You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@spamassassin.apache.org by bu...@bugzilla.spamassassin.org on 2006/10/12 13:21:31 UTC
[Bug 4628] remove some RFC-I rules
http://issues.apache.org/SpamAssassin/show_bug.cgi?id=4628
jm@jmason.org changed:
What |Removed |Added
----------------------------------------------------------------------------
Severity|normal |major
Summary|rfc-ignorant.org blacklists |remove some RFC-I rules
|are overzealous |
Target Milestone|Undefined |3.2.0
------- Additional Comments From jm@jmason.org 2006-10-12 04:21 -------
some up-to-date data from the ruleqa system --
http://ruleqa.spamassassin.org/?daterev=20061007-r453869-n&s_defcorpus=on&rule=%2FRFC&srcpath=&s_zero=on&s_detail=+&g=Change
:
0.00000 3.7247 0.0540 0.986 0.85 2.60 DNS_FROM_RFC_DSN
0.00000 2.2447 0.1700 0.930 0.73 1.94 DNS_FROM_RFC_BOGUSMX
0.00000 15.1533 4.6068 0.767 0.51 1.45 DNS_FROM_RFC_POST
0.00000 18.6219 8.6003 0.684 0.49 1.71 DNS_FROM_RFC_ABUSE
0.00000 6.4258 4.0476 0.614 0.48 0.20 DNS_FROM_RFC_WHOIS
measured on a 260k spam, 63k ham corpus with 7 contributors.
I think we should keep DNS_FROM_RFC_DSN and DNS_FROM_RFC_BOGUSMX (98.6% and 93%
accurate respectively), but drop the other 3 rules.
They have a very high ham hit-rate, but are still assigned relatively high
scores by the perceptron; I'd prefer if the perceptron didn't have the option of
being misled by them at all... as this bug and mailing traffic attests, it's a
PITA to support.
2. an alternative would be a way to hint to the perceptron that these are
untrustworthy rules, "tflags weak" or whatever. but that would require
additional code and I think we're better off dropping the rules.
3. actually, another alternative: make them into meta subrules, so that future
rules can use them in metas. that may be useful. They're "free" anyway, since
DNS_FROM_RFC_* are all from one DNS lookup.
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.