On Wed, 22 Dec 2004, Bruce Lilly wrote:
Yes. I remarked on that a half-year ago on the mail-ng
list (http://www.imc.org/mail-ng/mail-archive/msg00715.html),
and the issue came up recently on the ietf-822 list
(http://www.imc.org/ietf-822/mail-archive/msg05290.html).
I agree that the 822 message header is uncomfortably stuck doing two jobs,
but I disagree that this implies that the term "envelope" should include
any header fields.
I did a more thorough survey of RFCs related to internet email (I can send
people my notes if they want). I excluded X.400 and other non-internet
protocols, and MIME types not related to the transport - so I included
DSNs but excluded S/MIME (which only uses "envelope" to talk about
cryptographic encapsulation).
The first time the envelope is defined as the MAIL FROM and RCPT TO
commands is in RFC 976's definition of batch SMTP; however it's also
implicit in RFC 822.
The first time the SMTP envelope is properly defined is in RFC 1123. There
are 50ish email-related RFCs in the 15 years since then which mention the
word "envelope" and the VAST MAJORITY of them either explicitly or
implicitly use the same definition.
The exceptions include the IMAP RFCs (because of the name of an IMAP
protocol element which substantially predates 1123) and a few instances of
sloppy wording.
In RFC 2130 "envelope" is being used in the general sense of protocol
encapsulation - a place where you might include the character set
information related to the encapsulated data - and MIME encapsulation is
used as an example. This isn't strictly email-related so isn't really a
counter-example.
In RFCs 2442 and 2554 "envelope" is used where it would be more correct to
say "envelope and header". This is probably just an omission. In both
cases they are talking about security of message transport protocol
elements which are harder to secure than the message header.
Having said that I excluded X.400 RFCs (which exclusively use the term
"envelope" to refer to X.400 protocol elements, and are generally
abstracted away from SMTP) there's a useful comment in RFC 1506:
- RFC 1327 doesn't define how to map the MTS.OriginatorName and
the MTS.RecipientName (often referred to as the P1.originator
and P1.recipient), since this depends on which underlying 822-
MTS is used. In the very common case that RFC 821 (SMTP) is
used for this purpose, the mapping is normally as follows:
MTS.Originator-name <-> MAIL FROM:
MTS.Recipient-name <-> RCPT TO:
A lot of X.400 envelope information has to be put into the 822 header when
gatewaying because of the difference in architecture between X.400 and
internet email, but I don't see why this should have any bearing on the
meaning of "envelope" in the internet world - and indeed it doesn't seem
to have had any, judging by my surveys.
So I still think there is no lack of consensus in the meaning of
"envelope" in the context of internet email, and this has been the
case for at least 15 years.
Tony.
--
f.a.n.finch <dot(_at_)dotat(_dot_)at> http://dotat.at/
BERWICK ON TWEED TO WHITBY: WEST 6 TO GALE 8 BACKING SOUTHWEST 4 OR 5 LATER.
RAIN THEN SHOWERS. MODERATE OR GOOD. MODERATE OR ROUGH.