On Wed, May 12, 2010 at 02:19:06PM +0300, Take wrote: > On 05/12/2010 11:39 AM, Roman Medina-Heigl Hernandez wrote: > > I have a similar setup. What I do is using sieve. My script for mailbox > > > You have an extra advantage with this method: users may choose to edit > > his/her filter so he/she could exactly decide whether he/she want junk mail > > to be dropped, moved to a spam-folder, and the score threshold too. > > This is disadvantage in our case. We currently have spam filtering set > up like this, and it's of course working like it should. But now, since > our users aren't that technically advanced, it happens almost daily > basis that someone messes around with the filter rules via > managesieve-plugin on roundcube and destroys the spam-adjustment as well. > > Unfortunately those clueless users are the same people who spread their > address to newsletters, questionnaires etc, so they receive most spam as > well. > > This is why we'd need a way to do the rough clean before any user has > anything to say about that. If you give the user an possibility to shoot > him/herself into the foot they'll pull the trigger immediately.
I don't know if cyrus can do something similar, but I think you want something like sieve_before in Dovecot: http://wiki.dovecot.org/LDA/Sieve/Dovecot#Executing_Multiple_Scripts_Sequentially > So, thank you for your reply, but that's not the solution we're after. Dennis
