You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@spamassassin.apache.org by Thomas Harold <th...@nybeta.com> on 2014/04/15 05:49:40 UTC
Re: RP_MATCHES_RCVD
On 11/8/2013 4:38 PM, John Hardin wrote:
> On Fri, 8 Nov 2013, Kris Deugau wrote:
>
>> LuKreme wrote:
>>> Some spam has been matching the rule RP_MATCHES_RCVD which is worth
>>> -2.8 points. I wanted to look at this rule, so I went to
>>> /usr/local/etc/mail/spamassassin and gripped for the name, but no hits.
>>
>> There was a thread on this rule not too long ago; check the list
>> archives
>
> Yeah, I thought we'd killed that in favor of a subrule. I guess we never
> actually pulled the trigger on that change... Mark?
>
It seems to be back, and the value is changing from week to week.
Feb 28th - RP_MATCHES_RCVD = -0.001
(during first few weeks of March it was showing as T_RP_MATCHES_RCVD, -0.01)
Mar 24th - RP_MATCHES_RCVD = -0.535
Mar 27th - RP_MATCHES_RCVD = -0.371
Apr 7th - RP_MATCHES_RCVD = -0.271
Apr 14th - RP_MATCHES_RCVD = -0.989
Running 3.3.1 on CentOS 6 (from the @updates channel). Running
"sa-update" daily.
Re: RP_MATCHES_RCVD
Posted by "Kevin A. McGrail" <KM...@PCCC.com>.
On 4/17/2014 10:21 AM, Matus UHLAR - fantomas wrote:
>>> it's not corrected, that's the point...
>
> On 17.04.14 09:14, Kevin A. McGrail wrote:
>> The scoring occurs from automatic corpus checks. The best way to
>> help the rule score better is to help with masscheck.
>
> and still SA people tune some scores manually.
>
>> Looking at
>> http://ruleqa.spamassassin.org/?daterev=20140416-r1587834-n&rule=RP_MATCHES_RCVD&srcpath=&g=Change
>> there does appear to be a hamminess to the rule and it justifies a
>> negative score. A score of -1.05 seems appropriate to me.
>
> Not to me. The whole fact that @gmail.com spam comming from gmail.com
> servers does not mean it's not spam, only because millions of
> @gmail.com ham
> comming from gmail.com are ham...
>
> this logic is braindead to me
Then you will likely have to use manual tuning.
regards,
KAM
Re: RP_MATCHES_RCVD
Posted by Matus UHLAR - fantomas <uh...@fantomas.sk>.
>>it's not corrected, that's the point...
On 17.04.14 09:14, Kevin A. McGrail wrote:
>The scoring occurs from automatic corpus checks. The best way to
>help the rule score better is to help with masscheck.
and still SA people tune some scores manually.
>Looking at http://ruleqa.spamassassin.org/?daterev=20140416-r1587834-n&rule=RP_MATCHES_RCVD&srcpath=&g=Change
>there does appear to be a hamminess to the rule and it justifies a
>negative score. A score of -1.05 seems appropriate to me.
Not to me. The whole fact that @gmail.com spam comming from gmail.com
servers does not mean it's not spam, only because millions of @gmail.com ham
comming from gmail.com are ham...
this logic is braindead to me
--
Matus UHLAR - fantomas, uhlar@fantomas.sk ; http://www.fantomas.sk/
Warning: I wish NOT to receive e-mail advertising to this address.
Varovanie: na tuto adresu chcem NEDOSTAVAT akukolvek reklamnu postu.
- Holmes, what kind of school did you study to be a detective?
- Elementary, Watson. -- Daffy Duck & Porky Pig
Re: RP_MATCHES_RCVD
Posted by Benny Pedersen <me...@junc.eu>.
Thomas Harold skrev den 2014-04-17 19:01:
> (Hopefully next month I can help out with the mass-check.)
should it not be like
meta RP_UNLISTED_HAM (!RP_MATCHES_RCVD)
if it should score as spam ?
if just scores are changed, then its another problem imho
Re: RP_MATCHES_RCVD
Posted by Thomas Harold <th...@nybeta.com>.
On 4/17/2014 9:14 AM, Kevin A. McGrail wrote:
>
>> it's not corrected, that's the point...
>>
> The scoring occurs from automatic corpus checks. The best way to help
> the rule score better is to help with masscheck.
>
It's not really a good indicator of spam/ham here either. A moderate
amount of spam is being marked as ham due to that rule's weight.
This rule was discussed back in Oct/Nov 2013, after which the rule was
manually set to -0.001. And it stayed that way until at least Feb 28th
of this year. Then during the first few weeks of March 2014, someone
converted it to a T_ rule before re-releasing it.
(Hopefully next month I can help out with the mass-check.)
Re: RP_MATCHES_RCVD
Posted by "Kevin A. McGrail" <KM...@PCCC.com>.
> it's not corrected, that's the point...
>
The scoring occurs from automatic corpus checks. The best way to help
the rule score better is to help with masscheck.
Looking at
http://ruleqa.spamassassin.org/?daterev=20140416-r1587834-n&rule=RP_MATCHES_RCVD&srcpath=&g=Change
there does appear to be a hamminess to the rule and it justifies a
negative score. A score of -1.05 seems appropriate to me.
Regards,
KAM
Re: RP_MATCHES_RCVD
Posted by Matus UHLAR - fantomas <uh...@fantomas.sk>.
>Thomas Harold skrev den 2014-04-15 05:49:
>>Mar 24th - RP_MATCHES_RCVD = -0.535
>>Mar 27th - RP_MATCHES_RCVD = -0.371
>>Apr 7th - RP_MATCHES_RCVD = -0.271
>>Apr 14th - RP_MATCHES_RCVD = -0.989
>>
>>Running 3.3.1 on CentOS 6 (from the @updates channel). Running
>>"sa-update" daily.
On 15.04.14 07:18, Benny Pedersen wrote:
>what is the problem ?, the scores is adjusted by public corpus, so
>if there is score that is not correct its a sign of missing ham/spam
>to correct it
the problem with this rule is (and was) that it often pushes score under the
spam threshold. It was complained here more times IIRC.
I have complained about this too, and I still have in my cf:
/etc/spamassassin/local.cf:score RP_MATCHES_RCVD 0
This rule is imho just something that should not be used as a whole.
No complaints against metas for now.
>other then that spamassassin does not just counts on one rule, so
>even if that rule seems incorrect hitting then it corrected by other
>rules
it's not corrected, that's the point...
--
Matus UHLAR - fantomas, uhlar@fantomas.sk ; http://www.fantomas.sk/
Warning: I wish NOT to receive e-mail advertising to this address.
Varovanie: na tuto adresu chcem NEDOSTAVAT akukolvek reklamnu postu.
Posli tento mail 100 svojim znamim - nech vidia aky si idiot
Send this email to 100 your friends - let them see what an idiot you are
Re: RP_MATCHES_RCVD
Posted by Benny Pedersen <me...@junc.eu>.
Thomas Harold skrev den 2014-04-15 05:49:
> (during first few weeks of March it was showing as T_RP_MATCHES_RCVD,
> -0.01)
note rules that begins with T_ is corpus testing rules, also why it
score just 0.01 here
> Mar 24th - RP_MATCHES_RCVD = -0.535
> Mar 27th - RP_MATCHES_RCVD = -0.371
> Apr 7th - RP_MATCHES_RCVD = -0.271
> Apr 14th - RP_MATCHES_RCVD = -0.989
>
> Running 3.3.1 on CentOS 6 (from the @updates channel). Running
> "sa-update" daily.
what is the problem ?, the scores is adjusted by public corpus, so if
there is score that is not correct its a sign of missing ham/spam to
correct it
other then that spamassassin does not just counts on one rule, so even
if that rule seems incorrect hitting then it corrected by other rules