Tony Finch writes:
The envelope and the header are separate in Internet email, and this
distinction is much more precisely defined than any division of the
header into transport-related and user-related parts.
Is the message size part of the envelope? Your ``precise definition''
gives a wishy-washy answer. Sometimes SMTP clients put the size on the
MAIL line; sometimes they don't. It doesn't make sense for this minor
transport detail to dictate our terminology.
The fundamental responsibility of an MTA is to deliver a specified
message to a specified recipient, or to notify a specified sender if
delivery is not made. This basic feature of the infrastructure had some
influence on the structure of an SMTP conversation, obviously; but the
feature is much more important than SMTP and would exist even if SMTP
had (e.g.) encoded the sender address in a Return-Path field, as some
delivery protocols do.
Similarly, Received lines could have been transmitted through a separate
SMTP RCVD command, and then inserted into the header the same way that
Return-Path is. This wouldn't make any difference in the role that
Received lines play in the infrastructure.
I'm not saying that ``envelope'' should be given a narrow definition
(sender and recipient address). I'm not saying that it should be given a
broad definition (including postmarks, for example, and return-receipt
requests). What I'm saying is that your whole approach to the question
is broken. You're wildly exaggerating the importance of certain syntax
decisions in one protocol---decisions that could clearly have been
changed with no serious effect.
---D. J. Bernstein, Associate Professor, Department of Mathematics,
Statistics, and Computer Science, University of Illinois at Chicago