diff options
| author | Eric Wong <e@80x24.org> | 2014-04-21 00:18:33 +0000 |
|---|---|---|
| committer | Eric Wong <e@80x24.org> | 2014-04-21 00:23:33 +0000 |
| commit | 62f57b7545eedf6a781efb2f5269b0a38ef2cd4c (patch) | |
| tree | b561952c8a3fa1ac8f033966e5971f3f89ec3751 /Documentation | |
| parent | 1c910ebd5d088b68fd1cec41285ad62641673845 (diff) | |
| download | public-inbox-62f57b7545eedf6a781efb2f5269b0a38ef2cd4c.tar.gz | |
The ~/.dc-dlvr.pre script for my public-inbox user does this.
Diffstat (limited to 'Documentation')
| -rw-r--r-- | Documentation/dc-dlvr-spam-flow.txt | 38 |
1 files changed, 38 insertions, 0 deletions
diff --git a/Documentation/dc-dlvr-spam-flow.txt b/Documentation/dc-dlvr-spam-flow.txt new file mode 100644 index 00000000..2cdcefa5 --- /dev/null +++ b/Documentation/dc-dlvr-spam-flow.txt @@ -0,0 +1,38 @@ +dc-dlvr spam/ham training system flow +------------------------------------- + +An overview of the Maildir + inotify-based spam training system Eric +uses on his mail server. This idea may be implemented for kqueue-based +systems, too. + +The idea is to use inotify (via incron) to watch for new files appearing +in Maildirs. We only want to train seen messages as ham, and old (but +not necessarily seen) messages as spam. The overall goal of this is to +allow a user to train their filters without leaving his favorite mail +user agent. + +Every message written to Maildir involves a rename, so we only +have incron watch for IN_MOVED_TO events. + +The generic flow is as follows, all for a single Unix user account: + + incron -> report-spam +-> sendmail -> MTA -> dc-dlvr -> spamc -> spamd + | + V + ... + +For public-inbox, Eric uses a separate Unix account ("pi") to add a +layer of protection from fat-fingering something. So his report-spam +script delivers to a second recipient for training, the "pi" user: + ... + | + +-> sendmail -> MTA -> dc-dlvr + | + V + ~pi/.dc-dlvr.pre + | + V + public-inbox-learn + +public-inbox-learn will then internally handle the "spamc -> spamd" +delivery path as well as calling ssoma-rm on falsely trained |