[net.mail.headers] "From " vs. "From:" in Sendmail

gnu@sun.uucp (John Gilmore) (02/24/84)

	From: eric%ucbarpa@Berkeley.ARPA (Eric Allman)
	Subject: Re: "Return-Path" vs. "From"

	Gosh Chris, I'm awfully sorry that I have tried to do the right
	thing instead of perpetuating the insanity that has characterized
	(and sadly, continues to characterize) the UNIX mail system.

	Sendmail passes the information it gets from the received envelope
	into the transmitted envelope.  If the mailer is a program rather
	than an SMTP channel, the "envelope" consists of the command line
	arguments.

Unfortunately the information provided by uucp in the "envelope" is
often wrong.  This causes the "From " line and the "From:" line to
differ, which is very confusing to the end-users, mail-readers, etc.

For mail generated on the Arpanet and relayed to Sun via uucp, the
From: line is right (or at least can be munged to be right, by adding
".arpa"), but the "From " line often claims the message is from
"daemon" or "uucp" or whatever user happened to do something that
started this uucp connection.  E.g.:
	From uucp Thu Feb 23 00:27:47 1984
	From: Ed Pattermann <PATTERMANN@SUMEX-AIM.ARPA>
	To: "sun!gnu@shasta"@SUMEX-AIM.ARPA

For mail which has been passed solely thru uucp sites, the "From " line
is clearly correct, since the "From:" line has not been maintained at
(all) intermediate sites.  E.g.:
	From dual!allegra!cbosgd!fair Thu Feb 23 17:01:34 1984
	From: dual!cbosgd!fair

The win would have been to make the envelope right for all uucp
traffic, (which it already was for strictly-uucp traffic) and have
sendmail fix the From: line to match the "From " line.  This has not
been done -- yet.  Is there a uucp/mail expert in the house who can tell
us WHY the envelope is "uucp" or some user, instead of the right thing?

fair@ucbvax.UUCP (Erik E. Fair) (02/26/84)

Uh, I have a small nit to pick with John Gilmore. The example he used
for showing the proper interpretation of `correct' mail headers from
the UUCP land was wrong:

>>For mail which has been passed solely thru uucp sites, the "From " line
>>is clearly correct, since the "From:" line has not been maintained at
>>(all) intermediate sites.  E.g.:
>>	From dual!allegra!cbosgd!fair Thu Feb 23 17:01:34 1984
>>	From: dual!cbosgd!fair
>>
>>The win would have been to make the envelope right for all uucp
>>traffic, (which it already was for strictly-uucp traffic) and have
>>sendmail fix the From: line to match the "From " line.  This has not
>>been done -- yet.  Is there a uucp/mail expert in the house who can tell
>>us WHY the envelope is "uucp" or some user, instead of the right thing?

I have been many things in my time, but never have I been cbosgd!fair.
I'm dual!fair or ucbvax!fair or fair@ucb-arpa.ARPA. The problem is that
some 4.2 BSD sites out there are running bogus versions of sendmail which
puts out the return address in the envelope in domain format. It gets
worse when the letter passes through a working sendmail sites, because
they try and undo the damage, and only succeed in hiding it. That letter
there started life

	From dual!fair

cbosgd changed that to

	From cbosgd!fair@dual.UUCP

and allegra changed that to 

	From dual!allegra!cbosgd!fair

It gets really fun when someone tries to reply to that. This bug
has been fixed by steve@nbires.UUCP, tjt@masscomp.UUCP,
postmaster@uwisc-rsch.ARPA, and eric@BERKELEY.ARPA
Some by my prodding them, some not.

If you spot one of these, complain until it gets fixed, because
all mail passing through the suspect site will have bogus return
addreses, and only the silly people who keep the entire routing map in
their heads (like me) will be able to get anything through the Emails.
(Mark Horton, are you listening? Do I have to beat this to death?)

	Please, Help Stamp Out Sendmail Errors by sending your
	tax deductible contributions to 

		Save the Network
		Box 10.2.0.78
		Berkeley, California 94720

Thank you.

	Erik E. Fair	ucbvax!fair	fair@ucb-arpa.ARPA

P.S.	For those on USENET: 10.2.0.78 is ucbvax's ARPANET address. :-)

chris@umcp-cs.UUCP (02/27/84)

Well, on our system the uucp "envelope" info is usually wrong
because uux uses getpwuid(getuid())->pw_name, but it is invoked
on behalf of the fake user "mmdf" when MMDF dequeues UUCP channel
mail.
-- 
In-Real-Life: Chris Torek, Univ of MD Comp Sci
UUCP:	{seismo,allegra,brl-bmd}!umcp-cs!chris
CSNet:	chris@umcp-cs		ARPA:	chris.umcp-cs@CSNet-Relay

tjt@kobold.UUCP (02/27/84)

For anyone who is interested:

I fixed the sendmail bug described by Erik Fair (sending out domain
based address in the unix-style "From " line) by switching from the
4.1c version of sendmail to the 4.2 version.  Looking at the Version.c
file indicated that this was explicitly fixed and is not an artifact of
an aberrant configuration file.
-- 
	Tom Teixeira,  Massachusetts Computer Corporation.  Westford MA
	...!{ihnp4,harpo,decvax}!masscomp!tjt   (617) 692-6200 x275

gjm@ihnp4.UUCP (Gary J. Murakami) (02/28/84)

Last time I looked at cbosgd's config file, my analysis agreed with Erik
Fair's analysis of the situation.  Empirical evidence seems to point out
that the path reversal problem still exists at more than one host.  Part
of the problem stems from the inconsistent treatment of a!b@c. 
Unfortunately the 4.2 distribution contains config files that do a
considerable amount of header munging, and more hacking will probably
make matters worse.

Sendmail is still a mixed blessing, however some of us hope to
eventually fix the (few) cursed parts.

Gary Murakami
AT&T Bell Laboratories
ihnp4!gjm

tjt@kobold.UUCP (02/29/84)

Although I missed Rob Pike's "cat -v considered harmful" talk at the
Toronto USENIX conference, I understood he mentioned that the current
trend in arcane command switches would no double lead to an extension
language for cat. It occurred to me the other day after wrestling with
configuration files that sendmail must be the aforementioned extension
language (where is TECO now that we need you?).

-- 
	Tom Teixeira,  Massachusetts Computer Corporation.  Westford MA
	...!{ihnp4,harpo,decvax}!masscomp!tjt   (617) 692-6200 x275