I’m using spambayes for a long time now. But when i decide to install it for all the curent users of my setup (w/ some virtual domains..), i just discover that spambayes don’t have a system-wide deamon like spamassasin (spamd).
So the first try:
- install spamassassin :) .. This mail filtering is just a bulshit ! Even w/ the training done on a mail, it achieve to deliver it as ‘unsure spam’ !!
- put spamassin away .. but keep piece :)
I first decided to write another client / server for spambayes. but looking at all the stuff writen in spamc (spamassassin client) i discover i will need a lot of nights (i’m not a C guru . even if the little try i wrote works perferctly )
Nice try, but why i shouldn’t simply write a server that use spamc as client ? .. could be really easy and efficient too (spamc is really efficient) ..
i just finish to write this and submit to the dev list
I’m using the “sb_filter.py” command instead of starting a pop3 proxy. That way I can implement a white list before spambayes has a crack at it, and my database is unique for each user.
My .spambayesrc:
[Storage]
persistent_use_database = False
persistent_storage_file = ~/.spambayes_database.pck
Stephen
Please run the thread until the end. You will discover severall things:
- sb_filter.py have some drawback, and should be replaced w/ sb_bn[filter|server].
- my piece of code can be hacked to support ~/.spambayesrc parsing too. This is just a mather of implementation
Wow, this is really cool, I was searching for something like this.
Can you please explain your setup a little bit more? How do you do the training? Which parameters do you use when calling spamc? I’m currently running spamc like this:
spamc -u $USER@$HOST -U /tmp/spamd.sock
thanks
I run spamc like that :
- spamc -u $USER -U /tmp/spamd.sock
- or spamc -u $USER -d ip -p port
(default ip = localhost -p = 1783)
The training is done throught a cron. I put a HAM in each Maildir folder, wrong classified mail should be put by users in.