[Rspamd-Users] rewrite subject and bayes learning process

[denny.friebe at icera-network.de]

I am also interested in it.
I need to add another domain to my mailserver soon. The mailboxes of this
domain are fetched by a groupware which unfortunately only speaks pop3. So I
am forced to dig out pop3 again soon. For pop3 I will simply merge all
folders of the mailboxes via virtual namespace so that all folders like
inbox and spam can be fetched. So that the groupware moves the spam mails
already filtered by rspamd also into their own spam folder I must
unfortunately fall back to the subject rewrite function since the groupware
does not support distribution on the basis of header flags.

 

If rspamd now flags such an email via rewrite, how does the learning behave?
Currently I trigger an automatic relearn when an imap user moves a mail from
the spam folder. I would like to keep it that way. What happens if a mail
that was previously marked as ***spam*** by rspamd is to be learned as HAM.
Does this affect the learning of rspamd or is this flag ignored in the
subject?