> On Sep 27, 2018, at 1:24 PM, Gary Chambers <[email protected]> wrote:
>
> Viktor,
>
>>> -Queue ID- --Size-- ----Arrival Time---- -Sender/Recipient-------
>>> 77E39131AF 2895 Thu Sep 27 11:42:12 [email protected]
>>> (host smtp.example.com[00.000.00.000] said: 450 4.1.8
>>> <[email protected]>:
>>> Sender address rejected: Domain not found (in reply to RCPT TO command))
>>> [email protected]
>>
>> Note that the *sender* domain is "cuddy.localdomain". Note also that
>> the reject message is "*Sender* address rejected". Does "cuddy.localdomain"
>> resolve via DNS on your system?
>
> I did note the sender domain, and cuddy.localdomain does not resolve via DNS
> on my system. My question is why the identical configuration works in
> Postfix 3.1 and 2.9.
>
>> Actually, it is the *remote* system "smtp.example.com" that did not accept
>> this mail, because the *sender* address is invalid. Not much you can do
>> about that on the sending end. Either use a valid sender address, or
>> convince the remote system to accept the mail.
>
> I control smtp.example.com, so I'm able to make it accept whatever I'm
> sending to it. I also don't mind making the changes, but I'd like to
> understand why I'm making them because of the apparent different behavior in
> the previous Postfix versions.
The difference between Postfix versions you speak of is mostly an illusion.
You have a message in your queue with a sender address that the remote
system is unwilling to accept. That's all. If the sender address was not
supposed to be "cuddy.localdomain", you need to have suitable settings for
some combination of:
myorigin
append_dot_mydomain
smtp_generic_maps
to ensure that outbound mail carries a valid domain. Note that Debian
and Ubuntu systems set "myorigin" outside of Postfix:
myorigin = /etc/mailname
Most likely the new system has a different /etc/mailname than the old.
The only pertinent change in Postfix is that the default value of
append_dot_mydomain changed to "no" as of Postfix 3.0.
--
Viktor.