Re: [vps-mail] sa-learn

[Jonathan Duncan <jonathan@xxxxxxxxxx>]

On Tue, 19 Apr 2005, Abigail Marshall wrote:

> --- Jonathan Duncan <jonathan@xxxxxxxxxx> wrote:
> > I ran "sa-learn --spam spamfile" on a file with
> > 1000+ messages in it.  It
> > took a minute or so and then said something like it
> > found 1 message and
> > learned 1 message.  Is this normal?
>
> How was that file compiled? Spamassassin will not
> re-learn email it has already learned automatically,
> so unless the spam in the email box is fresh, running
> sa-learn will not accomplish anything.  Also, if you
> had a situation where the Bayes databases became
> corrupted and you were trying to recreate a database,
> it would be important to delete all the old database
> files -- or else SA will use the old bayes_toks or
> bayes_journal file and treat the spam as "read" even
> though the bayes_seen file may no longer exist.
Ok, so I used the --mbox option that Jim mentioned and it learned all the 
messages in the file.  That is great, thanks!


> > Is there some
> > server-wide bayes database?  If so, where are the
> > files for it?  I did a
> > locate on bayes_tok and it found one for each of my
> > users, but that is it.
>
> That depends on your set up -- if you configured SA
> for server wide use, then there will be a server-wide
> database, usually in
> /usr/local/etc/mail/spamassassin/bayes on a VPS2 where
> SA was set up using vinstall.  The local.cf file could
> specify a different location.  On a VPS2, it would be
> in ~/etc/mail/spamassassin/bayes
>
> If SA is configured on a per-user basis, then you
> would not see a server-wide database.
Hmmm, there are no bayes files or directories in the 
/usr/local/etc/mail/spamassassin directory and each user has these files. 
So I guess if I want everyone to have access to these I need to reinstall 
or reconfigure or copy these to everyones home dirctories or make symbolic 
links from these to everyones home directories.  What is the best option?

I do not see my users adding much if anything to their own bayes files. 
Probably because they do not know about them and do not know how to use 
them.  I think CPX may incorporate bayes training into the mail client but 
that is not here yet.  So I figure that since I get most of the spam on my 
servers I might as well do the bayes training and share that with 
everyone.  Except, the intention of bayes is for everyone to be able to 
train it to their own needs.  What I think is spam may not be what someone 
else thinks is spam.  Highly unlikely, but possible.

Jonathan
======================================================================
This is <vps-mail@xxxxxxxxxxxx>       <http://www.perlcode.org/lists/>
Before posting a question, please search the archives (see above URL).