My apologies -- I should have been more precise about chronology.
When we _first_ did mail-over-FTP, the norm was to deliver more
or less directly into the user's file system. The notion of
"spooling" or "mail store" mail-receiving processes came later --
in the Multics case, not much later, as it became clear that
direct-to-user-space delivery raised some security issues that no
one was happy about-- but well before SMTP.
So Dave, and his chronology, are quite correct.
john
--On Saturday, 27 January, 2001 06:26 -0600 Dave Crocker
<dhc2(_at_)dcrocker(_dot_)net> wrote:
At 03:14 PM 1/26/2001 -0500, John C Klensin wrote:
With FTP, the mail was delivered more or less into the space of
the receiving user. So any conversations that were done (and I
can't remember much, if anything) would have needed to be done
in what we would now call the receiving MUA -- there really
was no _mail_ transport process.
architecturally, the MAIL commands within FTP were identical to
SMTP. They were an email transport protocol.
Both delivermail/Sendmail and MMDF were alive an kicking before
RFC821. The introduction of SMTP did not alter the roles or
basic system processing of either of these applications. It
just added one more transport protocol to their set. That is,
however we would characterize their behaviors now, such as
distinguishing activities within the MUA versus elsewhere -- it
was the same before SMTP.
There were other email processes that worked the same way, but
the only one I know any details about was the MMDF predecessor
that we did at Rand in 1978. My impression is that the Multics
NCP email software had a similar architecture.