Advice on Using sa-learn

[Errol Neal]

Hi folks,

A week or two ago, I posted a question about properly getting messages into
sa-learn.
I didn't get any answers and I was hoping for some advice from somebody on
this topic.
I've searched the archives, but didn't come up with a clear cut way to
proceed forward on
this. Could someone give me some suggestions based upon what they are doing
and their
results?

I am trying to create a standard method for our clients to report SPAM to
us that makes it thorough
the scanners for some reason. This would involve clients somehow
redirecting the sample to
a special e-mail account. The e-mail account would be reviewed daily to
make verify each sample.
A cron job on our mailscanner would execute a perl script that would log
into the pop3 server and
download the messages into a mbox file and feed those messages into
sa-learn. My questions
revolve around  the best ways to get those samples to that special e-mail
account. I understand that for
obvious reasons, forwarding messages do not work; they mangle the message
headers. However, what
about re-directing the message? It modifies the messages headers somewhat.
Also, these messages,
since they have been scanned already by mailscanner have additional headers
added to them. Should
I configure mailscanner to ignore those message headers when it's using the
bayes database?
I saw on the list some speaking about having users send the original
message as an attachment.
I don't know of any MuA that does this (at least the most popular windows
ones). The users would have
to do something more manual like copying and pasting the message source
including full headers into
a text file and attaching it to an e-mail. Is that pretty much what
everyone is doing?


Best Regards,

Errol Neal