aem@mthvax.cs.miami.edu (a.e.mossberg) (07/11/90)
Many of the better BITNET hosts also are on Internet. Would having a table of these for sendmail be a good idea, and if so, how would one set up a rule to map a $1.bitnet to the corresponding internet host name from the file? The reason I ask is that using the ip address when available would be a whole lot quicker than sending it to cunyvm and letting it make its way down the bitnet links. aem -- a.e.mossberg / aem@mthvax.cs.miami.edu / aem@umiami.BITNET / Pahayokee Bioregion Philosophy only works when it is shared. - Andrew Mossberg
gs26@prism.gatech.EDU (Glenn R. Stone) (07/11/90)
In <1990Jul10.212525.21017@mthvax.cs.miami.edu> aem@mthvax.cs.miami.edu (a.e.mossberg) writes: >Would having a table of these for sendmail be a good idea, and if so, >how would one set up a rule to map a $1.bitnet to the corresponding >internet host name from the file? Nope, bzzzt, wrong. The whole point of having .bitnet is to force it to route thru the bitnet links... so that when things screw up, mail will still go thru. >The reason I ask is that using the ip address when available would be >a whole lot quicker than sending it to cunyvm and letting it make >its way down the bitnet links. Bitnet sites that are on the internet also have a domain-ized name that is used for sending mail to it thru the internet. e.g. GITVM1 (bitnet) is vm1.gatech.edu (internet); DRYCAS (bitnet) is DRYCAS.CLUB.CC.CMU.EDU (internet), etc.... Admittedly, when everything is peachy-keen, ip point-to-point is an order of magnitude faster, but when the chips are down, a slow RSCS link is better than no link at all.... forcing a reroute thru IP regardless of the address form introduces a single point of failure. -- Glenn R. Stone gs26@prism.gatech.edu
emv@math.lsa.umich.edu (Edward Vielmetti) (07/11/90)
In article <11197@hydra.gatech.EDU> gs26@prism.gatech.EDU (Glenn R. Stone) writes:
etc.... Admittedly, when everything is peachy-keen,
ip point-to-point is an order of magnitude faster, but
when the chips are down, a slow RSCS link is better than
no link at all.... forcing a reroute thru IP regardless
of the address form introduces a single point of failure.
Many (most?) long-haul 'bitnet' links are running RSCS on top of IP,
and I don't think it's exactly common to keep the dedicated leased
point to point link around once the VMNET software is going. So a
forced route through bitnet doesn't necessarily buy you anything.
--Ed
Edward Vielmetti, U of Michigan math dept <emv@math.lsa.umich.edu>
comp.archives moderatorkarl_kleinpaste@cis.ohio-state.edu (07/11/90)
aem@mthvax.cs.miami.edu writes:
Many of the better BITNET hosts also are on Internet.
Would having a table of these for sendmail be a good idea, and if so,
how would one set up a rule to map a $1.bitnet to the corresponding
internet host name from the file?
Heh heh heh...
Danger -- nasty mail hack approaching...
Fake yourself a "bitnet" top-level domain. It should look something
like (gads, I almost can't believe I'm doing this...)
bitnet. in soa mthvax.cs.miami.edu. aem.mthvax.cs.miami.edu. (
900711 ; serial
86400 ; refresh
300 ; retry
3600000 ; expire
86400 ) ; minimum
bitnet. in ns mthvax.cs.miami.edu.
gitvm1 in cname vm1.gatech.edu.
drycas in cname drycas.club.cc.cmu.edu.
ohstvma in cname ohstvma.ircc.ohio-state.edu.
ohstpy in cname ohstpy.mps.ohio-state.edu.
; etc...Where DO you get a list of DNS equivalents for BITNET hosts?
* in mx 0 cunyvm.cuny.edu.
(Thanx to Glenn for a couple of extra sample bitnet hostnames to put
in there...)
Now you don't even have to notice $1.bitnet; just make sure that you
pass hostnames through $[$] and they'll magically turn into The Right
Thing.
Now, of course, if anything outside your own nameserver ever discovers
that something claims to know about a ".bitnet" top-level domain,
well, there'll be hell to pay...you also take on a non-trivial task of
maintaining a reasonable list of CNAMEs...I don't recommend this, but
it oughta work...
--karl
armed and dangerous
with a nameserverCraig_Everhart@TRANSARC.COM (07/11/90)
If you're going to define a fake domain full of CNAMEs, at least stay out of everybody else's way by defining them within ``your local domain,'' whatever that is. So if you're ohio-state.edu., and your name lookups will append .ohio-state.edu to any names going through, then define a zone bitnet.ohio-state.edu with all the CNAMEs you want. People's attempts to resolve drycas.bitnet will try appending your local domain (or a sequence of local domains), and all you have to do is ensure that one of those attempts turns drycas.bitnet into drycas.bitnet.ohio-state.edu, from which point you've defined a CNAME for that host as being drycas.club.cc.cmu.edu. Of course, this screws up the case where foo.bitnet doesn't have an Internet equivalent; ``foo.bitnet'' could get rewritten as your fake foo.bitnet.ohio-state.edu domain before being passed on elsewhere, but maybe you could stand on your head enough in your sendmail.cf to detect that case. Craig
gs26@prism.gatech.EDU (Glenn R. Stone) (07/11/90)
In <EMV.90Jul11013021@urania.math.lsa.umich.edu> emv@math.lsa.umich.edu (Edward Vielmetti) writes: >Many (most?) long-haul 'bitnet' links are running RSCS on top of IP, >and I don't think it's exactly common to keep the dedicated leased >point to point link around once the VMNET software is going. So a >forced route through bitnet doesn't necessarily buy you anything. hmmm, interesting. but the "necessarily" is the keyword.... routing gets weird around Atlanta. Northbound BITNET from computer center machines here goes thru GITVM1->UGA, but Internet traffic goes gatech.edu->hubcap.clemson.edu (I think... but the point is it goes thru gatech, currently a VAX 11/780 that tends to wheeze every once in a while).... I imagine things get similarly weird around Maryland, NYC, etc.... Besides, I thought it was bad form to dynamically reroute someone's mail if it was anything other than user@host.do.main? DWIM,D! :) -- Glenn R. Stone gs26@prism.gatech.edu "When the going gets weird, the weird turn pro."
ittai@shemesh.GBA.NYU.EDU (Ittai Hershman) (07/12/90)
> Bitnet sites that are on the internet also have a domain-ized name that > is used for sending mail to it thru the internet. e.g. > GITVM1 (bitnet) is vm1.gatech.edu (internet); > DRYCAS (bitnet) is DRYCAS.CLUB.CC.CMU.EDU (internet), The problem is that end-users have no way of knowing this. NYU, for instance, is very well connected to the Internet (two T1's to two different regionals), yet we receive tons of mail on our 9600 baud RSCS connections from other institutions which, like us, are on both BITnet and the Internet. At present, at my site, mail sent out on the RSCS link gets the RSCS/BITnet nodename, while mail sent out on the Internet gets a fully qualified domain name. I have been very tempted to have all outgoing mail sent with the fully qualified domain name which would force replies from dual BITnet/Internet sites to favor the Internet. The dwindling population of BITnet only sites would presumably route the mail via a BITnet/Internet gateway. Comments? No flames please -- this is a topic some people feel is religous, and none of us need a flame war. Come to think of it, shouldn't this be on comp.mail.misc: I've crossposted and directed followups to there. -Ittai
bob@MorningStar.Com (Bob Sutterfield) (07/12/90)
In article <KARL.90Jul11094604@giza.cis.ohio-state.edu> karl_kleinpaste@cis.ohio-state.edu writes:
Now, of course, if anything outside your own nameserver ever
discovers that something claims to know about a ".bitnet" top-level
domain, well, there'll be hell to pay...
SHHH! Don't anybody say anything, so that our little secret stays
safe. I know we can trust the 10,000 readers of this group, but who
knows about all the other rabble? Good think Karl didn't tell a lot
of people about it - they might discover something, and, well, there'd
be hell to pay...Makey@Logicon.COM (Jeff Makey) (07/13/90)
In article <KARL.90Jul11094604@giza.cis.ohio-state.edu> karl_kleinpaste@cis.ohio-state.edu writes: >Now, of course, if anything outside your own nameserver ever discovers >that something claims to know about a ".bitnet" top-level domain, >well, there'll be hell to pay... It's not that bad, really. Here's the complete, unedited zone file I use: ----------------------------------------------------------------- ; ; @(#)bitnet.zone 1.1 (Logicon) 02/23/90 ; @ IN SOA Snoopy.Logicon.COM. Postmaster@Logicon.COM. ( 1 ; Serial Number 7200 ; Refresh Interval (2 hours) 1800 ; Retry Interval (30 minutes) 1209600 ; Secondary Expiration (14 days) 86400 ) ; Minimum Expiration (24 hours) IN NS localhost. ; ; advertised Internet<->BITNET mail bridges ; * IN MX 100 CUNYVM.CUNY.EDU. IN MX 100 MITVMA.MIT.EDU. ; ; Some BITNET hosts that are also known by real domain names, ; and their respective canonical names. ; BROWNVM IN CNAME BROWNVM.BROWN.EDU. CUNYVM IN CNAME CUNYVM.CUNY.EDU. ----------------------------------------------------------------- Note the use of "localhost" as the nameserver to make it inconvenient for other sites to depend on me. The CNAME list is short because I handle very little BITNET traffic, and the wildcard MX entries take care of all of the other hosts. If the official Internet<->BITNET gateway sites would get together and register an honest BIT.NET zone then all of us Internet-only sites could put R$*<@$+.BITNET>$* $:$1<@$2.BIT.NET>$3 into our .cf files and we'd never have to worry about it again. :: Jeff Makey Department of Tautological Pleonasms and Superfluous Redundancies Department Disclaimer: All opinions are strictly those of the author. Internet: Makey@Logicon.COM UUCP: {nosc,ucsd}!logicon.com!Makey
ittai@shemesh.GBA.NYU.EDU (Ittai Hershman) (07/13/90)
> If the official Internet<->BITNET gateway sites would get together and > register an honest BIT.NET zone then all of us Internet-only sites > could put For years now, the idea of a BIT.NET zone (or any permutation thereof) has been rejected by the "Name Czars". I was hoping that the merger of CSnet and BITnet into CREN would have straightened out this whole BITnet-Internet interoperability issue, but as far as I know nothing has been done to tackle that problem (nor any significant chunk of it). It is rather sad that with all the talk of gigabit networks and interoperability, no one seems to care enough about this issue to make something happen. I suspect that of those who do care, half are waiting for BITnet to go away, and the other half are waiting for the Internet to go away :-) -Ittai
Makey@Logicon.COM (Jeff Makey) (07/14/90)
In article <3803@shemesh.GBA.NYU.EDU> ittai@shemesh.GBA.NYU.EDU (Ittai Hershman) writes: >For years now, the idea of a BIT.NET zone (or any permutation thereof) >has been rejected by the "Name Czars". I don't think they could really prevent *any* permutation. If someone with the appropriate responsibility and authority would maintain the zone database, they could call it BITNET.CUNY.EDU for all I care. As I see it, the hard problem is getting someone to commit to maintaining the zone. >It is rather sad that with all the talk of gigabit networks and >interoperability, no one seems to care enough about this issue to make >something happen. Yup. :: Jeff Makey Department of Tautological Pleonasms and Superfluous Redundancies Department Disclaimer: All opinions are strictly those of the author. Internet: Makey@Logicon.COM UUCP: {nosc,ucsd}!logicon.com!Makey
cmf@obie.cis.pitt.edu (Carl M. Fongheiser) (07/16/90)
In article <KARL.90Jul11094604@giza.cis.ohio-state.edu> karl_kleinpaste@cis.ohio-state.edu writes: >; etc...Where DO you get a list of DNS equivalents for BITNET hosts? Simple; you fetch a copy of BITEARN NODES, and grab the :internet tag for each node. >Now, of course, if anything outside your own nameserver ever discovers >that something claims to know about a ".bitnet" top-level domain, >well, there'll be hell to pay...you also take on a non-trivial task of >maintaining a reasonable list of CNAMEs...I don't recommend this, but >it oughta work... Assuming you can get the information at all (more like know *how* to get the information), it's easy to generate the list. NOTE: You will make some system maintainers *very* unhappy by doing this. I know of certain EARN administrators, in particular, who would *much* rather receive their mail from the US via BITNET/EARN. This was all hashed out on the NODMGT-L list a couple months ago. Be careful! Carl Fongheiser cmf@unix.cis.pitt.edu
Makey@Logicon.COM (Jeff Makey) (07/17/90)
In article <25918@unix.cis.pitt.edu> cmf@obie.cis.pitt.edu (Carl M. Fongheiser) writes: >In article <KARL.90Jul11094604@giza.cis.ohio-state.edu> karl_kleinpaste@cis.ohio-state.edu writes: >>; etc...Where DO you get a list of DNS equivalents for BITNET hosts? > >Simple; you fetch a copy of BITEARN NODES, and grab the :internet tag for >each node. [...] >Assuming you can get the information at all (more like know *how* to get >the information), it's easy to generate the list. > >NOTE: You will make some system maintainers *very* unhappy by doing this. > I know of certain EARN administrators, in particular, who would > *much* rather receive their mail from the US via BITNET/EARN. > >This was all hashed out on the NODMGT-L list a couple months ago. Be >careful! This is exactly why I think BITNET should administrate an official Internet domain for themselves. Any site that prefers their BITNET interface to their Internet interface simply doesn't have to have a CNAME record, or else has their MX record point to their nearest INTERBIT site. Either way, thousands of Internet-only sites simply let the Domain Name System do the job it is designed to do. No muss, no fuss, and the mail gets through. :: Jeff Makey Department of Tautological Pleonasms and Superfluous Redundancies Department Disclaimer: All opinions are strictly those of the author. Internet: Makey@Logicon.COM UUCP: {nosc,ucsd}!logicon.com!Makey