Author: Marc Perkel Date: To: exim-users Subject: Re: [exim] Alternate callout verification sequence suggestion
Yes - it could work but loop avoidance would be such a support issue
that its not worth it. I'm getting more inclined that if they don't play
be the rules then they don't get to send email to me.
But - auto reporting is still on the table. The very reason for
accepting nul sender is to avoing looping. So -
I thing we need to pressure those who don't play be the rules to fix it.
Wakko Warner wrote:
>Please keep me in CC and don't top post.
>
>
>
>>You could have a special callout from address that didn't do a callout
>>itself.
>>
>>
>
>That's not really possible.
>server A talks to server MARC. MARC does a callout with sender <> to server
>A. server A thinks <> is invalid. MARC server decides to use a callout
>with sender <callouttest@MARC> (for example) instead of <>. server A
>decides to verify that <callouttest@MARC> is a valid address (callout if you
>will) and server MARC notes that this was a callout test it had done and
>immediately says this sender is OK (making note to drop actual mail to it in
>the data phase, just incase this is a spammer!). server A decides that
>since <callouttest@MARC> is valid, it returns ok. in the callout MARC did,
>it sees it as ok and proceeds as a valid address to the original smtp
>session from server A.
>
>Now if instead that MARC was not configured to say ok to <callouttest@MARC>
>and did a callout, you would run into an infinite loop (infinite being smtp
>timeout or not. I think theoretically, it'd stop at the default of 5
>minutes or whatever the servers are using). This scenario here is what you
>want to avoid.
>
>I have stated this at least 2 times prior on the list. Phil does not want
>this capability in exim. I would assume that this is because it can easily
>cause loops, which I would agree with. The responsibility is upon the
>administrator who is setting up their machine to use a sender callout using
>a non-null address to avoid the callout loop problem. *IF* all servers that
>do callouts always use <> first and always accept <> as a valid sender, this
>probably really would not be an issue. I guess that would just be wishful
>thinking.
>
>
>
