Tracking Compound Meta's

[Dan Patnode <da...@patnode.net>]

I've confirmed that meta's within meta's within meta's work well:

body __testA /\ba/i
body __testB /\bb/i
body __testC /\bc/i
body __testD /\bd/i
body __testE /\be/i
body __testF /\bf/i
body __testG /\bg/i
body __testH /\bh/i
body __testI /\bi/i
body __testJ /\bj/i
body __testK /\bk/i
body __testL /\bl/i

meta __META_ABC (__testA || __testB || __testC)
meta __META_DEF (__testD || __testE || __testF)
meta __META_GHI (__testG || __testH || __testI)
meta __META_JKL (__testJ || __testK || __testL)

meta META_ABCDEF (__META_ABC || __META_DEF)
meta META_GHIJKL (__META_GHI || __META_JKL)
meta META_ABCDEFGHIJKL (META_ABCDEF || META_GHIJKL)


but, combined with neutralized (__) tests, the score line doesn't  
show which individual tests were triggered:

X-SpamAssassin: score=3.0  
tests=META_ABCDEF,META_ABCDEFGHIJKL,META_GHIJKL


Leaving them non scoring (__), what are all the ways of determining  
which sub tests (body in this case) participated in triggering a  
given meta?  I'm looking for all the places that store this  
information as well as any way to show the individual tests in the  
score line itself (while still scoring 0 each).

Thanks,
Dan

Re: Tracking Compound Meta's

[Theo Van Dinter <fe...@apache.org>]

On Fri, Apr 28, 2006 at 12:37:40AM -0400, Matt Kettler wrote:
> do such detailed analysis on live email... If you really must do it
> live, the 0.001 score seems to be your best option.

If you really really wanted it in the headers, you could open a BZ ticket
requesting an enhancement to create a template tag for the subtests.

In the meantime, you could create a plugin to do that too. :)

-- 
Randomly Generated Tagline:
"Good news, good news abounds."            - Prof. Branche

Re: Tracking Compound Meta's

[Matt Kettler <mk...@comcast.net>]

Dan wrote:
>> Is 0.001 a low enough score to be considered to be zero?
>
> I'm building a very precise configuration so any variations create
> headaches, but if I can't find any alternatives, I may have to.
>
>
>> You can run the message through spamassassin -tD. The debug output will
>> include the full list of tests, including subtests, matched.
>
> The messages I will need this for often won't include headers, and
> many of my tests are header based, so reprocessing the message wont be
> luxury I'll have.
>
> Any other ways to create or access the information historically? 

Not really..  Most folks that actually need to record sub-tests of a
meta rule wind up using mass-check, or spamassassin -tD. Very few people
do such detailed analysis on live email... If you really must do it
live, the 0.001 score seems to be your best option.

Re: Tracking Compound Meta's

[Dan <a...@patnode.net>]

> Is 0.001 a low enough score to be considered to be zero?

I'm building a very precise configuration so any variations create  
headaches, but if I can't find any alternatives, I may have to.


> You can run the message through spamassassin -tD. The debug output  
> will
> include the full list of tests, including subtests, matched.

The messages I will need this for often won't include headers, and  
many of my tests are header based, so reprocessing the message wont  
be luxury I'll have.

Any other ways to create or access the information historically?

Thanks,
Dan

Re: Tracking Compound Meta's

[Matt Kettler <mk...@comcast.net>]

Dan Patnode wrote:
> I've confirmed that meta's within meta's within meta's work well:
>
> body __testA /\ba/i
<snip>
> but, combined with neutralized (__) tests, the score line doesn't show
> which individual tests were triggered:
>
> X-SpamAssassin: score=3.0 tests=META_ABCDEF,META_ABCDEFGHIJKL,META_GHIJKL
>
>
> Leaving them non scoring (__), what are all the ways of determining
> which sub tests (body in this case) participated in triggering a given
> meta?  I'm looking for all the places that store this information as
> well as any way to show the individual tests in the score line itself
> (while still scoring 0 each).

You can run the message through spamassassin -tD. The debug output will
include the full list of tests, including subtests, matched.

Re: Tracking Compound Meta's

[jdow <jd...@earthlink.net>]

From: "Dan Patnode" <da...@patnode.net>

> I've confirmed that meta's within meta's within meta's work well:
> 
> body __testA /\ba/i
> body __testB /\bb/i
> body __testC /\bc/i
> body __testD /\bd/i
> body __testE /\be/i
> body __testF /\bf/i
> body __testG /\bg/i
> body __testH /\bh/i
> body __testI /\bi/i
> body __testJ /\bj/i
> body __testK /\bk/i
> body __testL /\bl/i
> 
> meta __META_ABC (__testA || __testB || __testC)
> meta __META_DEF (__testD || __testE || __testF)
> meta __META_GHI (__testG || __testH || __testI)
> meta __META_JKL (__testJ || __testK || __testL)
> 
> meta META_ABCDEF (__META_ABC || __META_DEF)
> meta META_GHIJKL (__META_GHI || __META_JKL)
> meta META_ABCDEFGHIJKL (META_ABCDEF || META_GHIJKL)
> 
> 
> but, combined with neutralized (__) tests, the score line doesn't  
> show which individual tests were triggered:
> 
> X-SpamAssassin: score=3.0  
> tests=META_ABCDEF,META_ABCDEFGHIJKL,META_GHIJKL
> 
> 
> Leaving them non scoring (__), what are all the ways of determining  
> which sub tests (body in this case) participated in triggering a  
> given meta?  I'm looking for all the places that store this  
> information as well as any way to show the individual tests in the  
> score line itself (while still scoring 0 each).

Is 0.001 a low enough score to be considered to be zero?

{^_^}