Am 03.09.2014 um 19:16 schrieb Ted Mittelstaedt:
> 
> 
> On 9/2/2014 1:52 PM, Reindl Harald wrote:
>>
>> Am 02.09.2014 um 22:32 schrieb Ted Mittelstaedt:
>>> On 9/2/2014 4:59 AM, Reindl Harald wrote:
>>>>> just get a proper MTA, enable debug logging
>>>>> and watch the commands / responses between
>>>>> client and server due a message transmission
>>>>
>>>> and to make it clear for you:
>>>>
>>>> until after end of data itslef is responded with success
>>>> the message is *undelivered* and tried again from the
>>>> sendig client if it is a proper MTA
>>>>
>>>
>>> However you have GIVEN THE SPAMMER AN OK that they have a
>>> valid victim address.  You had to issue an OK to the
>>> RCPT TO: to get that DATA from them.  You just told them "you
>>> got a good email address"
>>
>> child you do not realize that all you claim below has
>> nothing to do with SA, nor did you understand how
>> a *layered* spam protecton works nor did you try
>> to understand *anything* i explained you
>>
> 
> You changed the discussion from what SA does to your Postfix
> pre acceptance filter solution many posts ago.  Weren't you 
> paying attention to your own posts?

i explained you a *complete setup* and why SA is only a piece

>> so what - otherwise i had even accepted the message as you do
>>
>> your setup:
>>   * not on RBL
>>   * accept it and drop it silently because the score
>>   * issue "250 OK i even took the whole message"
>>   * how do you think you don't leak the RCPT case
>>   * frankly with the 250 OK you invite to send more spam
>>
>> my setup:
>>   * not on a RBL
>>   * reject it
>>   * don't issue "250 OK i even took the whole message""
>>   * if it was not a spammer trigger a bounce on the
>>     senders server so that he don't think it was
>>     successful delivered and can even prove it by logs
>>
> 
> Your words:
> 
> "i am saying to my milter "above score XYZ reject the message" because
> only the [SPAM] in the subject is worthless, it just indicates which
> messages should go to "spam" or "ham" fpr bayes training"
> 
> I interpreted this to mean you are attempting to content filter on
> the Subject: line with a milter.  That is a post-acceptance, 
> post-recipient-address exposure-leak kind of filter
> with the drawbacks I already illustrated

what SA typically does is add the [SPAM] to subejct and
a header to allow filtering for MUA or Sieve - and
that is not really helpful

> Once more your treating postscreen like a black box and conflating
> it with the discussion of what pre acceptance filtering is and 
> what content filtering is.  It's just more Postfix
> evangelizing.

no - the be behavior of a complete setup is that only
a small piece makes it to the extensive content filter

wether you use postfix , exim or something else
well, postfix indeed has some unique features

> if postscreen rejects on a DNSBL without issuing an OK to the
> RCPT TO: then it's NOT content filtering or leaking addresses

the *whole* point was to explain you that content filtering
is only one part of the game

> It is pre acceptance filtering.  This is the same as greylisting 
> (which I do and which mainly eliminates dumb bots that ignore 
> SMTP response codes), and many other kinds of pre-acceptance
> filters (PTR checks, milter-callback schemes, SPF and DKIM, etc.)

it has *nothing* to do with greylisting
greylisting answers with a 4xx code

http://www.postfix.org/postconf.5.html#smtpd_delay_reject

> And this discussion has little to do with SpamAssassin.  YOU dragged
> postscreen in here and the notion of pre-acceptance filtering
> (when you referred to postscreen using DNSBLs) several posts ago.

because i tried to explain you that SA alone is not
a ready solution as you expect and as many others
explained you: it is a framework and *part* of a
solution which needs knowledge to setup

> If postscreen rejects on a scan of the Subject: line then it
> has issued an OK to a RCPT TO: and leaked an address - and it is
> content filtering the same as SpamAssassin.

that's how a spamfilter works: different stages:

* IP / DNSBL -> OK / Reject
* PTR: OK / Reject
* Sender address: OK / Reject
* SPF: OK / Reject
* Subject: OK / Reject
* Contentfilter: Ok / REJECT within the session
* Virus Scanner: OK / REJECT within the session

> At this point I think that further demonstrations from each of us as to what 
> the
> SMTP handshake is are rather pointless.

yes

> In reviewing your posts and my responses it's obvious that your posts are 
> all based on a Postfix/postscreen setup and you are making an assumption 
> that everyone out there works with this software, and can understand what 
> your saying because they understand that context.

i expect if someone talks about a contentscanner he understands
the context independet of the software and keep in mind milters
are not postfix-specific, frankly they are a sendmail thing
originally

http://www.postfix.org/MILTER_README.html
http://en.wikipedia.org/wiki/Milter

*that* would have been postfix specific
http://www.postfix.org/SMTPD_PROXY_README.html

> This is the mark of a software zealot and evangelist.

it don't matter which MTA you use in combination with a contentfilter

the main difference is reject or accept and deliver anyways or drop
silent and the main point is that you don't want to handoff 90%
of your incoming mailflow to the contentscanner at all

> I am attempting to respond based on the normal understanding of what the
> basic blocks of email transmission are using vendor-neutral language
> as much as possible.  It is no wonder your getting terribly pissed off 
> about it.  My refusal to engage in this discussion by accepting all your 
> vendor-specific Postfix terms, you are interpreting as a slap in the face to
> Postfix and your enraged by this.

if we are talk about email you need to understand what a milter is, what
before-queue filtering is, what a queue is and on which levels of the
protocl you can snap in which filters and how "expensive" they are
to make decisions

> I want to talk about ideas

me too

> You want to talk about Postfix.  

no i explained on *examples* how things are woring

> That is fine except this isn't a Postfix mailing list.

that's not the point

you expect 90% filtering by a single contentfilter out of the box
and refuse to understand that this is just impossible and that
you pretend gmail can do that is a impudence because no user
ever had a touch with "oout-of-the-box" of gmail

> When you can drop all references to Postfix terms and drop the 
> insistence that all MTA's operate the same as an MTA running 
> Postfix and all it's associated programs, then I think we can get 
> somewhere.

no, you just could Google somethings, read what i talk
about *before* you respond and then seek for how to
do that with your MTA or even realize why i use
a specific one

> I will simply end by echoing something you said earlier:
> content filtering should be the LAST RESORT in the filtering chain

which i told you at the very beginning of that thread
and after you insited in SA has to catch more i explained you
why that is impossible and others tried that too

> and I will add my own to that:
> 
> NOBODY has come up with a filter yet that works well and 
> ONLY considers the SMTP handshake BEFORE issuing an OK on
> receipt of the recipient's address, that has any long-term 
> stick-tion

because it is technically impossible

* you don't know if you accept the body before you scan it
* you must answer the question for the valid RCPT before

> Every one that has been created has had countermeasures worked 
> out by spammers. The fact that many unsophisticated spammers 
> don't use these countermeasures is beside the point, enough do 
> that too much spam gets past to NOT use content filtering 
> (such as provided by SA)

there are no countermeasures for a spammer against make it
on a RBL or use a zombie on a infected machine and get
blocked by Dialup-RBL's before the first mail or by
get rejected because the dynamic PTR of the infected
zombie

you can setup your own servers but they will land in
a short on a RBL (the R is for real time) and they
have not the same throughbot than a zombie-network

> Thus, while SA (and any content filter) should be the last 
> filter in the chain, it's impossible to not have it - it 
> is required.

to filter out the remaining 5-10% percent

> In fact, (although I DO NOT do this) many people just accept 
> everything and put SA at the beginning of the filter chain
> and get almost the same results

no - read how a milter works (not postfix specific)

it's in the middle of the chain and get the whole data
from headers to body, but the MTA can reject a message
based on the envelope data and stop before body

Attachment: signature.asc
Description: OpenPGP digital signature

Reply via email to