spamassassin-dev December 2011 archive
Main Archive Page > Month Archives  > spamassassin-dev archives
spamassassin-dev: Re: Regarding Scoring of Mailspike

Re: Regarding Scoring of Mailspike

From: Axb <axb.lists_at_nospam>
Date: Mon Dec 12 2011 - 22:38:24 GMT
To: dev@spamassassin.apache.org

On 2011-12-12 23:26, Kevin A. McGrail wrote:
> On 12/12/2011 5:23 PM, Warren Togami Jr. wrote:
>> On Mon, Dec 12, 2011 at 12:15 PM, Kevin A. McGrail<KMcGrail@pccc.com>
>> wrote:
>>>> * Did you really intend for Mailspike to add a total of 4.1 or 4.5
>>>> points? (Given _BL is a composite rule that adds 1.0 on top of _L5 or
>>>> _ZBI.)
>>> Yes, I did.
>> Every time _L5 hits you have 4.1 points. Every time _ZBI hits you
>> have 4.5 points. This comes pretty close to exclusively trusting only
>> Mailspike. Mailspike we know is good, but does it really deserve a
>> static score this high? Has the Perceptron/GA balancing ever set any
>> DNSBL this high?
>>
>> Does this really make sense?
> Based on real-world usage with no real-world FPs, I would argue yes or I
> wouldn't have proposed the scores to begin with.

4.x for a new RBL score is scary... especially comparing with
RCVD_IN_SBL / RCVD_IN_XBL or even RCVD_IN_CSS whichis highly reliable

I'll put Mailspike to production (with low scores) as soon as I can get
my rsync stuff setup with Joćo