Re: Removing test for FROM/MX_MATCHES_NOT_HELO(DOMAIN)

2007-11-07 Thread Sahil Tandon

On Nov 7, 2007, at 7:25 AM, Robert Felber wrote:


when added to the 1 score for
$from_match_regex_verified_helo[0],


Which you have set to 0


I should've clarified, this was done after the instance of mail being  
discussed, so the change does not apply.


I don't think FROM/MX_MATCHES_NOT_HELO(DOMAIN)=1 was due to other  
scores, I

explained it rather to not raise wrong assumptions when setting
$from_match_regex_verified_helo[0] to 0 (also I haven't seen a log- 
excerpt,
which caused me to make that statement, I can neither say whether it  
is

impossible that FROM/MX_MATCHES_NOT_HELO(DOMAIN) can be "1" if
$from_match_regex_verified_helo[0] is set to 0, but I'd guess so).


No you are right it cannot be 1, because once the config changes were  
picked up after a full policyd-weight reload, that score is gone -- it  
was only after refreshing the config (but not reloading policyd- 
weight) that the score was still =1.


Thanks for your your help.


Policyd-weight Mailinglist - http://www.policyd-weight.org/


Re: Removing test for FROM/MX_MATCHES_NOT_HELO(DOMAIN)

2007-11-07 Thread Robert Felber
On Wed, Nov 07, 2007 at 07:08:44AM -0500, Sahil Tandon wrote:
> On Nov 7, 2007, at 3:16 AM, Robert Felber wrote:
> 
> >FROM/MX_MATCHES_NOT_HELO(DOMAIN) for instance is calculated such:
> >
> >$from_match_regex_verified_helo[0]  +
> >($total_dnsbl_score/4)  +
> >($bogus_mx_penalty * $bogus_mx_penalty) +
> >$glob_numeric_score
> 
> In the same email that had FROM/MX_MATCHES_NOT_HELO(DOMAIN)=1, there were 
> three negative DNSBL scores that 
> summed to -4.5.  That sum div by 4 = -1.125 which,

No, total_dnsbl_score is the score for dnsbls that have hit.

> when added to the 1 score for 
> $from_match_regex_verified_helo[0],

Which you have set to 0

> should've made the aggregate FROM/MX_MATCHES_NOT_HELO(DOMAIN)=-.125, no?  
> There was no positive bogus_mx_score/penalty on the email. 

> The only variable I do not understand is 
> $glob_numeric_score, iwhich may bring the aggregate back to 1.

Sorry, was from the devel version, which is 
$helo_numeric_score[0] * $total_dnsbl_score 

I don't think FROM/MX_MATCHES_NOT_HELO(DOMAIN)=1 was due to other scores, I
explained it rather to not raise wrong assumptions when setting 
$from_match_regex_verified_helo[0] to 0 (also I haven't seen a log-excerpt,
which caused me to make that statement, I can neither say whether it is
impossible that FROM/MX_MATCHES_NOT_HELO(DOMAIN) can be "1" if
$from_match_regex_verified_helo[0] is set to 0, but I'd guess so).


> >However: config changes are picked up by childs after MAINTENANCE_LEVEL
> >requests while each child has an own maintenance counter.
> >To speed up a complete reloading of all childs use policyd-weight reload.
> 
> That explains it.  Thanks for the clarification.
> 
> Sahil
> 
> 
> Policyd-weight Mailinglist - http://www.policyd-weight.org/

-- 
Robert Felber (PGP: 896CF30B)
Munich, Germany


Policyd-weight Mailinglist - http://www.policyd-weight.org/


Re: Removing test for FROM/MX_MATCHES_NOT_HELO(DOMAIN)

2007-11-07 Thread Sahil Tandon

On Nov 7, 2007, at 3:16 AM, Robert Felber wrote:


FROM/MX_MATCHES_NOT_HELO(DOMAIN) for instance is calculated such:

$from_match_regex_verified_helo[0]  +
($total_dnsbl_score/4)  +
($bogus_mx_penalty * $bogus_mx_penalty) +
$glob_numeric_score


In the same email that had FROM/MX_MATCHES_NOT_HELO(DOMAIN)=1, there  
were three negative DNSBL scores that summed to -4.5.  That sum div by  
4 = -1.125 which, when added to the 1 score for  
$from_match_regex_verified_helo[0], should've made the aggregate FROM/ 
MX_MATCHES_NOT_HELO(DOMAIN)=-.125, no?  There was no positive  
bogus_mx_score/penalty on the email.  The only variable I do not  
understand is $glob_numeric_score, which may bring the aggregate back  
to 1.


However: config changes are picked up by childs after  
MAINTENANCE_LEVEL

requests while each child has an own maintenance counter.
To speed up a complete reloading of all childs use policyd-weight  
reload.


That explains it.  Thanks for the clarification.

Sahil


Policyd-weight Mailinglist - http://www.policyd-weight.org/


Re: Removing test for FROM/MX_MATCHES_NOT_HELO(DOMAIN)

2007-11-07 Thread Robert Felber
On Tue, Nov 06, 2007 at 09:29:26PM -0500, Sahil Tandon wrote:
> This score penalizes legitimate email from people who send "From:" personal 
> domain names using their ISP's 
> SMTP server (which, generally, is not the authorized MX for 
> some-domain-name.org). 

A penalizing is "accepted" as long as this false penalizing does not lead to 
a reject. This is what policyd-weight is for - that checks may hit - but do no
harm.

> In order to disable this 
> test, based on 
> http://www.mail-archive.com/[email protected]/msg00525.html, I 
> added  
> @from_match_regex_verified_helo = (0,   -2   ); into 
> policyd-weight.conf.  /var/log/maillog 
> confirms that the new conf is loaded.  Nevertheless, new email is still being 
> tagged with a 
> FROM/MX_MATCHES_NOT_HELO(DOMAIN)=1.
> 
> What am I doing wrong?
> 

Probably nothing. Policyd-weight does score certain checks by using results
of other checks.

FROM/MX_MATCHES_NOT_HELO(DOMAIN) for instance is calculated such:

$from_match_regex_verified_helo[0]  +
($total_dnsbl_score/4)  +
($bogus_mx_penalty * $bogus_mx_penalty) +
$glob_numeric_score


There is no possibility to "disable" a check.


However: config changes are picked up by childs after MAINTENANCE_LEVEL
requests while each child has an own maintenance counter. 
To speed up a complete reloading of all childs use policyd-weight reload.


-- 
Robert Felber (PGP: 896CF30B)
Munich, Germany


Policyd-weight Mailinglist - http://www.policyd-weight.org/