Justin Mason wrote:

Rudy Rucker said:

I saw something about auto_learn...

(It will be used to auto-train the Bayesian learner from mails that are
"definitely spam" or "definitely nonspam", but isn't fully implemented
yet.)

As long as this feature is 'in the works' , let me put in my 2 cents: what will the learner mark as 'definitely spam'?

This also leads me back to my original questions, will the all_spam_to and whitelist_from scores affect the auto-learner. Furthermore, will the RBL stuff be added to the learner?
The RBL and LARGE scores (eg +-100) should be ignored. They do not demark spam, rather an RBL or whitelist item. RBLs may be from open relays, and once the RBL (eg osirusoft.com) database is updated due to a mislisting or closure of an open relay, SpamAssassin should no longer flag emails from that relay.

So, back to my original question, in Version 2.43, is there anyway to limit how the auto-whitelist learns? Well, doesn't sound like it. Here is a Version 2.43 patch:

# diff -u AutoWhitelist.pm.orig AutoWhitelist.pm
--- AutoWhitelist.pm.orig Fri Nov 29 16:31:49 2002
+++ AutoWhitelist.pm Fri Nov 29 16:41:07 2002
@@ -23,6 +23,7 @@
};

$self->{factor} = $main->{conf}->{auto_whitelist_factor};
+ $self->{scorelimit} = $main->{conf}->{auto_whitelist_scorelimit};

if (!defined $self->{main}->{pers_addr_list_factory}) {
$self->{checker} = undef;
@@ -90,6 +91,10 @@

sub add_score {
my ($self,$score) = @_;
+
+ if (defined $self->{scorelimit}) {
+ ($score < $self->{scorelimit} and $score > -$self->{scorelimit}) or return undef;
+ }

if (!defined $self->{checker}) {
return undef; # no factory defined; we can't check

# diff -u Conf.pm.orig Conf.pm
--- Conf.pm.orig Fri Nov 29 16:41:39 2002
+++ Conf.pm Fri Nov 29 16:52:19 2002
@@ -125,6 +125,7 @@
$self->{auto_whitelist_path} = "~/.spamassassin/auto-whitelist";
$self->{auto_whitelist_file_mode} = '0600'; # as string, with --x bits
$self->{auto_whitelist_factor} = 0.5;
+ $self->{auto_whitelist_scorelimit} = 40;

$self->{rewrite_subject} = 1;
$self->{detailed_phrase_score} = 0;
@@ -836,6 +837,20 @@
=cut
if (/^auto[-_]whitelist[-_]factor\s+(.*)$/) {
$self->{auto_whitelist_factor} = $1; next;
+ }
+
+=item auto_whitelist_scorelimit n (default: 40, range > 0)
+
+whitelists and all_spam_to scores don't add accurate information to your auto whitelist
+database. If C<scorelimit> > score > - C<scorelimit>, then the email is added to the
+running auto whitelist mean, otherwise, it is skipped. A value of 0 deactivates the limit
+check.
+
+=cut
+ if (/^auto[-_]whitelist[-_]scorelimit\s+(.*)$/) {
+ $self->{auto_whitelist_scorelimit} = $1;
+ $1 > 0 or undef $self->{auto_whitelist_scorelimit}; + next;
}

=item describe SYMBOLIC_TEST_NAME description ...




-------------------------------------------------------
This SF.net email is sponsored by: Get the new Palm Tungsten T handheld. Power & Color in a compact size! http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0002en
_______________________________________________
Spamassassin-talk mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/spamassassin-talk

Reply via email to