...
A-C are the use cases catered for by the current spamtrap system. D and E are new, and would be useful in order to provide a common, shared "meeting place" of spamtrap data.
SMTP LISTENER
Each trap box runs an SMTP listener which delivers all incoming messages with size under a set size limit, to a \[http://www.danga.com/gearman/ Gearman\] queue. A set of Gearman worker processes then take messages from that queue, processing them as follows: Wiki Markup
- discarding obvious viruses/malware
- discarding too-large messages (over 4MB in size?)
- removing forwarding headers, for spams forwarded from third-party addresses
- extracting attached message/rfc822 parts (for certain forwarders)
- discarding messages from now-unreliable addresses, identified using header regexps
- discarding frequently-appearing ham, identified using header regexps
- discarding bounces (scan with SpamAssassin for ANY_BOUNCE_MESSAGES)
...