header __HELO_NO_DOMAIN   X-Spam-Relays-External =~ /^[^\]]+ helo=[^\.]+ /

header FSL_HELO_NON_FQDN_1 X-Spam-Relays-External =~ /^[^\]]+
helo=[a-zA-Z0-9-_]+ /i
score FSL_HELO_NON_FQDN_1 2.361 0.001 1.783 0.001
# (note that no FSL_* rule has a description)

http://ruleqa.spamassassin.org/?rule=/HELO_NON_FQDN_1|__HELO_NO_DOM

These rules are virtually identical.  Sometimes one out-performs the
other by an insignificant margin.  I thought this was too high a ham%
(0.7795 for both) to get a decent score, yet it made it into 3.3.0
scored 2.361 on no-net/no-bayes.

To combat that, I refined the rule so as to discount anything that
identified as 'localhost' since ruleqa indicated that was a large
source of the FPs on the rules.  Looks like __HELO_NO_DOMAIN didn't
make it into 3.3.0 while FSL_HELO_NON_FQDN_1 did, so my derivative
HELO_NO_DOMAIN also failed to make the cut, and it wasn't alone:

  SPAM%     HAM%     S/O    RANK   SCORE  NAME
 3.5398   0.0056   0.998    0.91    0.00  DYN_RDNS_SHORT_HELO_HTML
15.0445   0.2587   0.983    0.73    1.00  HELO_NO_DOMAIN
 0.1007   0.0007   0.994    0.69    1.82  DYN_RDNS_SHORT_HELO_IMAGE
18.9905   0.7795   0.961    0.64   (n/a)  __HELO_NO_DOMAIN
18.9711   0.7795   0.961    0.63    2.36  FSL_HELO_NON_FQDN_1

The question:  Do we discard __HELO_NO_DOMAIN and rewrite the
dependent rules?  Is FSL_HELO_NON_FQDN_1 scored too high?  Again,
HELO_NO_DOMAIN was written to make a non-meta-rule out of that, and I
worry that the FSL version has side-stepped that.

Thoughts?

Reply via email to