Am 19.09.2014 um 14:42 schrieb Martin Gregorie:
> On Fri, 2014-09-19 at 13:47 +0200, Reindl Harald wrote:
>>> Most DNSBL tests are done on the last relay into the internal network.
>>>
>>> I'm not say this should be done, I'm saying that it's one reason why
>>> scanning mailing list can be more trouble than it's worth
>>
>> but how is that different to any other mail?
>>
> Probably because the last relay (the listserver) can be trusted (almost
> by definition) but its message sources cannot and some mailing lists
> don't check all submissions for spam. IME that applies especially to
> mailing lists that work alongside web forums to provide subscriber
> access by both web browser and e-mail.
> 
> Obviously it is as easy to scan incoming e-mails before submitting them
> to the listserver and copying them to the web forum as it is to scan any
> inbound mail stream. 
> 
> However, it seems to be a lot harder to scan web input before accepting
> it into the forum and copying it to the associated mailing list.
> Consequently, a large proportion of the spam in these combined lists
> seems to be posted via the web forum. While volunteers usually monitor
> for and remove spam from the forum, that doesn't prevent web submitted
> spam from going out on the mailing list before the volunteers can
> recognise and deal with it.
> 
> At the same time the web-sourced spam has no header-type information to
> show where it came from before it hit the listserver, so SA checks on it
> are pretty much limited to body, rawbody and URI rules: reliably scoring
> spam from these sources is hard. 

that may all be true but don't change anything how a mail is handeled
nor does it bother the topic - no idea why now everybody hangs on
*one example* instead the topic - and no - detect spam from the body
is *not* hard, it works perfectly with the existing tags and bayes

what i want is some levels for sender/sender-domains or rcpt/rcpt-domains
with several negative scores and a simple name in a webinterface working
similar to "more_spam_from" but *more different* and *self defined* levels

frankly the whole question is in the subject and if i have one working example
how to implement that in context of a single user "sa-milter" and assign
domains/mailaddresses to that groups in "local.cf" i can make multiple ones

so can we please stop that endless meta-discussions

* is it possible
* if yes how

i can read the complete documentation and sourcecode for my own
the intention of my question to a ML is "has anybdody done that
before and could provide a working example"

Attachment: signature.asc
Description: OpenPGP digital signature

Reply via email to