On 31.08.24 11:12, Stuart Henderson wrote:
> On 2024/08/31 10:14, Christian Schulte wrote:
>>
>> "In all cases, it is the meaning that the *sender* of the message wishes
>> to convey (i.e., whether this is the same message or a different
>> message) that determines whether or not the "Message-ID:" field changes,
>> not any particular syntactic difference that appears (or does not
>> appear) in the message."
> 
> If the sending MUA doesn't include Message-ID then presumably it's not
> trying to convey any information about that.
> 
>> I then mentioned that I think OpenSMTPD gets it right by not blindly
>> adding Message-ID headers because it is not the sender and thus cannot
>> decide about the identity of a message which made you question OpenSMTPD
>> to get it right.
> 
> It is standard in most MTAs (including opensmtpd) to add missing
> Message-ID at least to emails which they consider as locally originated.
> (Doing this for mail from external origins risks DKIM failure with some
> header signing setups).
> 
> This was broken in opensmtpd on little-endian machines from 2018-2023
> for mail received on 587, and if I read the code correctly it's still
> not done for mail received on 465. Though that shouldn't affect non-SMTP
> mail submission as done by mail(1).
> 

Maybe I try to picture it in some way. Clearly every physical newspaper
(mail) is a unique entity. The Message-ID of all this newspapers (mails)
would be identical as it is the same newspaper (mail) transported to
many recipients. An MTA cannot decide on this. Only the author/sender
can apply a unique ID to the content delivered/transported/sent.

> which they consider as locally originated

Agreed. Then. And only then, if the author did not supply it's own ID
(possibly chosen from some ID scheme like a barcode on a newspaper).

Random first search result:
<https://youtu.be/l6DEL9GY-VU?si=7HduSAT8Z7KibI4B>

-- 
Christian