Hello all,

I've setup DSPAM 3.8 as a replacement for our old amavis/SA system, approx
5.000 users, but the more I read about settings and training, the more I get
confused.

What I've done is:
- putting it behind amavis/SA on an intermediate server, so it can train
with SA's false positives
- backend and preferences are stocked in a MySQL database (separate server)
- training mode is TOE (well TEFT until I have 2500 ham/spam)
- I chose Inoculation group (with a virtual global user) so that users can
network their results
- no corpus fed for the moment, as I want users to retrain their mail from
the Web UI


I've tried MyISAM but I got performance issues, so now I'm trying InnoDB,
some readings said it may not be appropriate: some people suggest it may
enhance performance due to locking improvements, some others said they went
back to MyISAM

So now, I wonder:
- did I chose the right strategy (Inoculation) to give all users a starting
medium of filtering, or should I use a global group?
- am I right to only "TOE" via Web UI?

thanks in advance for any pointers

-- 
Alexandre Boulanger <[EMAIL PROTECTED]>


!DSPAM:1011,4916d128150921856817617!

Reply via email to