RE: new spamprobe BerkeleyDB on CygWin

Brian Burton
> > (Therefore I started over and ran it under 'nice'
> > at reduced priority - ignore the REAL time):
> >
> > $ time nice spamprobe train-spam spamtext.txt
> > real    44m31.719s
> > user    18m18.733s
> > sys     3m23.781s
> >
> > [Remember to ignore the real time, but...] almost 21 
> minutes user+sys 
> > time for one message.
> >
> 
> Wow.  How many messages did you have in that file?  What kind 
> of system was it (CPU, speed, memory)?  Using a larger -s 
> option can sometimes help with large training runs.  I use -s 
> 5000 in my larger test runs but you can use whatever works 
> best on your system.  The higher the number the more memory 
> SP will use and the less frequently it will update the database.

Sorry, Brian I should have been more clear (thinking
it was obvious when it is not was my fault) by saying
this was a single fairly short message used MERELY
to create the files.

The system is a reasonably fast, not cutting edge, 2 Ghz
Celeron with 1 MB of RAM.

Something is wrong -- it is likely something that I have
done (or not done) and/or the combination of CygWin,
Win2003 server, and SpamProbe since you presumably don't
have this problem usually AND CygWin runs most "Linux/GNU"
like software flawlessly (Exim, SpamAssassin, CRM114 are
among my production components.)

> Using auto-train might give better accuracy for you.  It will 
> take longer to run though (it pre-parses all of the mailbox 
> files to count  
> how many messages there are of each type before it begins 
> training).   
> I really need to change the README to recommend using it 
> instead of train-good/train-spam for new users.

Biggest problem for me initially was figuring out how
to "create a database" (that READ only thing drove me
nuts until I reviewed the archives of this list.)

Please understand that I really appreciate your software
is freely available so this is NOT a 'complaint'.

But here is my thinking:  Given that I am a reasonably
technical developer in a general sense but not a Gnu/Linux
etc expert specicifically if I am having trouble you may
be losing "users" if there are problems not getting 
reported to you.

Sure CygWin may be a bit 'odd' but generally it works;
and it would have been very easy for me to just compile,
fail, and move on to other software without giving you
a report.

Just so you won't feel alone, I am also having problems
making DSpam due to autoconf and libraries issues, and
although I have CRM114 working it did take a bit of
asking for help and massaging the make file to get it 
to compile/link,

BTW, today I passed my first 24 hours without a single
SPAM with no FPs.  No spam even in the server 'spam catch' 
accounts. Nothing, nada, nichts....(30+ hours right now.).

When I started my Exim/SA/greylistd/CRM114/etc project 
just 3-4 months ago we received about 1200 spams per
day.

I was hoping to use your (highly recommended) software
to improve the ability to distinguish spam/ham....

--
Herb Martin




-------------------------------------------------------
This SF.Net email is sponsored by:
Power Architecture Resource Center: Free content, downloads, discussions,
and more. http://solutions.newsforge.com/ibmarch.tmpl