On 1/24/2013 8:42 AM, Jeroen Geilman wrote:
> On 01/24/2013 07:08 AM, Stan Hoeppner wrote:
>> On 1/23/2013 2:23 PM, Grant wrote:
>>>>> I thought my postfix setup was configured to send mail on port 587 and
>>>>> receive mail on port 25, so I was surprised to find that I could send
>>>>> mail from the local machine on port 25.  Is my config OK?
>>>> Postfix never sends mail *from* TCP 25 or TCP 587.  These are receive
>>>> ports.  Outbound connections occur on high ports.  You're not properly
>>>> describing your use case, actually not at all.  Would you please?
>>> You're right, I didn't word that correctly.  I thought mail received
>>> on port 25 could only be delivered locally with my config, but I was
>>> able to send mail to any destination via port 25.  The mail client and
>>> mail server are on the same machine.
>> You haven't identified a problem Grant.  You've identified standard
>> Postfix behavior and told us it is confusing to you.  We have no idea
>> why that is confusing to you because you haven't told us exactly how you
>> are trying to use Postfix.  One thing I can tell you up front is that
>> using authentication between your MUA and Postfix on 587 is useless,
>> completely unnecessary, because the packets are transferred via machine
>> memory, never going over the wire.  The submission service exists
>> strictly for accepting authenticated connections over a network.  Your
>> connections exist entirely within on machine.
>>
> 
> If he is actually using SMTP submission on the local server, that is
> obviously untrue.

So you're saying all interprocess communication should require
authentication and encryption?  Hmm.. how many of the applications you
run do this Jeroen?

> The workings of SMTP submission are not dependent on where this happens
> from.
> 
> I would recommend submission regardless of goal or purpose, even on
> localhost.

That's because you seem to be looking at this backwards.

smtp over TLS with auth has a single goal:  security.  What additional
security is provided by using TLS and auth for interprocess
communication on a single user PC?  I.e. what is the attack vector here,
and how does 'submission' prevent such an atack?  Answer:  there is no
attack vector, thus it doesn't help.

-- 
Stan


Reply via email to