[Gllug] How do I do this?

[Nix]

On Mon, 17 Jan 2005, Martin A. Brooks suggested tentatively:
> sa-learn isn't the quickest running program ever, especially on large
> numbers of spam messages. If you run it often, it's probably worth
> adding a lock file check to avoid multiple runs on the same messages.

Er, it takes very little time for sa-learn to recognise that it's seen a
message again; it doesn't try to learn the same messages over and over.

What *is* dog slow is DB_File over NFS; it fsync()s up a storm. If you
make sure it's on a local drive, it's not that slow.

If Bayes is in an SQL database, it can be quite a lot faster (yes,
PostgreSQL and mysql both appear to be faster than Berkeley DB for this
application!)

-- 
`Blish is clearly in love with language. Unfortunately,
 language dislikes him intensely.' --- Russ Allbery
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug