connections per minute, you do the sums on the amount of > delays > that will incrementally occur testing in DATA to still deliver mail in > reasonable time after received, won't take long before those machines > are > to be running way behind and trigger Hi-Q alerts. > > > Loz > _______________________________________________ > dovecot mailing list -- dovecot@dovecot.org > To unsubscribe send an email to dovecot-le...@dovecot.org > _______________________________________________ > dovecot mailing list -- dovecot@dovecot.org > To unsubscribe send an email to dovecot-le...@dovecot.org _______________________________________________ dovecot mailing list -- dovecot@dovecot.org To unsubscribe send an email to dovecot-le...@dovecot.org
