DSPAM can give results of 99.995 accuracy. (I've had that at times), but the
best I've seen with SA is 95% correct.
We use DSPAM on a large number of customers, and as its not Heurestic, there
s no need to update anything (like you have to do with SA).
Training/retraining can be make to do it automaticly for customers (via
forwarding or via webmail etc)
Its just as good (or better) than other commercial software.
Thanks
Craig
----- Original Message -----
From: "Marc Perkel" <marc@???>
To: "Felix Schwarz" <Felix.Schwarz@???>
Cc: <exim-users@???>
Sent: Thursday, November 17, 2005 4:09 AM
Subject: Re: [exim] Re: Who likes DSPAM?
>
>
> Felix Schwarz wrote:
>
>> Marc Perkel wrote:
>>
>>> Just looking for a quick opinion - who likes DSPAM?
>>
>>
>> Me.
>>
>> I switched ca. 9 months ago when SpamAssassin failed to detect more and
>> more spam (had around 5-10 false negatives in a week with ca. 3500
>> mails/week). DSPAM is far better (2-3 false negatives) and produces less
>> false positives especially for unknown senders/uncommon subjects. DSPAM
>> classified messages as ham which I would have deleted without further
>> notice because they look so spammy but in reality they WERE ham.
>>
>> DSPAM needed quite a big training corpus for me (10000 ham/spam mails)
>> and you should test it thoroughly as it rewrites mails (signature in the
>> body) and imho these routines could be implemented more
>> conservatively/carefully. For example in 3.4 it destroyed subjects with
>> umlauts and S/MIME signed messages. While the first bug was fixed in 3.4,
>> S/MIME signed message were not (at least until 3.4.7, don't know newer
>> versions).
>>
>> fs
>>
>>
>
> Can it be made to not modify the message at all or to just add one header
> line with a score? I just want to use it to get a second opinion and then
> process the scroe with Spamassassin.
>
> --
> Marc Perkel - marc@???
>
> Spam Filter: http://www.junkemailfilter.com
> My Blog: http://marc.perkel.com
>
>
> --
> ## List details at http://www.exim.org/mailman/listinfo/exim-users ## Exim
> details at http://www.exim.org/
> ## Please use the Wiki with this list - http://www.exim.org/eximwiki/
>
