Re: Spam filter

hi again,
On 10.11.2006, at 08:48, Vidar Tyldum Hansen wrote:

> Morten Nilsen, 09.11.2006 16:45:

...
>
>>
>> smtpd_delay_reject = yes
>> smtpd_helo_required = yes
>> smtpd_helo_restrictions =
>>   reject_non_fqdn_hostname,
> This will cause a lot of false positives in my experience. DNS is  
> widely
> broken out there as it's put in the hands of too many clueless admins.

here in austria I found two broken dns configs in two major firms,  
the rest is spam that gets rejected. maybe this situation is not  
appliable to other countries b/c here internet experience is still  
largely influenced by the big universities, and their very strict  
overall configuration example.
so most small companies have their dsl/8 ips configuration, but use  
external smtp/http servers to rely on.
most business is concentrated in and around Vienna anyways.
>
> There are more risky settings here I would not recommend for business
> usage. The main problem with this strict aproach is that you reject or
> accept based on a single parameter. I find this far too risky!
>
> I would put more effort into amavisd and spamassassin where multiple
> checks are performed and no single rule will throwaway any mail. I
> removed all these extremist checks and all RBL checks from postfix and
> let spamassassin deal with them.
this did put so much load on my machines that I take all the strict  
blocklists to take the load of amavis. it even saved electrical  
power, it is so much load we are talking about.


>
> Of course, my subjective opinion.
granted, and I would like to have the possibility to run such a  
setup, but I would like to hear more about your experience in  
statistics of spam, how did you generate this graph (http:// 
www.tyldum.com/spam-year.png) you mentioned earlier?

thanks for bringing more light into the spam-statistics corner if  
possible ...

matthias

> _______________________________________________
> tsl-discuss mailing list
> [email protected]
> http://lists.trustix.org/mailman/listinfo/tsl-discuss