SpamSieve on multiple machines with IMAP - a workable approach
I have been running SpamSieve on three different boxes for several months with good results, and without using the “drone” approach (which incidentally looks interesting.)
Two of the machines are almost always running. One faces the intarweb and routes my household, and the other is a “communal” box in our living area. The third is my notebook, which is running only part of the time.
I got around the question of “how do you know which machine and filterset marked a given piece of mail as spam so you can train it” by telling each of the machines to put spam in a dedicated folder on the server.
When I look at my IMAP folders, I see something like this:
Spam - desktop box
Spam - communal box
Spam - notebook
For purposes of training SpamSieve that a given piece of spam has leaked through to my inbox, I don’t really care which machine gets the lesson. Someday when it is possible to merge the corpuses (corpii?) I’ll do it, but in the interim it doesn’t really matter.
What matters for me is training the specific machine which occasionally marks something good as spam.
When I’m scanning through the IMAP spam folders and find something which has been misclassified as spam I flag it and then tell SpamSieve that it is good next time I’m using the guilty machine so it doesn’t do it again.
While this approach lacks elegance, it gets the job done and serves to provide me with a usable inbox on those occasions when I don’t have access to any of my own machines and have to use the web interface to check my mail.
I suspect that something could be cobbled together using a variation of the “drone” approach to make this process cleaner.