[Spambayes] Re: Newbie experiences (Emacs + VM) and questions.

Prabhu Ramachandran prabhu at aero.iitm.ernet.in
Mon Jun 2 00:31:29 EDT 2003


Hi Neale,

Strangely, I did not receive your reply to my email but I did receive
Alex's reply to your email!  Anyway here is a reply using your message
from the archives.  Sorry about the messed up threading of the
messages, I don't have all the message ids. :(

>>>>> "NP" == Neale Pickett <neale at woozle.org> writes:

    NP> Hi Prabhu.  This is excellent!  I've added it to spambayes.el

Thank you!  I'm glad this was of use.

    NP> (included at the end of this message).  Please try this out
    NP> and let me know if it works for you.  If it does, I'll check
    NP> it in to CVS.

I don't use the spambayes.el directly but noticed some errors in the
version you posted.  I've attached a corrected version with some
documentation fixes.  If you think it is OK please go ahead and check
it in.

    >> 2. Is it a good idea to train on every incoming message or are
    >> there better approaches?

    NP> I did this (train on everything) for a while, but I've found
    NP> that mistake- and unsure-based training has been working very
    NP> well for me.  IIRC, this is even how Tim Peters does it.  So
    NP> you don't train on incoming messages, you only train on the
    NP> ones that spambayes got wrong.

Ah, OK, thanks for the hint!  I've also switched to using unsure and
mistake based training and so far its working admirably well.  What a
relief from disgusting spam!

    >> 3. At the moment I receive more HAM than SPAM.  Does this
    >> matter or make no difference?

    NP> It doesn't seem to matter for me.

OK.

    >> 4. I see that some folks retrain the whole database every so
    >> often.  Is this necessary?  If so why?

    NP> I think people do this as a way of pruning their database.

Thanks Neale and Alex for the information!

    >> 5. It would be nice if hammie.py would retrain on any messages
    >> that had an X-Spambayes-Trained header i.e. if
    >> Hammie.train(...) would check for the header and if found
    >> retrain.  If this is intentionally not done it needs to be
    >> documented somewhere.  Its a little confusing (for a newbie) if
    >> there are different scripts with similar options. :)

    NP> I know :( I may be able to get back in and fix this soon, as
    NP> $FIRM may let me start working on the project again soon.  But
    NP> in the meantime, feel free to hurl invectives my way about the
    NP> myriad confusing options to the myriad similar command-line
    NP> utilities :)

Well, I'll certainly send you no invectives but maybe a few humble
suggestions. :) For now I think just having one hammiefilter.py or
hammie.py should do the job.  If that is not desireable perhaps a few
lines in the usage string (or even in the documentation README.txt or
INTEGRATION.txt) specifying which is to be used under what
circumstance would be nice.

The INTEGRATION.txt file also mentions VM (does not mention Gnus
though) so perhaps it can point to spambayes.el and the contrib
directory.  

INTEGRATION.txt should really have sections on integration with Gnus,
VM, mutt and other MUA's.  If sections for each MUA is difficult
perhaps pointers to where info on integration for them is available
can be added.  I think this would make Spambayes a whole lot easier
for non-Outlook newbies to get going with.

Thanks!

cheers,
prabhu

-------------- next part --------------
A non-text attachment was scrubbed...
Name: spambayes.el
Type: application/octet-stream
Size: 5386 bytes
Desc: not available
Url : http://mail.python.org/pipermail/spambayes/attachments/20030601/23f954c8/spambayes.obj


More information about the Spambayes mailing list