Hello Everyone, So first about Zimbra and its Dspam integration. Zimbra uses only one local user (called zimbra) - now every mail goes first trough amavis new here we add all filters, blocklist whatever, each adds some score to decide its final destination. so dspam hangs somewhere in there.
the spam training is by user itself by moving their mails in and out the spamfolder. every time a user moves a mial in or out the spamfolder a caopy is sent to the spam/ham accounts for nightly training. the messages are within an attachment and get feed to all spam scanners installed. Well so far so good but something feels not really right here. The Original Zimbra Amavis configuration is this: %%uncomment LOCAL:amavis_dspam_enabled%% ['DSPAM', 'Amavis::SpamControl::ExtProg', $dspam, %%uncomment LOCAL:amavis_dspam_enabled%% [ qw(--stdout --classify --deliver=innocent,spam --mode=tum %%uncomment LOCAL:amavis_dspam_enabled%% --tokenizer=chained,noise --user), $daemon_user ], %%uncomment LOCAL:amavis_dspam_enabled%% mail_body_size_limit => 64000, score_factor => 1, %%uncomment LOCAL:amavis_dspam_enabled%% ], ); which cant be right, obivisously. since theres no option tokenizer :-))) also why in hell we want to give here another option than in the config? using tum by default but maybe toe in the config wont do any good - thankfully those options are in the amavisd.conf and not documented at all by zimbra they dont even tell ya to look there if you want enable dspam lol another question comes up with classify instead of process if classify is set we dont get dspam signature, so that cant be right either can it ? so the lines should be more like this, should they?: %%uncomment LOCAL:amavis_dspam_enabled%% ['DSPAM', 'Amavis::SpamControl::ExtProg', $dspam, %%uncomment LOCAL:amavis_dspam_enabled%% [ qw(--stdout --process --deliver=innocent,spam --source=error --user), $daemon_user ], %%uncomment LOCAL:amavis_dspam_enabled%% mail_body_size_limit => 64000, score_factor => 1, ], ); any ideas? as a side note also maybe using parsetoheaders if we want individual user based training which seems to be the only option todo that with zimbra. of course this would be make a giant amount of data stored - if parseto would be an option we maybe would need a 3rd option here right now we can only use xxx@ or xxx@xxx while we would need @xxx - so domains would work kinda groups and reduce the amount of data needed by dspam ------------------------------------------------------------------------------ This SF.net email is sponsored by Windows: Build for Windows Store. http://p.sf.net/sfu/windows-dev2dev _______________________________________________ Dspam-user mailing list Dspam-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/dspam-user
