jww@bonnie.UUCP (Joel West) (05/23/85)
A quick question for the gurus out there: What do the two "From" headers at the beginning of a mail message signify? Below are excerpts from my sample case (note redundant cbosgd in #7 (sample #1) From uucp Wed Feb 6 12:50 EST 1985 >From cbosgd!decvax.UUCP!genrad!charlie Wed Feb 6 12:50:03 1985 remote from clyde (sample #7) From uucp Fri May 10 01:07 EDT 1985 >From cbosgd.ATT.UUCP!db Fri May 10 01:07:10 1985 remote from cbosgd From: db@cbosgd.ATT.UUCP (Dave Bursik) (sample #8) From uucp Fri May 10 06:20 EDT 1985 >From uw-beaver!tektronix!davidl@orca Fri May 10 06:20:43 1985 remote from ihnp4 From: ihnp4!uw-beaver!tektronix!davidl@orca (sample #9) From uucp Mon May 13 05:23 EDT 1985 >From ut-sally!ut-ngp!werner Mon May 13 05:23:25 1985 remote from ihnp4 In particular, isn't the >From line redundant to an honest-to-goodness RFC822-std From: line (which I assume has been assaulted by each and every mailer along the way). Is it necessary to include both of these first two "From" lines (I'm writing my own mailer to drop mail at a sendmail site, so the question is not academic). And boy, the "remote from" is ugly! :-). is not academic.) -- Joel West (619) 457-9681 CACI, Inc. - Federal 3344 N. Torrey Pines Ct La Jolla 92037 jww@bonnie.UUCP (ihnp4!bonnie!jww) westjw@nosc.ARPA "The best is the enemy of the good" - A. Mullarney
avolio@decuac.UUCP (Frederick M. Avolio) (05/24/85)
In article <487@bonnie.UUCP>, jww@bonnie.UUCP (Joel West) writes: > What do the two "From" headers at the beginning of a mail message signify? > ... > In particular, isn't the >From line redundant to an honest-to-goodness > RFC822-std From: line (which I assume has been assaulted by each and > every mailer along the way). Is it necessary to include both of these > first two "From" lines (I'm writing my own mailer to drop mail at a > sendmail site, so the question is not academic). Well, From_ is for UUCP mailing/routing. From: should be the "ultimate" address of the person. Strictly speaking, it should be in a form such as "User@Domain" as in you sample #7 (which was "user@cbosgd.ATT.UUCP"). (RFC822-std.) So you are quite right -- UUCP mailers all over are "assaulting" that line. That line should never be rewritten. The way things *should* work is if you have a smart mailer, it uses the From: line. Since most UUCP mailers are not smart, they should use the From_ line. (But they do not always do this -- I don't think UCB Mail does, for example.) By the way, I think is is a mistake for a system to be tacking on an extra "From_" line (From uucp .....) as it appears yours is. At the least it is redundant. -- Fred Avolio {decvax,seismo}!decuac!avolio 301/731-4100 x4227
hokey@plus5.UUCP (Hokey) (05/26/85)
In article <487@bonnie.UUCP> jww@bonnie.UUCP (Joel West) writes: > (sample #7) >From uucp Fri May 10 01:07 EDT 1985 >>From cbosgd.ATT.UUCP!db Fri May 10 01:07:10 1985 remote from cbosgd >From: db@cbosgd.ATT.UUCP (Dave Bursik) There is no redundant cbosgd in the >From line. The first cbosgd is part of a domain address, and can only be sent directly to sites which know how to handle it. By prepending the domain name with the name of a site which *can* handle the domain name (in this case, the site happens to be the same one as that represented by the domain name), a good route is preserved. >In particular, isn't the >From line redundant to an honest-to-goodness >RFC822-std From: line (which I assume has been assaulted by each and >every mailer along the way). Is it necessary to include both of these >first two "From" lines (I'm writing my own mailer to drop mail at a >sendmail site, so the question is not academic). The >From lines trace a *route*, while the From: line is (almost always) an *address*. If you are invoking sendmail directly, you can only pass a single "From " line, so you will have to "crush" all >From lines into a single line. The biggest problem with this is that timestamp information is lost. Someday I hope to write some code to move the timestamp information into a Received: line (or something similar). (One must take pains to avoid creating duplicate Received: lines.) -- Hokey ..ihnp4!plus5!hokey 314-725-9492
honey@down.FUN (Peter Honeyman) (05/27/85)
regarding fred's assertion that "UUCP mailers all over are 'assaulting' [the From:] line," let's get our facts straight. uucp never ever ever looks at the body of a mail message. the culprits are those new improved cute-as-a-shithouse mailers like sendmail and smail, that resolve to foist their vile inconsistencies on the unsuspecting world. fear not, better ideas await in the wings. peter
honey@down.FUN (Peter Honeyman) (05/27/85)
regarding hokey's assertion that "[t]here is no redundant cbosgd in [cbosgd!cbosgd.ATT.UUCP!db] line," this is transparent doublespeak. smail, in an attempt to eliminate operator precedence ambiguity, is using overkill. there is no logical reason to name cbosgd twice in the route, just a lack of vision on the authors' parts. i see no contribution to robust mail routing here. peter
hokey@plus5.UUCP (Hokey) (05/30/85)
As entities with some intelligence, we can see that cbosgd!cbosgd.ATT.UUCP!db lists the same machine twice. However, I have not yet seen a sendmail.cf file which could determine this "fact". Several configuration files already assume that the second level domain is the site name, and thus try to send mail to the ATT machine. This is a problem with naming, not operator precedence. The recently posted uk.cf stuff handles domain names (I think, I haven't installed it yet, due to a problem with operator precedence and some missing connectivity information in my setup files). -- Hokey ..ihnp4!plus5!hokey 314-725-9492
gregbo@houxm.UUCP (Greg Skinner) (05/30/85)
I think we are seeing the From cbosgd.ATT.UUCP!user remote from cbosgd phenomenon because Mark is running sendmail RFC822 style and putting out mess- ages which only an RFC822 site can interpret. Any non-sendmail site which picks up a message from cbosgd will apply the UGLYUUCP rules to it, resulting in what we see above. Perhaps Mark could change his sendmail.cf to use the uucp mailer for hosts not in his RFC822 domain? -- It's like a jungle sometimes, it makes we wonder how I keep from goin' under. Greg Skinner (gregbo) {allegra,cbosgd,ihnp4}!houxm!gregbo gregbo%houxm.uucp@harvard.arpa
arash@jendeh.UUCP (Arash Farmanfarmaian) (05/31/85)
> regarding fred's assertion that "UUCP mailers all over are 'assaulting' > [the From:] line," let's get our facts straight. uucp never ever ever > looks at the body of a mail message. the culprits are those new > improved cute-as-a-shithouse mailers like sendmail and smail, that > resolve to foist their vile inconsistencies on the unsuspecting world. > fear not, better ideas await in the wings. > > peter [honeyman, princeton!down!peter] Oh fudge Peter. I don't see anything wrong with sendmail. Complaining about EMACS based mailers is one thing ( a problem you have fixed with that 'cute' and nasty message of yours about screwed-up mailers emanating from rmail ( remember the '#define bitch' in rmail.c !!)) but complaining about sendmail is another. Sendmail works and 'your' UUCP has never complained about it! And we are not even running the standard 'sendmail', but a hacked version of it. So please lay off. Arash Farmanfarmain'85 P.S. Please disregard this message until you have corrected my final exam. -- Arash Farmanfarmaian ...!allegra!princeton!jendeh!arash "Any man who lives within his means suffers from a lack of imagination"
arash@jendeh.UUCP (Arash Farmanfarmaian) (05/31/85)
> regarding hokey's assertion that "[t]here is no redundant cbosgd in > [cbosgd!cbosgd.ATT.UUCP!db] line," this is transparent doublespeak. > smail, in an attempt to eliminate operator precedence ambiguity, is > using overkill. there is no logical reason to name cbosgd twice in the > route, just a lack of vision on the authors' parts. i see no > contribution to robust mail routing here. > peter You know Peter, considering the number of times I have bitched at you you tonight, you should be happy to learn that for once I agree with you. And this is not the only time 'smail' uses overkill. Some people just like being on the safe side a bit too much. Reminds me of an IBM executive that once said that 'a little bit of conformism never hurts!' Makes you wanna puke. Arash Farmanfarmaian'85 -- Arash Farmanfarmaian ...!allegra!princeton!jendeh!arash "Any man who lives within his means suffers from a lack of imagination"
henry@utzoo.UUCP (Henry Spencer) (06/02/85)
> ... Oh fudge Peter. I don't see anything wrong with sendmail. ...
You are probably the only sendmail maintainer in the world with this view.
--
Henry Spencer @ U of Toronto Zoology
{allegra,ihnp4,linus,decvax}!utzoo!henry
gnu@sun.uucp (John Gilmore) (06/03/85)
> regarding hokey's assertion that "[t]here is no redundant cbosgd in > [cbosgd!cbosgd.ATT.UUCP!db] line," this is transparent doublespeak. > ...there is no logical reason to name cbosgd twice in the route... > peter Peter is correct; there is redundancy here. The first cbosgd! is for dumb mailers. If they need to send a reply back, they will just pull off the first name and ! and route the message there. The cbosgd.att.uucp part is for smart mailers. The convention is that a string between ! chars which contains a period is a fully-qualified domain based address. Whenever a smart uucp mailer sees this address, it can discard all the leading host!host!host! stuff and just use the domain based address. The reason both are there is that a smart mailer can't tell that "cbosgd!db" is the same as "cbosgd.att.uucp!db", since there may be many machines named "cbosgd" in different domains. [If the smart mailer is a neighbor of cbosgd's, it will know the full name of its neighbor, but if the message has gone through some dumb intermediate sites, it has no way to know.]
gjm@ihnp4.UUCP (Gary J. Murakami) (06/03/85)
> > ... Oh fudge Peter. I don't see anything wrong with sendmail. ... > > You are probably the only sendmail maintainer in the world with this view. I agree with Peter and Henry on sendmail -- sendmail has major problems. However I commend Eric on giving us the functionality available in sendmail, especially in such a controversial no-win subject as mail. But it's time to move on to the next generation of mailers. Too bad something better isn't available to public ... yet. -Gary (a sometimes sendmail maintainer by duress)
jim@hwcs.UUCP (Jim Crammond) (06/10/85)
In article <761@plus5.UUCP> hokey@plus5.UUCP (Hokey) writes: >As entities with some intelligence, we can see that cbosgd!cbosgd.ATT.UUCP!db >lists the same machine twice. However, I have not yet seen a sendmail.cf >file which could determine this "fact". I can't think of anyway that sendmail can compare two arbitrary names and strip one off if they are the same. Thus this must be dealt with in the rmail front-end. Once uucp headers like "From cbosgd.ATT.UUCP!db ... remote from cbosgd" become "standard" then rmail should probably ignore the "remote from system" part (except to set the sender macro) and just pass the user part of the header to sendmail as the "from" address. If the first part of the user address (i.e. the bit before the first '!') does not contain any domain seperators (i.e. '.'s) then the old approach of prepending the system name to the user part can be done. -Jim. -- ------------- -Jim Crammond jim@hwcs.uucp