ericp@beach.csulb.edu (Eric Pederson CSE) (11/15/90)
I am trying to send a letter to a UUCP-only machine from this machine (beach.csulb.edu) that is on the Internet. I know that the UUCP machine "dest" exchanges mail with a machine "orange.bigcorp.com" that is on the Internet. What address do I use to get to dest!user via orange.bigcorp.com? I have tried orange.bigcorp.com!dest!user (which RFC 976 says "class 3" mailers are supposed to handle) and this got bounced back to me by the local mailer (sendmail under ultrix 4.0 with a seemingly stock sendmail.cf file). The error was "unknown mailer error 101" and it came from uux. The To: header had been re-written "dest!user@orange.bigcorp.com", which looks seemingly ok, but obviously the wrong mailer was chosen since I need to get it to bigcorp.com via STMP. Am I overlooking something obvious, or maybe does our sendmail.cf file need some tweaking? Eric Pederson ericp@beach.csulb.edu
ericp@beach.csulb.edu (Eric Pederson CSE) (11/16/90)
In article <1990Nov15.033214.27771@beach.csulb.edu> I wrote: >I am trying to send a letter to a UUCP-only machine from this machine >... Thanks to all for the responses (and that's SMTP, not STMP, sorry about the typo). I found through experimenting that user%dest.uucp@bigcorp.com worked ok. My only problem now is that mail I recieve from UUCP sites here have a From: header like dest!user@bigcorp.com, which if I try to 'r'eply to, give the uux 101 unknown mailer error again. Obviously my sendmail.cf is going to have to be modified. Does anyone happen to have some sendmail rules that will handle uucpdest!user@knownhost.com by routing to knownhost.com first via SMTP? Eric Pederson ericp@beach.csulb.edu
gary@proa.sv.dg.com (Gary Bridgewater) (11/17/90)
In article <1990Nov15.205028.4567@beach.csulb.edu> ericp@beach.csulb.edu (Eric Pederson CSE) writes: >I found through experimenting that user%dest.uucp@bigcorp.com worked >ok. My only problem now is that mail I recieve from UUCP sites here >have a From: header like dest!user@bigcorp.com, which if I try to 'r'eply >to, give the uux 101 unknown mailer error again. Of course it is no help to you, but, "bigcorp" has a broken .cf file if they are abusing addresses like this. Their outgoing "From:" address rewriter should not be blindly appending their domain address - it should check to see if the current addresses is in "!" form and, if so, either 1) leave it alone, or 2) PRE-pend their UUCP address if they have one. This would leave the address as replyable to. Some number of sites do this awful thing - please stop. >Obviously my sendmail.cf is going to have to be modified. Does anyone >happen to have some sendmail rules that will handle >uucpdest!user@knownhost.com by routing to knownhost.com first via SMTP? Or you can just hand-type over the bogus address. It's hard to answer your question without knowing what kind of an address your site prefers. Are you a UUCP only site or are you connected to the backbone in some manner? Is "knownhost" different from "bigcorp"? Did we go from the pseudo-specific to the pseudo-generic? A UUCP onlu site, when presented with uucpdest!user@knownhost.com address, has to punt this to its smart upstream forwarder via smart!knownhost.com!uucpdest!user. A site connected to the backbone via some intermediate host (like we are, via relay.cs.net) needs to find out from their postmaster what sort of syntax works, e.g. @connectedhost.com:knownhost.com!uucpdest!user. But ask the postmaster (of connectedhost.com) since this is a somewhat religious thing. Smarter hosts and MX based hosts should just be able to send the answer back to bighost.com and let it worry about it since it generated it in the first place - you could try rewriting it as @bighost.com:uucpdest!user or user%uucpdest@bighost.com but that does not generalize to user%uucphop!uucpdest@bighost.com. But the bottom line is - you should not be getting mail with ambiguous From: addresses. Hosts that act as gateways are required to generate consistent addresses. -- Gary Bridgewater, Data General Corporation, Sunnyvale California gary@sv.dg.com or {amdahl,aeras,amdcad}!dgcad!gary C++ - it's the right thing to do.
fitz@wang.com (Tom Fitzgerald) (11/20/90)
> ericp@beach.csulb.edu (Eric Pederson CSE) writes: >> My only problem now is that mail I recieve from UUCP sites here >> have a From: header like dest!user@bigcorp.com, which if I try to 'r'eply >> to, give the uux 101 unknown mailer error again. gary@proa.sv.dg.com (Gary Bridgewater) writes: > Of course it is no help to you, but, "bigcorp" has a broken .cf file if they > are abusing addresses like this. Their outgoing "From:" address rewriter > should not be blindly appending their domain address - it should check to > see if the current addresses is in "!" form and, if so, either 1) leave it > alone, or 2) PRE-pend their UUCP address if they have one. If bigcorp was sending the message over a UUCP link, yes; but in this case the message is going over SMTP to beach.csulb.edu and bigcorp.com has to make it RFC822 compliant before sending it. The easiest way to do this is by appending @bigcorp.com to it. The only way bigcorp could avoid this is if the message arrived with "From: dest.dom.ain!user", in which case bigcorp could rewrite it to "From: user@dest.dom.ain" before passing it on. If there isn't already a domain somewhere in the From: line, bigcorp has to append one. > A UUCP onlu site, when presented with uucpdest!user@knownhost.com address, > has to punt this to its smart upstream forwarder via > smart!knownhost.com!uucpdest!user. Not if it has a set of maps, and the maps contain knownhost.com. In that case it might find a better path to knownhost that doesn't go through 'smart'. --- Tom Fitzgerald Wang Labs fitz@wang.com 1-508-967-5278 Lowell MA, USA ...!uunet!wang!fitz
gary@proa.sv.dg.com (Gary Bridgewater) (11/21/90)
In article <awe40u.g4u@wang.com> fitz@wang.com (Tom Fitzgerald) writes: >gary@proa.sv.dg.com (this is me) writes: >> are abusing addresses like this. Their outgoing "From:" address rewriter >> should not be blindly appending their domain address - it should check to >> see if the current addresses is in "!" form and, if so, either 1) leave it >> alone, or 2) PRE-pend their UUCP address if they have one. > >If bigcorp was sending the message over a UUCP link, yes; but in this case >the message is going over SMTP to beach.csulb.edu and bigcorp.com has to >make it RFC822 compliant before sending it. The easiest way to do this is >by appending @bigcorp.com to it. But this generates an address that will give a naive UUCP site fits. >The only way bigcorp could avoid this is if the message arrived with >"From: dest.dom.ain!user", in which case bigcorp could rewrite it to >"From: user@dest.dom.ain" before passing it on. If there isn't already a >domain somewhere in the From: line, bigcorp has to append one. Actually, they could just leave it alone. But, RFC-976 says: Because of the confusion surrounding hybrid addresses, we recommend that all transport layer software avoid the use of hybrid addresses at all times. A pure bang syntax can be used to disambiguate, being written c.d!a!b in the first case above, and a!c.d!b in the second. We recommend that all implementations use this "bang domain" syntax unless they are sure of what is running on the next machine. In conformance with RFC-822 and the AT&T Message Transfer Architecture, we recommand that any host that accepts hybrid addresses apply the (a!b)@c.d interpretation. >> A UUCP onlu site, when presented with uucpdest!user@knownhost.com address, >> has to punt this to its smart upstream forwarder via >> smart!knownhost.com!uucpdest!user. > >Not if it has a set of maps, and the maps contain knownhost.com. In that >case it might find a better path to knownhost that doesn't go through >'smart'. Sure. I thought that went without saying since the original poster was asking how to deal with such addresses - by munging his .cf file. I keep hoping it will become that simple (I'm a simple minded guy) but sites come on line faster than the speed of knowledge and dealing with their mail is a recurring pain for sites that are gateways. But if, as you say, the answer is to use pathalias then I re-iterate that bighost should just leave "!" paths alone so as to not prejudice any pathalias reply. Or pre-pend its registered UUCP address including its domain and following "!" syntax. Appending an "@" address to a "!" path to somehow "satisfy" RFC-822 knowing full well that leaf sites will choke on it invokes the "Foolish consistency is the hobgoblin of small minds" rule with a vengence. There are just too many of these binary-only, under-documented, just-plain-users sites to continue to hope that new sites will eventually get with it. This is not a knock on them, just a result of the explosive acceptance of unix and/or the usefulness of e-mail. So - to the original poster - get pathalias and build the maps. Get smail3 (or, possibly, deliver) and pass your mail off to one of them. They can be configured to deal with domainish addresses and to look up a path for them. It is probably best to leave your .cf file alone in this case since you would have to do ugly things that will, without doubt, trip up you or your successor in the future. This pushes the delivery problem off onto software designed to to delivery. -- Gary Bridgewater, Data General Corporation, Sunnyvale California gary@sv.dg.com or {amdahl,aeras,amdcad}!dgcad!gary C++ - it's the right thing to do.
shawn@marilyn.UUCP (Shawn P. Stanley) (11/23/90)
In article <659186205.1594@proa.sv.dg.com> gary@proa.sv.dg.com (Gary Bridgewater) writes: >In article <awe40u.g4u@wang.com> fitz@wang.com (Tom Fitzgerald) writes: >>The easiest way to do this is by appending @bigcorp.com to it. >But this generates an address that will give a naive UUCP site fits. Indeed! Actually, I continually run into the following problem. A "From " line is not guaranteed to be an actual return path; rather, it's at best a transmittal route. In several situations I run into, there is at least one site that can't talk back to a site it received mail from. The only lines I can use at that point are "From:" or "Reply-To" lines. However, most messages I receive have no "Reply-To" lines in them. Thus, I'm left with "From:" lines. Yet many messages I receive have "addr3!addr2!spam@addr1.com" or somesuch in the "From:" line, and most are missing a prepended site name from one of my neighbors, whether it's a full "!" path or not. Replying to mail has become a very manual process for me. It's quite frustrating. -- Shawn P. Stanley shawn@marilyn.marilyn.mn.org bungia!marilyn!shawn {rosevax,crash}!orbit!marilyn!shawn
karl_kleinpaste@cis.ohio-state.edu (11/24/90)
The old religious war begins again...
shawn@marilyn.uucp writes:
Indeed! Actually, I continually run into the following problem. A
"From " line is not guaranteed to be an actual return path; rather,
it's at best a transmittal route.
...
Yet many messages I receive have "addr3!addr2!spam@addr1.com" or
somesuch in the "From:" line, and most are missing a prepended site
name from one of my neighbors, whether it's a full "!" path or not.
I really don't want to get into the whole blasted argument again, but
the short form is important.
"addr3!addr2!spam@addr1.com" is a perfectly legitimate address. It is
of a form that an Internet host would prefer to see. It can be
translated into English as
send to addr1.com, who should then send via UUCP (presumed,
that is) to addr3, who should then send via UUCP to addr2, who
will deliver to user "spam," who lives there on addr2.
It's quite unambiguous. Any Internet site can deal with it without so
much as a heart flutter. Any UUCP site with a modern mailer (smail
2.5 is more than sufficient) can deal with it just as easily.
If you're running a system with a Seriously Stupid Mailer (SSM), e.g.,
all you've got is /bin/mail as a transport, and mailx calls it
directly, then you shouldn't be set up to be replying to From: or
Reply-To: lines in the first place. Such SSM configurations can only
be depended upon to understand the UNIX From_ line. The From_ line
will almost certainly contain nothing but a pure !-path, and in the
case of your example above, it will probably look something like
From your-uucp-relay-host!addr1.com!addr3!addr2!spam <date>
or some minor variation on that theme. Your SSM can reply to that
just fine.
Blind prepending of intermediate sitenames in the From: line is widely
viewed as bogus behavior. Site prepending belongs in the From_ line
where SSMs need it, and no one else needs to bother with it.
No, a From: line is not at all guaranteed to be a return path, and it
was never intended as such. Nor should it be a transmittal route. It
should be an _address_, that is, an indication of _who_ you're trying
to reach, not _how_ you're trying to reach them.
The old analogy that applies is that of sending snail mail to someone
across the country or around the world. It is unreasonable and
insulting to expect the writer of a letter to have to provide
indications to the postal system workforce for how to get mail from
Hither to Yon. The writer just puts it in the mail and expects it to
get there. Similar thoughts apply to mail in this environment; a
From: should indicate to a recipient who you are, and not how a mail
carrier finds your street and puts mail in your mailbox.
--karlshawn@marilyn.UUCP (Shawn P. Stanley) (11/24/90)
In article <KARL.90Nov23115753@giza.cis.ohio-state.edu> karl_kleinpaste@cis.ohio-state.edu writes: >The old religious war begins again... I believe you misunderstood my problem. Maybe you could re-read my message and pick up my problem points; I need a solution, not a firm belief that nothing is wrong. >"addr3!addr2!spam@addr1.com" is a perfectly legitimate address. It is >of a form that an Internet host would prefer to see. I can agree to that; no problem. >If you're running a system with a Seriously Stupid Mailer (SSM), e.g., >all you've got is /bin/mail as a transport, and mailx calls it >directly, then you shouldn't be set up to be replying to From: or >Reply-To: lines in the first place. Such SSM configurations can only >be depended upon to understand the UNIX From_ line. Maybe implication wasn't enough. I'm not using a "SSM", unless you call Elm a SSM. I implied I have control over how my software can interpret headers, and so I do. >The From_ line will almost certainly contain nothing but a pure !-path, >and in the case of your example above, it will probably look something like > From your-uucp-relay-host!addr1.com!addr3!addr2!spam <date> >or some minor variation on that theme. Your SSM can reply to that >just fine. As I explained, it can't. There is at least one site in the address that can get toward me but not backward from me. For instance, a message can get to me from "addr1!sialis!pwcstoo!addr2...", but when sent back along that path, pwcstoo informs me it can't send to addr2. >Blind prepending of intermediate sitenames in the From: line is widely >viewed as bogus behavior. Site prepending belongs in the From_ line >where SSMs need it, and no one else needs to bother with it. >No, a From: line is not at all guaranteed to be a return path, and it >was never intended as such. Nor should it be a transmittal route. It >should be an _address_, that is, an indication of _who_ you're trying >to reach, not _how_ you're trying to reach them. Okay. >It is unreasonable and insulting to expect the writer of a letter to >have to provide indications to the postal system workforce for how to >get mail from Hither to Yon. Is that what I said? I don't believe so. I didn't realize the "From:" line was considered an unchangable part of the message, but now I do. What I WISH is that the sites along the way would take care of how I can get back to the author. But even that may seem unreasonable. And it probably is. Sigh... I guess I'm going to have to somehow start handling maps, eh? Or write myself some software to handle the specific cases that bother me the most. -- Shawn P. Stanley shawn@marilyn.marilyn.mn.org bungia!marilyn!shawn {rosevax,crash}!orbit!marilyn!shawn