Mike Cardwell wrote:
Steve Bertrand wrote:
Hi everyone,
I aggregate my work and personal email accounts within the same email
client. All accounts are IMAP-based.
My $work employs a Barracuda cluster, and of course my box runs SA.
From time-to-time, I'll get a SPAM message come through the 'cuda's.
From there, I move the message from one IMAP folder in my MUA into
another SPAM folder, which essentially is a transfer from a work storage
server onto my server.
Every few days, I run sa-learn against the collected SPAM messages.
My question is, given that the messages have already been processed by
the 'cuda's (with their header stamps in place), am I damaging, or at
risk of confusing the learning process of SA when I classify these
messages as SPAM?
Are there any negative consequences by doing this?
You should configure bayes to ignore those headers. In your local.cf,
list each of the cuda headers like this:
bayes_ignore_header X-CudaHeader1
bayes_ignore_header X-CudaHeader2
bayes_ignore_header X-CudaHeader3
I have a similar setup. If a Spam message makes it to my inbox with less
than the required_score, I put it into a SPAM folder and run sa-learn on
the folder. Should I also implement the following ignore rules?
bayes_ignore_header X-Spam-Flag
bayes_ignore_header X-Spam-Level
bayes_ignore_header X-Spam-Status
bayes_ignore_header X-Spam...etc.
--
Dan Schaefer
