"qsf", a light-weight alternative to bogofilter?
C. Fischer
ino-qc at spotteswoode.dnsalias.org
Sat Feb 26 10:51:04 CET 2005
<URL:http://www.ivarch.com/programs/qsf.shtml>
(disclaimer: i'm not the author of "qsf"!)
"qsf" has some interesting features:
- it can be built using third-party SQL databases, but also using only libc's
binary tree routines.
- MIME-decodes all text/* parts with HTML tags removed, others are
md5-checksummed and the checksums entered into the token database,
- all tokens are stored as md5 values, and a platform independant dump format
exists. thus users can share and merge databases without fear of disclosing
private information(!),
- Users without write access to the system-wide database get their private
databases in their $HOME. read-only use of two system databases is
possible.
- rudimentary whitelisting of sender emails is already integrated. "qsf" will
never mark emails from certain senders as spam.
- filters and pairs
- builtin filters tag special tokens such as images or executable attachments,
whose appearance will eventually grade messages as spam,
- not only single tokens but also token pairs (i even saw groups of up to four
words) become database entries. this is a simple step towards phrase
recognition.
i have done but few experiments with "qsf", but plan to train it once i found
a not-too-large, comprehensive spam-corpus fur further testing. since "qsf"
documentation never mentions bayes, it would be interesting to hear other
people opinions about this similiar/dissimiliar product!
clemens
_______________________________________________
Bogofilter mailing list
Bogofilter at bogofilter.org
http://www.bogofilter.org/mailman/listinfo/bogofilter
More information about the Bogofilter
mailing list