Fatal error

[David Relson]

On Sat, 26 Feb 2005 22:23:50 +1000
Mark Constable wrote:

> Matthias Andree wrote:
> >>Sometimes it seems RTFM can be more dangerous than not.
> > How so?
> 
> Well I've done quite a bit of RTFM and now I've totally
> hosed my setup, it appears. I should have left well enough
> alone.
> 
> > Check the current README.db file I posted three weeks ago,
> 
> That may have been just before I rejoined this list. I
> have tried to find a searchable mailing-list archive.
> 
> > <http://article.gmane.org/gmane.mail.bogofilter.general/9265>, and
> > review it for recovery procedures. From the errors you posted, I'd guess
> > you need to remove the directory and then retrain from scratch.
> > 
> > Did your computer crash previously?
> 
> No. That's not likely with our co-lo.
> 
> > If so and it was a crash due to a
> > blackout, please switch off the write caches, also in the README.db
> > file.
> 
> Most suggestions involve turning off the mailserver which
> is close to impossible in my case... it's way too busy.
> I have a feeling I'm going to have to rebuild most of 4k
> databases and I have no corpus to start from so are there
> any recommended bogofilter friendly ham/spam collections
> out there ?
> 
> I have been googling but ham/spam/corpus and other keywords
> have not got me anything decent so far.
> 
> --markc

Hi Mark,

Sorry to hear about your database problems.  Unfortunately they
happen :-<

Bogofilter does a good job of discriminating between ham and spam with
very little training.  It helps to save some of your old spam in case
of problems like you've had.  However, if you've not done that, so a
different strategy is needed.

One simple method would be to turn off bogofilter for a day or two,
manually sort all incoming mail as ham or spam, train with that ham and
spam, then turn bogofilter on.

Another method would be to turn bogofilter on, have your MUA filter on
Unsure and Spam classifications, verify the classification of each
incoming message, and train on errors and unsures.  After a day or two
(perhaps 100 each of ham and spam), bogofilter should be doing a pretty
good job.  After a week or so, it should be doing very well.

The key to bogofilter's working well is the care with which training is
done.

If you _really_ want, I can send you some spam messages of the mortgage,
drug, porn variety.  However it's best to use your local mail feed.

HTH,

David


_______________________________________________
Bogofilter mailing list
Bogofilter at bogofilter.org
http://www.bogofilter.org/mailman/listinfo/bogofilter