Re: [cvs] bogofilter/src globals.c,1.67,1.68
Fast Bayesian spam filter along lines suggested by Paul Graham
Brought to you by:
m-a
From: Matthias A. <mat...@gm...> - 2005-03-01 02:23:35
|
Matthias Andree <mat...@gm...> writes: > We might make a new poll, along the lines of "do you want fast & more > complex to operate (BDB) or slower & maintenance free (SQLite3), at the > same level of safety". It's a bit too early to ask that question just > now, as SQLite3 may need some tuning for performance still, I'm a bit > unhappy with the lock contention, t.lock{1,3} take very long with > SQLite3 and we might offer a fast mode that locks out multiple writers > at the same time, or optionally reverts to the classic > readers-only-or-one-writer exclusive locking we had before starting to > hack BDB. Given the recent change to Berkeley DB that addresses the "log files clog up my hard disk" problem by removing inactive log files automatically, I hope we got rid of 2/3rds of the users' woes about disk space use and maintenance. I recall one problem report in the past four weeks that should be resolvable with bf_resize though. This is my .bogofilter directory, annotated, I haven't run bogoutil to sweep out here in a long time; I have several trainings of the database daily but no bogofilter -u. Evidently, the wordlist.db size dominates over everything else: $ LANG=en_US ls -sh1 --si total 18M 4.1k DB_CONFIG 13k __db.001 <- environment 5.3M __db.002 <- cache/memory pool 46k __db.003 <- log administration 975k __db.004 <- lock tables 13k __db.005 <- transaction region 0 lockfile-d <- global lock (-d for "directory" or "database") 4.1k lockfile-p <- crash detector table (-p for "processes") 41k log.0000000218 12M wordlist.db $ cat DB_CONFIG set_lg_max 262144 <- default 1048576 in bogofilter (manual entry) set_lk_max_locks 3874 <- set by bf_resize (semi-automatic) set_lk_max_objects 3874 <- set by bf_resize (semi-automatic) -- Matthias Andree |