Mailing-List: contact dev-help@spamassassin.apache.org; run by ezmlm
Precedence: bulk
Received-SPF: pass (athena.apache.org: domain of matt@coders.co.uk designates
 12.158.191.97 as permitted sender)
Message-ID: <485F588B.4080106@coders.co.uk>
Date: Mon, 23 Jun 2008 09:02:19 +0100
From: Matt Hampton <matt@coders.co.uk>
User-Agent: Thunderbird 2.0.0.14 (Windows/20080421)
MIME-Version: 1.0
To: Justin Mason <jm@jmason.org>
CC: dev@spamassassin.apache.org
Subject: Re: Creating auto-generated rule sets.....
References: <20080620090902.37ADD30C0F3@jmason.org>
In-Reply-To: <20080620090902.37ADD30C0F3@jmason.org>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

Hi

Latest version is now avaliable

>> http://www.coders.co.uk/80_sane.cf
>>
>>     
I have expanded the support for clamav rule types so have increased the 
base ruleset generated by about 40%.
> It might be worthwhile discarding rules that are less than a certain
> length, in characters.  That's another thing the "sought" ruleset does.
> it, again, reduces FPs nicely.
>   
Which reduces it by 15%
> Also is there any way to get it to produce _more_ rules? that 80_sane.cf
> seems pretty short, compared to the 60k-rules input ;)   Sounds like
> 0.5% spam hits is too high a threshold, I think.
>
>   
Have tweaked this to 0.1% and is this is the basis for the ruleset 
avaliable above.

>> It isn't automatically updating at the moment and all of the scores are 
>> set to 0.01
>>     
>
>   
Still set to 0.01 - until I can work out a better scoring mechanism

> btw you can also safely drop the "require_version" line, that only makes
> sense as part of the SpamAssassin source tree.
>
>   
Haven't done that yet!


matt