vixie@palo-alto.DEC.COM (Paul Vixie) (08/05/88)
R U T G E R S
brisco@pilot.njin.net (Thomas Paul Brisco) writes:
# In fact, there is nothing wrong will letting your MTA do all
# the routing for you. I've had very few sites bouncing back to me,
# and I _usually_ route through rutgers.
This depends on what you mean by "all your routing".
Rutgers does it WRONG. BIG TIME EL WRONGO. Always has, always will.
If I send to <...!rutgers!foo!bar!person>, and rutgers looks in its map
database and says "oh, site <<<bar>>> can be reached through ...!baz!bar!user",
IT HAS JUST DONE THE WRONG THING.
<baz!bar!user> is never nec'ily the same person as <foo!bar!user>. <<<bar>>>
can be contextual: <baz!bar> and <foo!bar> aren't nec'ily the same machine.
This is not "routing", it's <<<rerouting>>> and it's EVIL and RUDE.
Now, if I send mail to <rutgers!bar!user>, then rutgers is free to send to
whatever entity it calls "bar", whether a direct UUCP neighbor or an indirect
neighbor reachable only through a routing database. This is "routing" and
it's very convenient. Many sites do this. Smail does this (by default,
unless misconfigured with a stupid option that shouldn't be in there).
Rule of thumb:
Don't look for a route in your UUCP map database / pathalias database
unless you are about to throw the message away out of unreachability
of the "next host" in the source route.
S U N
Sun thinks it owns the world. I send mail from my home machine (vixie.uucp)
with a From: line like this:
From: vixie!paul
or like this:
From: paul@vixie.UUCP (this is ugly and bad)
If this message passes through pacbell, or decwrl, or ubvax, or uunet, or
half a dozen other major sites within one or two hops of vixie.UUCP, the
message is forwarded on with a From: line that looks like this:
From: vixie!paul
Note that it's the same as I left it. Note that the From_ line (sometimes
called the UUCP From_ line) is growing and probably looks like:
From pacbell!vixie!paul <date>
Which is okay. But if this message goes through Sun.COM, it will be forwarded
on with a From: line like this:
From: vixie!paul@Sun.COM
This is EVIL and RUDE. Sun doesn't talk to vixie.UUCP; replies to the
message are broken, Sun.COM bounces things that come to it looking like:
To: vixie!paul@Sun.COM
The mistake? Sun is rewriting HEADER sender addresses; it's only supposed
to rewrite ENVELOPE sender addresses unless the message is passing into the
internal network (inside Sun).
Paul
--
Paul Vixie
Digital Equipment Corporation Work: vixie@dec.com Play: paul@vixie.UUCP
Western Research Laboratory uunet!decwrl!vixie uunet!vixie!paul
Palo Alto, California, USA +1 415 853 6600 +1 415 864 7013bill@ssbn.WLK.COM (Bill Kennedy) (08/05/88)
I won't repeat what Paul Vixie said (substantially, I agree) but I'd like
to know how a message through sun gets a ^M appended to every line. I
get some from a site, not through sun, no ^M, send the same note through
sun, ^M every time.
Paul says it's OK to open the envelope to change the path information,
fair enough. How come they have to scribble on the contents of the rest
of the message?
--
Bill Kennedy usenet {killer,att,rutgers,sun!daver,uunet!bigtex}!ssbn!bill
internet bill@ssbn.WLK.COMdavid@ms.uky.edu (David Herron -- One of the vertebrae) (08/05/88)
hmmm ... the Rutgers way of doing things wrong and the Sun way of doing things wrong. Maybe there's a connection here because Rutgers has a lot of Sun equipment as I understand it ... ? Nevermind ... In article <3674@palo-alto.DEC.COM> vixie@palo-alto.DEC.COM (Paul Vixie) writes: > > R U T G E R S > >brisco@pilot.njin.net (Thomas Paul Brisco) writes: ># In fact, there is nothing wrong will letting your MTA do all ># the routing for you. I've had very few sites bouncing back to me, ># and I _usually_ route through rutgers. > >This depends on what you mean by "all your routing". > >Rutgers does it WRONG. BIG TIME EL WRONGO. Always has, always will. > >If I send to <...!rutgers!foo!bar!person>, and rutgers looks in its map >database and says "oh, site <<<bar>>> can be reached through ...!baz!bar!user", >IT HAS JUST DONE THE WRONG THING. > ><baz!bar!user> is never nec'ily the same person as <foo!bar!user>. <<<bar>>> >can be contextual: <baz!bar> and <foo!bar> aren't nec'ily the same machine. how very true. It isn't done here, I don't know if rutgers does it or not. But you are right, it is the "wrong" thing to do. On the other hand I don't think you're taking it the way brisco@pilot.njin.net meant it. The only excuse I've ever heard for this is that "replies along news paths are always sub-optimal", never mind that you're not ever supposed to use news paths for replies ... NOTE that if 'bar' above were a full.domain.name then it's legit to re-route because full.domain.name's are unambiguous whereas unadorned uucp names in a uucp route *are* ambiguous. It seems appropriate to say somewhere in here that the only sites which have the right to use user@host.uucp are the ones who appear in the maps. > S U N > >Sun thinks it owns the world. I send mail from my home machine (vixie.uucp) >with a From: line like this: > > From: vixie!paul > >or like this: > > From: paul@vixie.UUCP (this is ugly and bad) [Most sites leave it alone, in one of those two forms ... oh, and why is the pseudo-domain version ugly? It's better than the bang form as I see it, that is it will be interpretable by non Unix machines.] [The From_ line is being added to as the message passes through the net, as described in both rmail source code and rfc976.] [Sun rewrites the header to give a reference relative to some internet host rather than leaving it as a bare thing which might or might not be known at the recipients machine ... namely, > From: vixie!paul@Sun.COM ] >This is EVIL and RUDE. Sun doesn't talk to vixie.UUCP; replies to the >message are broken, Sun.COM bounces things that come to it looking like: > > To: vixie!paul@Sun.COM In this case EVIL and RUDE are much too strong. It *is* EVIL and RUDE that Sun is munging the header and then NOT accepting the form into which it munges the header. But it is not such a bad thing to munge headers, especially when the message is going to be passing from a network where the addressing style works into one where the addressing style doesn't work or is only supported part of the time. A case in point which worked for many many years. The mailer at relay.cs.net used to accept mail from csnet members and then rewrite the header of the message to use the %-hack. That is they would end up with something like: david%uky.csnet@csnet-relay.arpa or david%ms.uky.edu@relay.cs.net (after domain conversion) Why did they do this? Well, in the beginning it was because a lot of the mailers on the arpanet wouldn't have been able to handle "david@uky.csnet" directly. But they *could* handle "something@csnet-relay.arpa" because csnet-relay.arpa is in HOSTS.TXT and can easily receive mail. If they were thinking of poor UUCP people at all they surely figured that the UUCP person would know of a uucp<->arpa gateway and use that to reach csnet-relay.arpa. Later on a lot of the csnet members acquired proper domain names but weren't directly connected to the Internet. MX records were advertised for these people, but not all mailers knew how to handle MX records. Again the %-hack comes to the rescue. Nowadays I think that relay.cs.net has stopped doing the %-hack for sites which have domain names, but still do for the few who do not. What does this have to do with vixie!paul@Sun.COM? A WHOLE LOT! The '!' thingie in there is very much the same thing as the %-hack, just in a slightly different form. Some random Internet site, say psuvm.psu.edu which I happen to know is an IBM mainframe, might not (probably) have any notion of what paul@vixie.uucp might mean. I know from personal painful experience that the mailer in the IBM TCP/IP package cannot use ANY host names which aren't in HOSTS.TXT *unless* the site has plunked down the many thousands of dollars they need to have a copy of SQL on site. They *only* way which that mailer will be able to have a chance of handling paul@vixie.uucp is *if* that data-base manager is on site. That's an example. All the reasons I gave above for csnet-relay using the %-hack also apply for Sun rewriting the headers as Paul says they do. BUT BUT BUT (I'm reminded here of Dr. Strangelove at the end of the movie when he's asking the Russian Ambassador *why* they would create such a good deterrent without *telling*anyone*) the employing [!%]-hack works a *whole*lot*better* if you support it coming back. [In fact, I'm really surprised that the !-hack isn't supported at Sun, it'd be such a trivial thing to put into a sendmail configuration, and is directly supported in MMDF]. >The mistake? Sun is rewriting HEADER sender addresses; it's only supposed >to rewrite ENVELOPE sender addresses unless the message is passing into the >internal network (inside Sun). NO. gateway sites are supposed to do appropriate header munging so that sites on both sides of the gateway can understand what the other side is saying. But like any language translation excercise it is non-trivial to do in practice and is a whole lot easier when both sides speak almost the same language. -- <---- David Herron -- The E-Mail guy <david@ms.uky.edu> <---- ska: David le casse\*' {rutgers,uunet}!ukma!david, david@UKMA.BITNET <---- <---- Looking forward to a particularly blatant, talkative and period bikini ...
wisner@killer.DALLAS.TX.US (Bill Wisner) (08/05/88)
Some sites believe that ANYTHING coming in via UUCP must have a bang path. For example, my outgoing mail leaves killer with this header. From: Bill Wisner <wisner@killer.dallas.tx.us> If I send mail to a mailing list, and get a copy back, it often looks like this: From: Bill Wisner <foo!bar!gronk!killer.dallas.tx.us!wisner> As far as I'm concerned, this is EVIL and RUDE. killer.dallas.tx.us is a perfectly valid domain name, known to UUCP routers and to MX mailers. IT SHOULD *NOT* BE CONVERTED INTO A BANG PATH! And while I'm bitching about misbehaving mailers, remind me to tell you about mcvax sometime..
rsalz@bbn.com (Rich Salz) (08/05/88)
. oh, and why is . the pseudo-domain version [pual@vixie.uucp] ugly? The question contains its answer; .UUCP is not a domain. You do not help The Cause by putting out illegal addresses. [ Sun turns vixie!paul into vixie!paul@sun.com; replies to that address will bounce. This has been called Evil and Rude.] . It *is* EVIL and RUDE that .Sun is munging the header and then NOT accepting the form into which it .munges the header. Exactly. Sun is *TRASHING* mail addresses. If they acted as a responsible gateway you could say that they are MUNGING headers. But they don't, they rewrite everything and won't accept what they put out. Either be a full gateway, or leave that game to the Big Boys (harvard, rutgers, uunet, etc). . I know from personal .painful experience that the mailer in the IBM TCP/IP package cannot use .ANY host names which aren't in HOSTS.TXT *unless* the site has plunked .down the many thousands of dollars they need to have a copy of SQL on .site. Well, I feel pity for the users on that site, but such is the price of progress. Any host that still relies on HOSTS.TXT is so woefully behind the times that I gotta wonder if they could possibly have anything worthwhile to say to me (semi :-). As for "-d rutgers" well, I have no problem with Rutger's policies. /rich $alz -- Please send comp.sources.unix-related mail to rsalz@uunet.uu.net.
gore@eecs.nwu.edu (Jacob Gore) (08/05/88)
/ david@ms.uky.edu (David Herron -- One of the vertebrae) / Aug 4, 1988 / > >What does this have to do with vixie!paul@Sun.COM? A WHOLE LOT! The >'!' thingie in there is very much the same thing as the %-hack, just in >a slightly different form. Some random Internet site, say psuvm.psu.edu >which I happen to know is an IBM mainframe, might not (probably) have >any notion of what paul@vixie.uucp might mean. That's all fine and dandy, if you stay out of the UUCP mailers. But when a message from "vixie!paul@Sun.COM" winds up in the UUCP world, it's quite likely to be interpreted as vixie-->Sun.COM-->mailbox paul instead of as Sun.COM-->vixie-->mailbox paul So, if sun does insist on performing this "service" (which, apparently, not everybody wants anyway), they should use "paul%vixie@Sun.COM" instead. Seems to me, though, that at this stage in the game, sites that play by the rules should expect sites that don't to bend over backwards to assure connectivity, instead of it being the other way around. Any reason why a machine like psuvm.psu.edu can't refer addresses it can't handle to a machine that can? Jacob Gore Gore@EECS.NWU.Edu Northwestern Univ., EECS Dept. {oddjob,gargoyle,att}!nucsrl!gore
wcf@psuhcx.psu.edu (William C. Fenner) (08/07/88)
In article <3400003@eecs.nwu.edu> gore@eecs.nwu.edu (Jacob Gore) writes: |/ david@ms.uky.edu (David Herron -- One of the vertebrae) / Aug 4, 1988 / |> |>What does this have to do with vixie!paul@Sun.COM? A WHOLE LOT! The |>'!' thingie in there is very much the same thing as the %-hack, just in |>a slightly different form. Some random Internet site, say psuvm.psu.edu |>which I happen to know is an IBM mainframe, might not (probably) have |>any notion of what paul@vixie.uucp might mean. Actually I think it has a slight notion. | |Seems to me, though, that at this stage in the game, sites that play by the |rules should expect sites that don't to bend over backwards to assure |connectivity, instead of it being the other way around. Any reason why a |machine like psuvm.psu.edu can't refer addresses it can't handle to a |machine that can? Nope. Actually I think it might... psuvax1.cs.psu.edu knows (or thinks it knows) all about all the UUCP sites there are... so there's no reason why it shouldn't (except that IBM writes drain-bamaged software) Bill
robert@setting.weitek.UUCP (Robert Plamondon) (08/08/88)
In article <3400003@eecs.nwu.edu> gore@eecs.nwu.edu (Jacob Gore) writes: >But when a message from "vixie!paul@Sun.COM" winds up in the UUCP world, >it's quite likely to be interpreted as > vixie-->Sun.COM-->mailbox paul >instead of as > Sun.COM-->vixie-->mailbox paul Not only that, but Sun's mailer is very good at ignoring the hostname part and delivering the mail to some local user with the same login name. robert@regulus.Sun.COM has occaisionally been plagued with masses of mail meant for me (weitek!robert). Sending mail through sun is a good way to ensure that people you don't know will read it. -- Robert Plamondon robert@weitek.COM pyramid!weitek!robert "You canna mix matter and antimatter cold"
jc@minya.UUCP (John Chambers) (08/09/88)
> The only excuse I've ever heard for this is that "replies along news paths > are always sub-optimal", never mind that you're not ever supposed to use > news paths for replies ... You're wrong there. For instance, try responding by email to this article. You will use a news path. Why? Well, this machine has 4 neighbors, and the news (e.g., this article) is sent to all of them. There is therefore no mail path to this machine that is not a reverse news path, and if you respond successfully, you will have used a news path. Q.E.D. > It seems appropriate to say somewhere in here that the only sites which > have the right to use user@host.uucp are the ones who appear in the maps. Huh? I'll use any mail notation I can get my mailer to accept; it's my machine (;-). How are you going to tell what notation I used, anyway? If my mailer converts the original notation to some canonical form, like host.uucp!user, there's no way you can reconstruct what I typed, and I'll have violated your rule with impunity. > ... many lines deleted... > [Sun rewrites the header to give a reference relative to some internet > host rather than leaving it as a bare thing which might or might not > be known at the recipients machine ... namely, > > > From: vixie!paul@Sun.COM > > >This is EVIL and RUDE. Sun doesn't talk to vixie.UUCP; replies to the > >message are broken, Sun.COM bounces things that come to it looking like: It is something even more sinful in a computing environment: it is ambiguous. One of the truly general rules is that you should never mix ! notation with @ notation. Such a mail path inherently has two valid meanings, and you are trusting the mailer to pick the right one. Anyone who's successfully written even one program knows how likely it is that the little monster will pick the meaning you intended. (I've wasted a lot of my time illuminating email users as to the consequences of such a path. Most of them understand, and wonder aloud how the email community allows such silliness to continue.) Isn't email fun? It's even more fun that theology! [Though sometimes I wonder if they aren't converging. ;-] -- John Chambers <{adelie,ima,maynard,mit-eddie}!minya!{jc,root}> (617/484-6393) [Any errors in the above are due to failures in the logic of the keyboard, not in the fingers that did the typing.]
david@ms.uky.edu (David Herron -- One of the vertebrae) (08/09/88)
In article <60@minya.UUCP> jc@minya.UUCP (John Chambers) writes: >> The only excuse I've ever heard for this is that "replies along news paths >> are always sub-optimal", never mind that you're not ever supposed to use >> news paths for replies ... > >You're wrong there. For instance, try responding by email to this article. >You will use a news path. Why? Well, this machine has 4 neighbors, and >the news (e.g., this article) is sent to all of them. There is therefore >no mail path to this machine that is not a reverse news path, and if you >respond successfully, you will have used a news path. Q.E.D. > Just because you were careful and made sure that all your news neighbors are also mail neighbors doesn't mean that will ALWAYS be the case. One way it will not work is if the article happens to pass through the link between here and psuvm.bitnet. THe Path: line will be something like: Path: ...!ukma!psuvm.bitnet!...!user When it arrives here rmail will see "psuvm.bitnet!...!user" and will attempt to route to "...!user@psuvm.bitnet". Now, it will succeed in doing so since we're directly connected to BITNET and have up-to-date maps of bitnet and such. However since the lowest common denominator of mail on bitnet (which is all I've had the energy to support so far) limits the local-part to 8 characters, this will fail a lot. (and does fail a lot from the number of mail messages which bounce into my mailbox). (er.. 1 or 2 times a month anyway) In general there are a number of links on the net which are not accompanied by UUCP mail links. One of the older examples is/was the sites at brl. Also you should go and check the news installation documents, rfc976, and whatever the current rfc for the news format. The news documents EXPLICITLY say that you should not ever use Path: lines for replies. And have said that for as long as I can remember (my experience with this stuff dates back to v2.10.1 beta). >> It seems appropriate to say somewhere in here that the only sites which >> have the right to use user@host.uucp are the ones who appear in the maps. >Huh? I'll use any mail notation I can get my mailer to accept; it's my >machine (;-). How are you going to tell what notation I used, anyway? >If my mailer converts the original notation to some canonical form, like >host.uucp!user, there's no way you can reconstruct what I typed, and I'll >have violated your rule with impunity. I meant in their outgoing headers, silly Oh, I know full well that "a!b@d.dom" is ambiguous ... I also know that the mailer here generates headers which say that when it's doing rewriting. One of my backburner projects is to do a rewrite of the UUCP channel in MMDF and that's one of the things I may touch on. Perhaps put in a routine to do a mapping to "b%a@d.dom". -- <---- David Herron -- The E-Mail guy <david@ms.uky.edu> <---- ska: David le casse\*' {rutgers,uunet}!ukma!david, david@UKMA.BITNET <---- <---- Looking forward to a particularly blatant, talkative and period bikini ...
page@swan.ulowell.edu (Bob Page) (08/11/88)
[isn't this fun? Why don't we have a talk.routes] >"replies along news paths are always sub-optimal", jc@minya.UUCP (John Chambers) replies: >You're wrong there. For instance, try responding by email to this article. >You will use a news path. Why? Well, this machine has 4 neighbors, and Nope. Here's the Path: line as it arrived here: Path: ulowell!bbn!uwmcsd1!ig!agate!ucbvax!decwrl!decvax!ima!minya!jc A reply from here would go to mit-eddie!minya!jc. Not a news path. >[Any errors in the above are due to failures in the logic of the keyboard, Obviously. ..Bob -- Bob Page, U of Lowell CS Dept. page@swan.ulowell.edu ulowell!page "What a wonder is USENET; such wholesale production of conjecture from such a trifling investment in fact." -- Carl S. Gutekunst
brian@ncrcan.Toronto.NCR.COM (Brian Onn) (08/18/88)
In article <8528@swan.ulowell.edu> page@swan.ulowell.edu (Bob Page) writes: >jc@minya.UUCP (John Chambers) replies: >> [stuff about replies using a news path] > >Nope. Here's the Path: line as it arrived here: >Path: ulowell!bbn!uwmcsd1!ig!agate!ucbvax!decwrl!decvax!ima!minya!jc gawd! How could anyone want to reply using that path?!? >A reply from here would go to mit-eddie!minya!jc. Not a news path. Much better. Score one for smart mailers choosing the best path. My thoughts are with Bob on this one. Brian. -- +-------------------+--------------------------------------------------------+ | Brian Onn | UUCP:..!{uunet!mnetor, watmath!utai}!lsuc!ncrcan!brian | | NCR Canada Ltd. | INTERNET: Brian.Onn@Toronto.NCR.COM | +-------------------+--------------------------------------------------------+