pja@ralph.Lafayette.LA.US (Pete Alleman) (08/16/90)
I would like to setup smail 2.5 to send any mail addressed to a FQDN to my nearest Internet gateway, and only source route mail destined for non-Internet sites. I have not come up with any simple way to achieve this. I would think that any uucp site closely connected to the Internet would want to route mail destined for an Internet site to the nearest gateway. Does anyone have a standard solution? -- Pete Alleman pja@ralph.lafayette.la.us
peter@ficc.ferranti.com (Peter da Silva) (08/17/90)
In article <3827@ralph.Lafayette.LA.US> pja@ralph.Lafayette.LA.US (Pete Alleman) writes: > I would think that any uucp site closely connected to the Internet > would want to route mail destined for an Internet site to the > nearest gateway. Well, that seems reasonable. However, a FQDN does not imply that a site is an internet site. For example, consider: uhnix1.uh.edu---Internet | <-- local call sugar.hackercorp.com | <-- long distance call uunet.uu.net---Internet, and MX for .hackercorp.com Suppose someone at uhnix1.uh.edu wanted to send mail to someone at sugar.hackercorp.com. What do you think their mail router should do? -- Peter da Silva. `-_-' +1 713 274 5180. 'U` peter@ferranti.com (currently not working) peter@hackercorp.com
emv@math.lsa.umich.edu (Edward Vielmetti) (08/18/90)
In article <FK959_@xds13.ferranti.com> peter@ficc.ferranti.com (Peter da Silva) writes: uhnix1.uh.edu---Internet | <-- local call sugar.hackercorp.com | <-- long distance call uunet.uu.net---Internet, and MX for .hackercorp.com Suppose someone at uhnix1.uh.edu wanted to send mail to someone at sugar.hackercorp.com. What do you think their mail router should do? If uhnix1.uh.edu cares enough, and is capable of it, they should act as though they were an MX for sugar.hackercorp.com. The risk they take in doing that is that hackercorp.com might unilaterally start MX'ing themselves to uhnix1.uh.edu without asking for permission or paying for any of the costs, and people might get upset. --Ed Edward Vielmetti, U of Michigan math dept <emv@math.lsa.umich.edu>
peter@ficc.ferranti.com (Peter da Silva) (08/19/90)
In article <EMV.90Aug17164108@stag.math.lsa.umich.edu> emv@math.lsa.umich.edu (Edward Vielmetti) writes: > In article <FK959_@xds13.ferranti.com> peter@ficc.ferranti.com (Peter da Silva) writes: > Suppose someone at uhnix1.uh.edu wanted to send mail to someone at > sugar.hackercorp.com. What do you think their mail router should do? > If uhnix1.uh.edu cares enough, and is capable of it, they should act > as though they were an MX for sugar.hackercorp.com. Uh, huh. In fact they should do that for any site that's accessible via local calls. Which they do. The easy way to do this is run pathalias for local sites (say, Houston). But the point in question is that having an FQDN doesn't mean you're on the internet. Simple point. Why do people keep forgetting it? -- Peter da Silva. `-_-' +1 713 274 5180. 'U` peter@ferranti.com (currently not working) peter@hackercorp.com
pja@ralph.Lafayette.LA.US (Pete Alleman) (08/21/90)
In article <5IA5TY9@ggpc2.ferranti.com> peter@ficc.ferranti.com (Peter da Silva) writes: >But the point in question is that having an FQDN doesn't mean you're on the >internet. Simple point. Why do people keep forgetting it? Maybe I missed something. Under what circumstances would a site with a FQDN ending with an Internet Top Level Domain not be accessible from the Internet? Do many sites use names that are indistinguishable from proper Internet names and have no Internet Name Server records? -- Pete Alleman pja@ralph.lafayette.la.us
mju@mudos.ann-arbor.mi.us (Marc Unangst) (08/21/90)
pja@ralph.Lafayette.LA.US (Pete Alleman) writes: > Maybe I missed something. Under what circumstances would a site with > a FQDN ending with an Internet Top Level Domain not be accessible from > the Internet? Do many sites use names that are indistinguishable from > proper Internet names and have no Internet Name Server records? Well, let's take my site as an example. "mudos.ann-arbor.mi.us" is a FQDN, and ends with an official Internet top-level domain name. If you try "nslookup mudos.ann-arbor.mi.us", you'll be told that it's an MX record, with forwarders leebai.aa.ox.com and itivax.iti.org. But there isn't an A record for mudos, meaning you can't access it through telnet, FTP, or anything else. It isn't directly connected to the Internet. In fact, it's an XT clone running MS-DOS and Waffle, and talks to the outside world solely through UUCP. Maybe you're getting confused. There's a difference between "accessible through the Internet" and "having an Internet nameserver record". In order to be accessible through the Internet, a site must have a valid "A" nameserver record -- but there's nothing to prevent a site that's *not* accessible through the Internet from having any other kind of nameserver record. -- Marc Unangst | mju@mudos.ann-arbor.mi.us | Angular momentum makes the world go 'round. ...!umich!leebai!mudos!mju |
jim@tiamat.fsc.com (Jim O'Connor) (08/21/90)
In article <3890@ralph.Lafayette.LA.US>, pja@ralph.Lafayette.LA.US (Pete Alleman) writes: > In article <5IA5TY9@ggpc2.ferranti.com> peter@ficc.ferranti.com (Peter da Silva) writes: > >But the point in question is that having an FQDN doesn't mean you're on the > >internet. Simple point. Why do people keep forgetting it? > > Maybe I missed something. Under what circumstances would a site with > a FQDN ending with an Internet Top Level Domain not be accessible from > the Internet? Do many sites use names that are indistinguishable from > proper Internet names and have no Internet Name Server records? Just so Peter da Silva doesn't feel alone in this discussion, I'll throw my two cents in: There is a BIG difference between being "accessible" from the Internet and being "on" the Internet. Thanks to many different gateways, user of Compuserve, MCI, and even (I think) ATT Mail are "accessible" from the Internet, but that doesn't mean they are "on" the Internet. My systems are all under the domain fsc.com and tiamat.fsc.com is a FQDN with respect to mail only, since uunet.uu.net will answer queries with an MX record. But something like "ftp tiamat.fsc.com" will fail miserably, since I am not on the Internet. ------------- James B. O'Connor jim@tiamat.fsc.com Ahlstrom Filtration, Inc. 615/821-4022 x. 651
michaelb@wshb.csms.com ( WSHB Operations Eng) (08/21/90)
> uhnix1.uh.edu---Internet > | <-- local call > sugar.hackercorp.com > | <-- long distance call > uunet.uu.net---Internet, and MX for .hackercorp.com > > Suppose someone at uhnix1.uh.edu wanted to send mail to someone at > sugar.hackercorp.com. What do you think their mail router should do? > > If uhnix1.uh.edu cares enough, and is capable of it, they should act > as though they were an MX for sugar.hackercorp.com. The risk they > take in doing that is that hackercorp.com might unilaterally start > MX'ing themselves to uhnix1.uh.edu without asking for permission or > paying for any of the costs, and people might get upset. ^^^^^^^^^^^^^^^^^^^^^^ This is a very valid point about costs. The 'cost' isn't always just a matter of dollars and cents for the phone calls. I have an unpublished link that connects regularly, doesn't cost me very much phone bill wise, and is very reliable. The political cost is prohibitive. As far as I'm concerned, if anyone else tried to use the link it should just as well be marked DEAD. I sometimes don't think about 'cost' in this light when I'm talking about YOUR system, but I do think about it when I'm talking about MY system. What it boils down to is this: even though there may be another 'cheaper' way to send me mail if you knew the links, realistically I'm willing to pay the phone bill to uunet every day to get you to send me mail through them. Michael -- Michael Batchelor--Systems/Operations Engineer #compliments and complaints WSHB - An International Broadcast Station of # letterbox@csms.com The Christian Science Monitor Syndicate, Inc. #technical questions and reports michaelb@wshb.csms.com +1 803 625 4880 # letterbox-tech@csms.com
peter@ficc.ferranti.com (Peter da Silva) (08/22/90)
In article <3890@ralph.Lafayette.LA.US> pja@ralph.Lafayette.LA.US (Pete Alleman) writes: > In article <5IA5TY9@ggpc2.ferranti.com> peter@ficc.ferranti.com (Peter da Silva) writes: > >But the point in question is that having an FQDN doesn't mean you're on the > >internet. Simple point. Why do people keep forgetting it? > Maybe I missed something. Under what circumstances would a site with > a FQDN ending with an Internet Top Level Domain not be accessible from > the Internet? Excuse me. *ARGH* Sorry. I had to get that out. You misunderstand what I mean by "on the internet". I mean having a dedicated line, rather than dialing up a real internet site that's serving as an MX. It's not that a site isn't *accessible* from the internet, it's that the internet link might just be an expensive one for that site, and that an overall cheaper route might be available using pathalias. As, for example, from just about anywhere in Houston to any Houston site using UUNET or PSI as an MX, since Houston has the second largest local calling area in the U.S. (measured by number of customers), probably in the world. It's just about a requirement these days to have a domain. That doesn't mean that domain name system should be considered the best way to get mail to someone. -- Peter da Silva. `-_-' +1 713 274 5180. 'U` peter@ferranti.com
pja@ralph.Lafayette.LA.US (Pete Alleman) (08/23/90)
In article <UZC5=UA@xds13.ferranti.com> peter@ficc.ferranti.com (Peter da Silva) writes: >*ARGH* >You misunderstand what I mean by "on the internet". I mean having a >dedicated line, rather than dialing up a real internet site that's >serving as an MX. It would help if there were (or I knew) the standard terminology. I once heard that "The Internet" was the collection of hosts with registered A records, whereas the "internet" was "The Internet" plus all those sites with registered MX records. >It's not that a site isn't *accessible* from the internet, it's that the >internet link might just be an expensive one for that site, and that an >overall cheaper route might be available using pathalias. As, for example, >from just about anywhere in Houston to any Houston site using UUNET or >PSI as an MX, since Houston has the second largest local calling area >in the U.S. (measured by number of customers), probably in the world. Now the real point. Some sites would prefer smail style uucp routes for their mail. On the other hand, I see in my routing database some 6 hop routes to "Internet" sites when my neighbor could deliver directly. Maybe we need to start putting more information into the uucp maps. It would be nice if pathalias could be set up to "do the right thing." -- Pete Alleman pja@ralph.lafayette.la.us
peter@ficc.ferranti.com (Peter da Silva) (08/24/90)
In article <801@wshb.csms.com> michaelb@wshb.csms.com ( WSHB Operations Eng) writes: > What it boils down to is this: even though there may be another 'cheaper' > way to send me mail if you knew the links, realistically I'm willing to > pay the phone bill to uunet every day to get you to send me mail through > them. That's true for *unpublished* links. But presumable links published in the maps are intended to be used, and the cost (in all senses) to the systems involved has presumably been included in the map entry. Which is (back to the point) why smail's action in not assuming that an FQDN should go staight to the internet is correct, and that simple-minded munging of the PATHS file based on who has an FQDN should be discouraged. -- Peter da Silva. `-_-' +1 713 274 5180. 'U` peter@ferranti.com
peter@ficc.ferranti.com (Peter da Silva) (08/24/90)
In article <3932@ralph.Lafayette.LA.US> pja@ralph.Lafayette.LA.US (Pete Alleman) writes: > On the other hand, I see in my routing database some > 6 hop routes to "Internet" sites when my neighbor could deliver directly. So add another "u." file, that just does this: internet = { list the sites you know to be on the internet here } (LOCAL) > Maybe we need to start putting more information into the uucp maps. > It would be nice if pathalias could be set up to "do the right thing." It can. -- Peter da Silva. `-_-' +1 713 274 5180. 'U` peter@ferranti.com
scott@wiley.uucp (Scott Simpson) (08/24/90)
In article <3932@ralph.Lafayette.LA.US> pja@ralph.Lafayette.LA.US (Pete Alleman) writes: >It would help if there were (or I knew) the standard terminology. >I once heard that "The Internet" was the collection of hosts with >registered A records, whereas the "internet" was "The Internet" >plus all those sites with registered MX records. I was under the impression that the Internet was composed of those hosts that spoke IP and could talk IP to everybody else on the Internet. This would rule out MX/uucp sites. Of course you can address MX/uucp sites, but name addressing and IP packet routing are two orthogonal issues. Remember the goals of the Internet: 1. Universal connectivity. 2. Hiding underlying technology. 3. Not mandating a specific network topology. (There may be more.) The word "internet" I think of as a general term for interconnected networks (e.g., separate ethernets) and not a reference to a specific internet in particular. Scott Simpson TRW scott@coyote.trw.com
fitz@wang.com (Tom Fitzgerald) (08/28/90)
> pja@ralph.Lafayette.LA.US (Pete Alleman) writes: >> On the other hand, I see in my routing database some >> 6 hop routes to "Internet" sites when my neighbor could deliver directly. peter@ficc.ferranti.com (Peter da Silva) writes: > So add another "u." file, that just does this: > internet = { > list the sites you know to be on the internet here > } (LOCAL) I've been trying this, and it doesn't work too well. I'm not sure what the reason is - I just gave up when I ran into problems. First of all, this isn't enough. You may also need: internet.site internet(LOCAL) another.site internet(LOCAL) ... for each machine you might be using to inject stuff into the Internet. Other problems are caused by sites which describe themselves in the UUCP maps in nonstandard ways. Even when it works right, it often causes mail to take a different path than it would have. I'd like to have a technique that just removes unnecessary hops from the paths, but doesn't force any machine to deal with mail that it wouldn't have gotten anyway, and messing with the pathalias input sometimes has this effect. I think that it's necessary to crunch the pathalias output rather than adding stuff to the input to get this done right. --- Tom Fitzgerald Wang Labs fitz@wang.com 1-508-967-5278 Lowell MA, USA ...!uunet!wang!fitz
peter@ficc.ferranti.com (Peter da Silva) (08/30/90)
pja@ralph.Lafayette.LA.US (Pete Alleman) writes: > On the other hand, I see in my routing database some > 6 hop routes to "Internet" sites when my neighbor could deliver directly. peter@ficc.ferranti.com (Peter da Silva) writes: > So add another "u." file, that just does this: > internet = { > list the sites you know to be on the internet here > } (LOCAL) In article <as2jn2.1b3@wang.com> fitz@wang.com (Tom Fitzgerald) writes: > I've been trying this, and it doesn't work too well. We do this internally for our internal internet (some 40 machines running 2 variants of OSI and TCP/IP) and it works fine. We have: opennet = { xds1 xds2 xds3 ... ris0 ris1 ... } (LOCAL) xds1 = xds1.ferranti.com xds2 = xds2.ferranti.com ... (this is generated automatically from /usr/lib/smail/nodes) > internet.site internet(LOCAL) > another.site internet(LOCAL) I don't see why. This will be ignored if you have the costs right. > I think that it's necessary to crunch the pathalias output rather than > adding stuff to the input to get this done right. No, because *if* you crunch the pathalias output you are guaranteed to get this done wrong. -- Peter da Silva. `-_-' +1 713 274 5180. 'U` peter@ferranti.com
chip@tct.uucp (Chip Salzenberg) (08/30/90)
I, also, have concluded that reasonable behavior will not be
forthcoming from raw pathalias output. I massage the output with this
script, named "honey" after Peter Honeyman who gave me the idea:
================================================================
eval 'exec /bin/perl -S $0 ${1+"$@"}'
if 0;
# honey
# Fix up a pathalias listing by removing .com and .edu sites,
# except for those nearby.
die "usage: honey internetpath\n" unless @ARGV == 1;
$INET = shift;
while (<>) {
chop;
next unless ($host, $path) = split(/\t/);
# KLUDGE ALERT
# Kill USF.
next if $path =~ /\busfvax2!/;
if ($host =~ /.+\./) {
next if ($path =~ y/!/!/) > 3 || $path =~ /^${INET}!/o;
}
elsif ($host =~ /^\./) {
$path = "$INET!%s";
}
print $host, "\t", $path, "\n";
}
================================================================
This script strips all paths to FQDNs unless the bang path is three
hops or less. It also replaces all paths to domain gateways with the
path given on the command line. In my case, $INET is "pdn!uunet".
(The "KLUDGE ALERT" removes all paths that include the defunct host
named "usfvax2").
Don't be a slave to what pathalias gives you. Take control of your
own destiny. Er, your own mail routing. :-)
--
Chip Salzenberg at Teltronics/TCT <chip@tct.uucp>, <uunet!pdn!tct!chip>joe@cbnews.att.com (Joseph Judge) (08/30/90)
We have an INTERNET host, so I want pathalias to *use* it.
All I can think of is grabbing all the alias records in the
maps and change the: name = name.alias
and make it: our_host name.alias (some_low_cost)
What I need is a better way to "know" which machines in the maps
are actually accessible from the INTERNET (better than the
preprocessing I am doing).
Peter, Tom, how do you get these "lists" of hosts to put
into your "INTERNET = { ...}" or "site internet(local)" records?
Joseph Judge postmaster@ATT.COM
This all would be sent in with the rest of the records.
In article <as2jn2.1b3@wang.com>, fitz@wang.com (Tom Fitzgerald) writes:
> peter@ficc.ferranti.com (Peter da Silva) writes:
> > So add another "u." file, that just does this:
> > internet = {
> > list the sites you know to be on the internet here
> > } (LOCAL)
>
> I've been trying this, and it doesn't work too well. I'm not sure what
> the reason is - I just gave up when I ran into problems. First of all, this
> isn't enough. You may also need:
>
> internet.site internet(LOCAL)
> another.site internet(LOCAL)
> ...
>
> I think that it's necessary to crunch the pathalias output rather than
> adding stuff to the input to get this done right.
> peter@ficc.ferranti.com (Peter da Silva) (08/31/90)
In article <26DC0A0E.1352@tct.uucp> chip@tct.uucp (Chip Salzenberg) writes: > # Fix up a pathalias listing by removing .com and .edu sites, > # except for those nearby. How about just culling the maps you feed into pathalias to only include nearby sites (say, the same state)? That's what I do. -- Peter da Silva. `-_-' +1 713 274 5180. 'U` peter@ferranti.com
joe@cbnews.att.com (Joseph Judge) (09/01/90)
So, email comes in to your gateway for "utai" - it will still take the "hop1!hop2! ... !utai!%s" You've just made it so that folks can send email to neat.cs.toronto.edu over the internet. That's what we do also. What we need is *any* email for "mcsun" goes over the INTERNET, not just if they send to "mcsun.eu.net" or "mcsun.nl". Thanks for the perl script! Joseph Judge postmaster@ATT.COM In article <26DC0A0E.1352@tct.uucp>, chip@tct.uucp (Chip Salzenberg) writes: > ================================================================ > > This script strips all paths to FQDNs unless the bang path is three > hops or less. It also replaces all paths to domain gateways with the > path given on the command line. In my case, $INET is "pdn!uunet". >
urlichs@smurf.sub.org (Matthias Urlichs) (09/02/90)
In comp.mail.uucp, article <1990Aug30.152241.28240@cbnews.att.com>,
joe@cbnews.att.com (Joseph Judge) writes:
<
< We have an INTERNET host, so I want pathalias to *use* it.
<
< All I can think of is grabbing all the alias records in the
< maps and change the: name = name.alias
< and make it: our_host name.alias (some_low_cost)
<
I am doing it the other way 'round.
I wrote a Perl script which extensively digests the maps. It finds entries of
"host = some.fq.domain" form, and uses these to generate routes which start with the last FQDN entry. Consider the following map:
one = one.edu
one two
zero one
me one
Then the output would be
zero zero!%s
one one.edu!%s
two one.edu!two!%s
Or at least it would be -- however, I'm using MMDF which allows a lot of short
cuts. For instance, one.edu isn't locally connected. I'm not on the Internet
either, so MMDF will consult its tables recursively to find a route to the
internet, and then append the above routes to "one" and "two" to that.
It also allows me to keep the processing of local and regional maps totally
separate, while considering them together when routing (but still giving
priority to the regional map).
For use with smail, someone would have to add additional postprocessing which
integrates these tables into one smail-style routing table. Any takers?
(This would probably require a 200-line Perl program.)
All of the above presupposes that every FQDN which is advertised in the maps
does have an Internet MX record pointing to it (or to the wildcard of one of
its superdomains). It also assumes that you want to send mail across the
Internet instead of generating ten-bang-paths which tend to get routed to
/dev/null somewhere. :-(
--
Matthias Urlichs -- urlichs@smurf.sub.org -- urlichs@smurf.ira.uka.de
Humboldtstrasse 7 - 7500 Karlsruhe 1 - FRG -- +49+721+621127(Voice)/621227(PEP)peter@ficc.ferranti.com (Peter da Silva) (09/06/90)
In article <1990Aug30.152241.28240@cbnews.att.com> joe@cbnews.att.com (Joseph Judge) writes: > Peter, Tom, how do you get these "lists" of hosts to put > into your "INTERNET = { ...}" or "site internet(local)" records? When I see a new site name in a message that indicates the site is on the internet, I add it. The list doesn't have to be complete to be useful. -- Peter da Silva. `-_-' +1 713 274 5180. 'U` peter@ferranti.com
rickert@mp.cs.niu.edu (Neil W. Rickert) (09/06/90)
In article <G4P5X93@xds13.ferranti.com> peter@ficc.ferranti.com (Peter da Silva) writes: >In article <1990Aug30.152241.28240@cbnews.att.com> joe@cbnews.att.com (Joseph Judge) writes: >> Peter, Tom, how do you get these "lists" of hosts to put >> into your "INTERNET = { ...}" or "site internet(local)" records? > >When I see a new site name in a message that indicates the site is on the >internet, I add it. The list doesn't have to be complete to be useful. My method was as follows: 1. Search for strings of the form name1 = name2 where at least one of the names contains a '.' and does not begin with a '.' 2. Eliminate all name so constructed which do not end in a known top level domain. 3. Build the pathalias map. 4. Sort, and accumulate the freqency of use of each of these presumed internet gateways. Eliminate all but the top 200 from my list. 5. Build the pathalias map again, and sort again. Now start checking, in order of frequency of use, as to whether they have Internet addresses. Stop when you have enough. (The last step isn't as bad as it seems, since many names are well known so don't need to be looked up). I finished up with almost 100 hosts. -- =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= Neil W. Rickert, Computer Science <rickert@cs.niu.edu> Northern Illinois Univ. DeKalb, IL 60115. +1-815-753-6940
fitz@wang.com (Tom Fitzgerald) (09/08/90)
joe@cbnews.att.com (Joseph Judge) writes: > Peter, Tom, how do you get these "lists" of hosts to put > into your "INTERNET = { ...}" or "site internet(local)" records? This is the wrong answer, but you just gotta know. I've got a map of NEARnet here I use to identify local sites. Some people advertise their IP addresses in their map entries. You can pretty much assume that anyone with a domain is not more than one hop off the Internet, though as Peter and the gentleman at AthabascaU can testify, that can be one long hop. [Actually some domain addresses are more than one hop off, but very very few]. It would be _real_nice_ if this stuff was in the map entries.... #F Internet forwarder (destination for MX record) #I IP address, for connected sites Some people use the #F line, but it's not real consistent. With the #F line we could tell for sure whether changing "site1!site2!host" to "site1!host.com" would be harmful. --- Tom Fitzgerald Wang Labs fitz@wang.com 1-508-967-5278 Lowell MA, USA ...!uunet!wang!fitz