Re: [Evolution-hackers] Bogofilter junk plugin patch

[Mikhail Zabaluev <mhz altlinux org>]

В Втр, 20/12/2005 в 10:57 -0500, Daniel Gryniewicz пишет:
> On Tue, 2005-12-20 at 15:29 +0300, Mikhail Zabaluev wrote:
> > I've massaged my Bogofilter junk plugin into a patch for inclusion in
> > the Evolution source tree, attached below.
> > The plugin is included in the experimental list.
> 
> Hi.
> 
> I have a very similar patch I've been using for a while.

Nice to hear from you. I dropped a few lines into your blog when I
discovered that we did parallel work :)

>   I would
> *highly* suggest using -u for the scan case, and -Ns and -Sn for the
> learn cases.  It makes a big difference in the learning ability.

I deemed this usage somewhat dangerous.
-u may silently reinforce false positives for an unsuspecting user.
It also writes every message's tokens into the database, so it is slower
and generates bloat.
I found no problems using bogofilter with the errors-only learning style
starting from a nearly empty wordlist, after correcting a few false
positives during the first week or so (to a total of 9 hams so far). My
database is mostly used read-only now.