Update: as of Oct 2007, this advice is obsolete. The Spamcop algorithms have been greatly improved, as far as I and others can tell.
I’ve been hearing increasing reports of false positives using bl.spamcop.net.
One today spurred me to check out exactly how many times it I’m seeing it misfiring on nonspam in my own mail collection. The results have been pretty astonishing.
In my nonspam collection, it fired on 1043 messages out of 8415 in July; 12.4% of the mail. It gets worse for August, though — 884 messages out of 3729 since the start of August. That’s a staggering 23% of my nonspam mail this month. ;)
Most of that is due to the listings of GMail and Yahoo! Groups, both of which seem to have been listed for large swathes of the past month and a half.
Now, an important point — it can work pretty well as a single input to a scoring system, like Spamcop itself or SpamAssassin. In fact, I didn’t lose any mail as a result of those listings; SpamAssassin assigns only 1.5 points to the RCVD_IN_BL_SPAMCOP_NET rule, so it’s easily corrected by other rules.
However, people using it to block or reject spam outright, or who’ve changed the score of the RCVD_IN_BL_SPAMCOP_NET rule, need to turn that off ASAP — as they are losing mail.
7 Comments