"jdow" <[EMAIL PROTECTED]> writes:
> Did this port the SURBL into the release or not?
Well, it's not listed in the announcement for a reason. :-)
Adding SURBL to 2.64 wasn't really an option we could consider. A few
reasons off of the top of my head:
- the 2.6x tree is our stable tree and that would be a major change
- the main purpose of 2.64 is fixing a specific security issue, adding
SURBL would have delayed the release and added unnecessary risk
- we don't have a old-style contributor agreement from Eric Kolve to
include the code
- 2.64 is going to be rather short-lived as the recommended stable
version, 3.0.0 is very close to release
- SURBL has a very large effect on scores, so it would have had a
negative impact on false positives
We did add a number of highly effective rules that have virtually no
false positives and fixed the most significant false positive problem
with 2.63 (the new SORBS and NJABL additions that cause the top-level
RCVD_IN_SORBS and RCVD_IN_NJABL rules to fire too often on ham).
Most of the rules are simple header rules, but we did add one new
network test which adds zero additional cost: XBL. I know it's not as
exciting as SURBL, but it works *very* well and we could add it with
minimal risk.
OVERALL% SPAM% HAM% S/O RANK SCORE NAME
151033 122586 28447 0.812 0.00 0.00 (all messages)
100.000 81.1650 18.8350 0.812 0.00 0.00 (all messages as %)
23.882 29.4218 0.0105 1.000 0.98 0.00 RCVD_IN_XBL
17.441 21.4837 0.0176 0.999 0.97 0.00 URIBL_SC_SURBL
53.206 65.5156 0.1582 0.998 0.90 0.00 URIBL_OB_SURBL
9.897 12.1449 0.2074 0.983 0.82 0.00 URIBL_AB_SURBL
58.068 71.4494 0.4043 0.994 0.79 0.00 URIBL_WS_SURBL
0.017 0.0204 0.0000 1.000 0.48 0.00 URIBL_PH_SURBL
Daniel
--
Daniel Quinlan
http://www.pathname.com/~quinlan/
