On Sun, Oct 06, 2002 at 10:59:08AM +0200, martin f krafft wrote: > also sprach Tom Marshall <[EMAIL PROTECTED]> [2002.10.06.0608 +0200]: > > I've written a Bayesian e-mail filter, > > http://sourceforge.net/projects/bmf/ From the project page: > > > > bmf is a self contained and extremely efficient Bayesian mail > > filter. See Paul Graham's article "A Plan for Spam" for background > > information. It aims to be faster, smaller, and more versatile > > than similar applications. > > bogofilter already implements this. > > > I would like to become a Debian developer to maintain my own package > > (and a few others) and find a sponsor. Toward that end, I am also > > in the process of contacting a current Debian developer to sign my > > GPG key but he has not yet responded. If anyone reading this > > message is in or near Seattle, WA and would like to sign my key, > > please let me know. > > I'll sponsor it if you can tell me why it's better than bogofilter. > Where are the Debian sources?
- It is small. bogofilter is over 30 times its size. spamprobe is over 7
times its size.
- It is versatile. It supports text files (compatible with bogofilter 0.6),
libdb v1 to v4 (compatible with bogofilter 0.7.x), and mysql. bogofilter
and spamprobe support only recent versions of libdb.
- It is efficient. The incoming text is not copied to form a data
structure. Sorted vectors are used to store data. The number of calls to
memory allocation functions is orders of magnitude less than bogofilter.
- It supports scoring and updating the word lists in a single invocation.
This was finally added less than two days ago in bogofilter cvs.
- It includes a utility to convert between supported formats. bogofilter
and spamprobe have no facility to import/export lists.
- The parser is handcrafted and is easily made to recognize the unique
format of an email message (multiline headers, case insensitivity in
header names, etc.) This would be difficult in a lex grammar.
- It does not rely on external data structure libraries. The current
release of bogofilter uses libJudy, which is a pain to download and
compile. Yes, I realize that CVS and the current .deb have removed
libJudy.
- It is highly portable. It's written in C and compiles cleanly with no
compiler warnings on several architectures.
- Its author promises not to break backward compatibility without providing
a clean upgrade path.
My initial shot at a Debian package is available on sourceforge.net along
with the tarball and rpm package. I freely admit that the current version
is hacked up from looking at existing packages. I'm here because I want to
learn the Right Way to do it.
--
Majority, n.:
That quality that distinguishes a crime from a law.
pgp1dB4FnmeIY.pgp
Description: PGP signature
