From: Karl Auer [mailto:[EMAIL PROTECTED] > > For some time now, I have been busily accumulating bayes data by running > sa-learn on various collections of emails. As myself, so I now have a > nice big chunk o'data in ~/.spamassassin. > > Since I am a newbie to SA, I didn't realise what was happening for some > time. I actually wanted that data to be used globally, for all mails > that spamassassin checks.
Since you are using per-user databases, there is no easy way to make that corpus available to every other SA user in your system. You may use sa-learn --backup and --restore facilities to copy all that knowledge to someone else's account, but this would wipe the previous contents of the destinating bayes db, which may be less than optimal. Due to how the bayes db is designed, you can't even "stack-up" the informations it contains in a multi-layered way, like, in example, by having a server-wide db and a per-user db which are inspected and updated in parallel: there is actually no way to merge data coming from multiple dbs as well as there is no way to update it. If you believe that each user gets more or less the same kind of e-mails (like, in example, when running a small-business MX), then you may think to switch to a per-system bayes db an preload that single db with the content of your own bayes. giampaolo > Is there some simple way to do this? I no longer have the email that I > used to train spamassassin, just ~/.spamassassin/bayes_seen and > ~/.spamassassin/bayes_toks. > > Regards, K. > > -- > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ > Karl Auer ([EMAIL PROTECTED]) +61-2-64957160 (h) > http://www.biplane.com.au/~kauer/ +61-428-957160 (mob) >
