1 ctdlsalearn is a program that scans a specified set of "spam" and "ham"
2 folders (rooms), and batch-submits their contents to the "sa-learn" program
3 in order to train its Bayesian filters.
5 This is a work in progress. It isn't finished. You are only permitted to
6 use this program if you promise not to whine about it if it doesn't work.
8 I mean it! If you complain in the Citadel Support forum that ctdlsalearn
9 isn't working, I will hunt you down and jam ice picks into your eye sockets
10 and smash your kneecaps with a ball pean hammer. THIS IS UNSUPPORTED.
12 Still with us? Ok, here's how to make it work.
14 * Your Citadel server needs to be r7737 or newer. That's where we fixed
15 a bug that prevents internal programs from deleting messages. Without
16 this fix, messages won't get deleted from your spam and ham folders,
17 and they'll just pile up forever and get resubmitted to sa-learn on
18 every run. Don't let this happen to you.
20 * Back up your data. I may have hidden an "rm -fr /" in here somewhere.
22 * Open up src/main.c with your favorite text editor (which ought to
23 be vi, not emacs). Yes, you're editing the source code to configure
24 the program. Like I said, it's a work in progress. I haven't written
25 a configuration tool yet. If this is a problem for you, please either
26 write the configuration tool, or jump off a cliff; either one will
29 * Look for the references to "ctdldir". Is one of the paths we're trying
30 a valid location for your Citadel installation? If not, add yours.
32 * Now look for the calls to do_room(). This is where you can specify the
33 locations of ham and spam folders. The SpamAssassin folks tell us that
34 you need both in order for Bayesian filtering to work well. ctdlsalearn
35 will attempt to *delete* messages from these folders after submitting
36 them, so don't be a schmuck and point it to your inbox. Create ham and
37 spam folders for this purpose. Probably you want to copy ham to your
38 ham folder, and move spam to your spam folder (unless you have some sick
39 fetish for spam and you like to keep it around). Also note that if you
40 specify any rooms that are users' private mailboxes, you'll need to
41 prepend the namespace prefix (which is their user number, padded with
42 zeroes to ten digits, followed by a dot).
44 Yes, you can specify as many spam and ham folders as you want.
46 * Compile the program and run it. If it works, great; if not, re-read
47 the warning at the top of this document.