UNC Linux Users Group

[Cristobal M Palmer <cmpalmer AT email.unc.edu>]

Sounds like you're right there in the trenches fighting this one.

I'd like to know what you think of bogofilter. I use aptitude instead of
apt-get, and a quick '/' (search) for 'filter' got me 'annoyance-filter,'
a similar Bayesian filter program. Anybody used that one? Anybody know if
there's a real difference?

I also got a response recommending mutt, an email client that
"sucks less than the others," and I was wondering if anyone else would
mind sharing her/his ~/.muttrc file.

Apparently in mutt you can skip the whole shell script thing and set up a
macro and assign it a keybinding that will let you delete as spam/ham.

http://www.rohrbach.de/files/rc/.muttrc

Thanks (anon.) for the link to the .muttrc above; anybody else wanna pitch
one in?

Also: thanks to those who recommended http://mail.unc.edu/spam

I set that up--we'll see how it performs.

                                                               .--.
                                                              |o_o |
                                                              ||_/ |
Peace,                                                       //   \ \
                                                            (|     | )
  CMP                                                      /'\_   _/`\
                                                           \___)=(___/


On Wed, 18 Feb 2004, Alan Hoyle wrote:

> I get lots of spam.  Since the beginning of Feb, I've received almost
> 10000 spams.  Yes, that's ten thousand.  My setup has caught and deleted
> 9668 of them at last count (True positive).  It didn't catch roughly 260
> for a false negative rate of around 3%.  Since the beginning of the year,
> classified as spam roughly 5 non-spam messages (false positive).  It's
> hard to tell given the sheer volume of spams that I get. I don't know how
> many "true negatives" I've had (i.e. non-spams classified as non-spams).
>
> Pine's "Rules" setup doesn't really work well.  I tried that for a long
> while.
>
> Here's what I did on my linux desktop.  I don't know if it would work on
> isis.  I run the pine out of AFS package space.
>
> 1.  Downloaded and installed IMAPSpamBegone
>        http://www.rogerbinns.com/isbg/
>
> 2.  Installed SpamAssassin (with RedHat9)
>
> 3.  Noticed that SA was an absolute memory and processor pig on my
>     desktop machine.
>
> 4.  Downloaded and installed Bogofilter (bogofilter.sf.net)
>
>    4a.  trained bogofilter with collection of spam and non-spam messages
>
> 5.  Modified ISBG (it's open-source Python) so it works with bogofilter
>     (in addition to some other changes)
>
> 6.  Installed isbg into my crontab so it runs every 20 seconds or so
>     during the day and every 5 minutes or so at night.  Used the pine
>     pipe command ('|') to train on messages in my inbox.  (bogofilter -Ns
>     is the most common command I use)
>
> ... time passes ...
>
> 7.  Noticed that every once in a while, ISBG hangs on certain messages
>     (particularly some weird virus messages) or when the mailserver or
>     network goes down.  the subsequent build-up of isbg processes
>     eventually crashes my machine if unchecked.
>
> 8.  Added a "killall isbg" every 5 min to my crontab.
>
> ... time passes ...
>
> 9.  Upgraded bogofilter.
>
> 10.  Noticed how it broke my ISBG setup.
>
> 11.  wrote "deisbg" script to remove it from my crontab temporarily and
>      add it back when I'm done.
>
> ... time passes ...
>
> 12.  got tired of typing "bogofilter" in the | dialog box, created "bf"
>      shell script to run it.
>
> 13.  Noticed that the .bogofilter word list gets big fast.
>
> 14.  Automated the "compact DB" function to run weekly in my crontab.
>      (weekly run, reduces size from ~100 MB to ~65 MB, yes, it grows
>      almost 40 MB/week)
>
> If anyone's interested, I can post more details on my setup.
>
>
> On Wed, 18 Feb 2004 at 12:09, Cristobal M Palmer wrote:
>
> > I've gotten used to using the isis-based Pine program for my UNC-CH email
> > account, but recently I've started getting a lot more spam.
> >
> > UNC has server-side filtering, but I can't figure out if there's a
> > user .config that I could use to train Pine to recognize junk. I know
> > there's a 'rules' option menu, but that looks like a lot of work.
> >
> > If that'd just be too much of a hassle, can anyone recommend a
> > **lightweight** email client with Bayesian filtering?
> >
> > I want to hit a 'delete as junk' button and not have to see the same
> > message again.
> >
> > Sick of hearing about NEW!!1 weight-|oss pi1ls...
>
>
> --
>     Alan Hoyle  -  alanh AT unc.edu  -  http://www.alanhoyle.com/
>       "I don't want the world, I just want your half." -TMBG
>                  Get Horizontal, Play Ultimate.
>
>
>
>
>
>