On Thu, 14 Jan 2010, LuKreme wrote: > On 14-Jan-2010, at 06:22, Robert Schetterer wrote: > > http://savannah.nongnu.org/projects/spamass-milt/ > > How efficient is spamass-milter? I've always been hesitant to try running SA > during the transaction because I was afraid it would take too long. >
I cannot speak for spamass-milter as I use a different milter (milterassassin) but the general concept of filtering at SMTP time is viable (depending upon your load) with a few considerations. 1) structure your filter stack so that SA runs after all lightweight filters (DNSBL, helo checks, valid recipient checks, gray-listing, etc). 2) Make sure that your milter uses SA intelligently, not opening premature connections to spamd (if your milter uses spamd rather than running SA directly in the milter; EG amavisd). The milter I use talks the SA net protocol directly (as opposed to forking spamc) and originally opened the connection to spamd when it got the receipt-from info. This would waste connections as unnecessary as recipient checks often would kill the SMTP transaction. I re-coded it to collect headers and not open the spamd connection until it reached the data-phase. 3) Adjust your MTA to limit the number of simultaneous incoming connections to the max number of spamd processes that your SA box(s) can reasonably handle. With these considerations we comfortable handle 100K messages/day with just one moderately sized SA box. -- Dave Funk University of Iowa <dbfunk (at) engineering.uiowa.edu> College of Engineering 319/335-5751 FAX: 319/384-0549 1256 Seamans Center Sys_admin/Postmaster/cell_admin Iowa City, IA 52242-1527 #include <std_disclaimer.h> Better is not better, 'standard' is better. B{