You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@spamassassin.apache.org by John Fleming <jo...@wa9als.com> on 2005/06/14 21:29:23 UTC
Bayes?
Not sure what I'm doing wrong, but noone has responded to my question -
If Bayes is working properly, am I supposed to have a Bayes_ entry in the
X-Spam_Status header info for every email? I think that's the way it used
to be - Bayes_00 - Bayes_99 etc. Lately it seems that only the spam emails
have those entries.
What am I missing? I apologize if this should be obvious, but if it is,
I've messed it and would really like to understand why my setup has changed.
My config is simple - calling spamc from procmail. Still working well, but
I have less entries in the headers and often no Bayes... Thanks - John
Re: SOLVED Bayes? (MORE)
Posted by Kai Schaetzl <ma...@conactive.com>.
John Fleming wrote on Thu, 16 Jun 2005 13:08:22 -0500:
> I'm not sure I fully understand why the single digit entries allowed BAYES_
> score headers on spam but not ham, but I'm happy to use the multiple scores
> and have all of my BAYES_ entries back.
Hm, just checked our setup and we changed one setting:
score BAYES_99 3.0
We don't have this problem. Are you sure there wasn't something wrong with the
added scores? Apart from that I don't recommend changing much in that area,
certainly not more than one or two scores. We changed 99, because it is so
reliable and a higher setting can grab those spams which otherwise slip thru
right below the threshold.
Kai
--
Kai Schätzl, Berlin, Germany
Get your web at Conactive Internet Services: http://www.conactive.com
IE-Center: http://ie5.de & http://msie.winware.org
SOLVED Re: Bayes? (MORE)
Posted by John Fleming <jo...@wa9als.com>.
----- Original Message -----
From: "Kai Schaetzl" <ma...@conactive.com>
To: <us...@spamassassin.apache.org>
Sent: Thursday, June 16, 2005 8:07 AM
Subject: Re: Bayes? (MORE)
> John Fleming wrote on Wed, 15 Jun 2005 18:25:28 -0500:
>
>> Note that there -is- a BAYES_ entry! I have some modified Bayes scores
>> in
>> my local.cf, and some of them only have one numerical entry after the
>> title
>> instead of the usual 4. Can anyone imagine that does something strange
>> to
>> the Bayes?
>
> Hm, I don't know. One of the figures *is* for running without BAYES. Maybe
> change to four figures and see if that helps? Or check the original rules
> in
> /usr/share/spamassassin for the syntax they used for scoring BAYES. At
> least
> *that* ham got a score ;-)
>
> Kai
THANKS, Kai - That was it! I removed my single score Bayes entries, and the
BAYES_ scores for my ham came back:
X-Spam-Status: No, score=-2.6 required=5.0 tests=BAYES_00 autolearn=ham
version=3.0.3
I'm not sure I fully understand why the single digit entries allowed BAYES_
score headers on spam but not ham, but I'm happy to use the multiple scores
and have all of my BAYES_ entries back.
- John
Re: Bayes? (MORE)
Posted by Kai Schaetzl <ma...@conactive.com>.
John Fleming wrote on Wed, 15 Jun 2005 18:25:28 -0500:
> Note that there -is- a BAYES_ entry! I have some modified Bayes scores in
> my local.cf, and some of them only have one numerical entry after the title
> instead of the usual 4. Can anyone imagine that does something strange to
> the Bayes?
Hm, I don't know. One of the figures *is* for running without BAYES. Maybe
change to four figures and see if that helps? Or check the original rules in
/usr/share/spamassassin for the syntax they used for scoring BAYES. At least
*that* ham got a score ;-)
Kai
--
Kai Schätzl, Berlin, Germany
Get your web at Conactive Internet Services: http://www.conactive.com
IE-Center: http://ie5.de & http://msie.winware.org
Re: Bayes? (MORE)
Posted by John Fleming <jo...@wa9als.com>.
----- Original Message -----
From: "Kai Schaetzl" <ma...@conactive.com>
To: <us...@spamassassin.apache.org>
Sent: Wednesday, June 15, 2005 4:32 PM
Subject: Re: Bayes? (MORE)
> John Fleming wrote on Tue, 14 Jun 2005 15:56:02 -0500:
>
>> Running spamassassin NOW results in ONLY the following, and then it
>> hangs:
>>
>> login as: root
>> Password:
>> Last login: Tue Jun 14 15:19:49 2005 from 66.209.225.25
>> Luke:~# spamassassin -D
>
> John, you have to pipe a message thru it or at least --lint.
>>>From your magic data it seems you have quite a big Bayes db. Run a forced
> expire to see how long it takes. It's possible that sa times out on it.
>
> Kai
Duh - thanks, I should've known that. Here's the end of the information
from spamassassin -D --lint:
debug: running uri tests; score so far=-3.281
debug: running raw-body-text per-line regexp tests; score so far=-3.281
debug: running full-text regexp tests; score so far=-3.281
debug: is spam? score=-3.281 required=5
debug:
tests=ALL_TRUSTED,BAYES_20,MISSING_HEADERS,MISSING_SUBJECT,NO_REAL_NAME,SARE_TOCC_NONE
debug:
subtests=__HAS_MSGID,__MSGID_OK_DIGITS,__MSGID_OK_HOST,__SANE_MSGID,__SARE_CC_NONE,__SARE_HTML_HAS_MSG,__SARE_TO_NONE,__UNUSABLE_MSGID
#
Note that there -is- a BAYES_ entry! I have some modified Bayes scores in
my local.cf, and some of them only have one numerical entry after the title
instead of the usual 4. Can anyone imagine that does something strange to
the Bayes?
Again, the observation is that all caught spam has a high BAYES_ score, but
all ham doesn't have any BAYES_ score entry in the header info. Thanks -
John
Re: Bayes? (MORE)
Posted by Kai Schaetzl <ma...@conactive.com>.
John Fleming wrote on Tue, 14 Jun 2005 15:56:02 -0500:
> Running spamassassin NOW results in ONLY the following, and then it hangs:
>
> login as: root
> Password:
> Last login: Tue Jun 14 15:19:49 2005 from 66.209.225.25
> Luke:~# spamassassin -D
John, you have to pipe a message thru it or at least --lint.
>From your magic data it seems you have quite a big Bayes db. Run a forced
expire to see how long it takes. It's possible that sa times out on it.
Kai
--
Kai Schätzl, Berlin, Germany
Get your web at Conactive Internet Services: http://www.conactive.com
IE-Center: http://ie5.de & http://msie.winware.org
Re: Bayes? (MORE)
Posted by John Fleming <jo...@wa9als.com>.
----- Original Message -----
From: "Matt Kettler" <mk...@evi-inc.com>
To: "John Fleming" <jo...@wa9als.com>
Cc: <us...@spamassassin.apache.org>
Sent: Tuesday, June 14, 2005 2:48 PM
Subject: Re: Bayes?
> John Fleming wrote:
>> Not sure what I'm doing wrong, but noone has responded to my question -
>>
>> If Bayes is working properly, am I supposed to have a Bayes_ entry in the
>> X-Spam_Status header info for every email?
> In 3.x, I think you should, but I can't speak with any authority here.
>
> In 2.6x, definitely not. Anything which is at the 50/50 mark won't show
> any
> bayes hits at all for 2.6x users.
>
>
>> I think that's the way it used to be - Bayes_00 - Bayes_99 etc. Lately
>> it seems that only the spam emails
>> have those entries.
>
> Hmm, that's a little odd. You might try hand running some of the messages
> through spamassassin -D and see if the debug gives you any clues.
>
Sorry, forgot to add -
Running spamassassin NOW results in ONLY the following, and then it hangs:
login as: root
Password:
Last login: Tue Jun 14 15:19:49 2005 from 66.209.225.25
Luke:~# spamassassin -D
debug: SpamAssassin version 3.0.3
debug: Score set 0 chosen.
debug: running in taint mode? yes
debug: Running in taint mode, removing unsafe env vars, and resetting PATH
debug: PATH included '/usr/local/sbin', keeping.
debug: PATH included '/usr/local/bin', keeping.
debug: PATH included '/usr/sbin', keeping.
debug: PATH included '/usr/bin', keeping.
debug: PATH included '/sbin', keeping.
debug: PATH included '/bin', keeping.
debug: PATH included '/usr/bin/X11', keeping.
debug: Final PATH set to:
/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/bin/X11
and dump magic results in:
dumpmagic
0.000 0 3 0 non-token data: bayes db version
0.000 0 37233 0 non-token data: nspam
0.000 0 60839 0 non-token data: nham
0.000 0 124202 0 non-token data: ntokens
0.000 0 1118109190 0 non-token data: oldest atime
0.000 0 1118782164 0 non-token data: newest atime
0.000 0 1118781255 0 non-token data: last journal sync
atime
0.000 0 1118756260 0 non-token data: last expiry atime
0.000 0 646651 0 non-token data: last expire atime
delta
0.000 0 29477 0 non-token data: last expire
reduction count
#
Any ideas?
Re: Bayes?
Posted by John Fleming <jo...@wa9als.com>.
----- Original Message -----
From: "Matt Kettler" <mk...@evi-inc.com>
To: "John Fleming" <jo...@wa9als.com>
Cc: <us...@spamassassin.apache.org>
Sent: Tuesday, June 14, 2005 2:48 PM
Subject: Re: Bayes?
> John Fleming wrote:
>> Not sure what I'm doing wrong, but noone has responded to my question -
>>
>> If Bayes is working properly, am I supposed to have a Bayes_ entry in the
>> X-Spam_Status header info for every email?
> In 3.x, I think you should, but I can't speak with any authority here.
>
> In 2.6x, definitely not. Anything which is at the 50/50 mark won't show
> any
> bayes hits at all for 2.6x users.
>
>
>> I think that's the way it used to be - Bayes_00 - Bayes_99 etc. Lately
>> it seems that only the spam emails
>> have those entries.
>
> Hmm, that's a little odd. You might try hand running some of the messages
> through spamassassin -D and see if the debug gives you any clues.
>
Here's my headers - SA 3.0.3 on Debian
X-Spam-Checker-Version: SpamAssassin 3.0.3 (2005-04-27) on Luke.wa9als.com
X-Spam-Level:
X-Spam-Status: No, score=0.0 required=5.0 tests=HTML_90_100,HTML_MESSAGE
autolearn=no version=3.0.3
Perl is 5.8.4 on Debian Stable (Sarge)
spamassassin --lint is OK; Bayes is turned on in local.cf and no recent
changes to that file.
A week or so ago I ran spamassassin -D, and the interesting thing I saw was
that the "Bayes test" (I don't know the specifics of this) worked fine and
came up with a percentage score.
All my caught spam has Bayes headers with high scores, but none of the ham
has any Bayes entries.
Other thoughts? Tnx - John
Re: Bayes?
Posted by Matt Kettler <mk...@evi-inc.com>.
John Fleming wrote:
> Not sure what I'm doing wrong, but noone has responded to my question -
>
> If Bayes is working properly, am I supposed to have a Bayes_ entry in the
> X-Spam_Status header info for every email?
In 3.x, I think you should, but I can't speak with any authority here.
In 2.6x, definitely not. Anything which is at the 50/50 mark won't show any
bayes hits at all for 2.6x users.
> I think that's the way it used to be - Bayes_00 - Bayes_99 etc. Lately it seems that only the spam emails
> have those entries.
Hmm, that's a little odd. You might try hand running some of the messages
through spamassassin -D and see if the debug gives you any clues.