$Id: cssclean.txt,v 1.01 2011/07/11 23:46:05 sbajic Exp $

HASH_DRV NIGHTLY MAINTENANCE

The tool for nightly maintenance - dspam_clean - does not work with hash_drv,
it does not do any cleaning.

You have to clean it by yourself. There are the two steps to do:

- first, you should remove old signature files. These files are located
in user.sig directory and have extension .sig. They are needed only
for dspam retraining, so you can remove them if they are older than two weeks.
- second, you should purge databases. They are located in user.css
files and contain a set of tokens with counters: SPAM and NONSPAM, which count
how many times the token appeared in spams and innocent mails.

There is a special tool for cleaning it - cssclean:

cssclean [file.css] {heavy}

Cssclean implements its own counter for each token. It increments every
cleaning - so if you run cssclean every night, it works like timestamp.
If DSPAM uses a token, it resets this counter. So cssclean knows which tokens
were not used for counted time.

Cssclean removes tokens, which are:
- not used for 15 cleans and ( NONSPAM + SPAM <= 1 ), or
- not used for 15 cleans and NONSPAM is equal or almost equal SPAM, or
- not used for 60 cleans and ( NONSPAM*2 + SPAM < 5 ), or
- not used for 120 cleans.

With special options - heavy - cssclean is more strict and removes
tokens for which:
- NONSPAM + SPAM <= 1
- NONSPAM is equal or almost equal SPAM