dwatts@ki.UUCP (Dan Watts) (10/18/90)
Has anybody gotten a sendmail.cf to work on the RS/6000?
Specifically, I've got an SGI which is my UUCP link to the outside world.
Locally, I've got the RS/6000 and a Sun 4. On the Sun 4, I defined my
SGI as my mail router and all mail to 'user@host' or 'host!user' gets
forwarded on to the SGI for delivery. Of course, if the host is local,
then the Sun just sends the email via TCP/IP on the Ethernet.
On the RS/6000, I can't make it do this. The only thing I've been
successful at is 'user@host' or 'host!user' where 'host' is a local
TCP/IP node on the Ethernet. If I try to send to 'user@remoteohst'
it won't forward. I tried defining the $U macro like they say, but
that doesn't work. When I send to 'host!user' I get:
No '@' foudn in SMTP recipient "host!user".
If I try to send to 'user@host' it tries to send the mail via TCP/IP!
Any help would be appreciated.
--
#####################################################################
# CompuServe: >INTERNET:uunet.UU.NET!ki!dwatts Dan Watts #
# UUCP : ...!{uunet | wgc386}!ki!dwatts Ki Research, Inc. #
############### New Dimensions In Network Connectivity ##############jackv@turnkey.tcc.com (Jack F. Vogel) (10/21/90)
In article <871@ki.UUCP> dwatts@ki.UUCP (Dan Watts) writes: > >Has anybody gotten a sendmail.cf to work on the RS/6000? >Specifically, I've got an SGI which is my UUCP link to the outside world. >Locally, I've got the RS/6000 and a Sun 4. On the Sun 4, I defined my >SGI as my mail router and all mail to 'user@host' or 'host!user' gets >forwarded on to the SGI for delivery. Of course, if the host is local, >then the Sun just sends the email via TCP/IP on the Ethernet. [....this isn't working on the 6000...] I can appreciate your frustration Dan. The sendmail.cf file shipped with AIX, and before any of you 6000 support types rise up in revolt this applies to AIX370 and PS/2 as well :-}, is extremely simpleminded. It has a chance of working OK on real internet-connected sites but that is about it. If you are in an increasingly common configuration, where you have a local domain and its network, but rely on a uucp-style connection to an internet MX fowarder for all other domain-style addresses, it just won't work for you as shipped. Basically, you need to do some surgery on Ruleset Zero. As shipped any domain-type address, meaning user@host will resolve to the tcp mailer, what you need to do is circumvent this. What you want in the sort of configuration I described above is basically to allow anything of the form 'user@host.mydomain' or 'user@host.mysubdomain.mydomain' to go ahead and resolve to tcp (assuming you talk smtp to any host in your domain), but for all other domain addresses to be sent to your internet forwarder (or what you may call relayhost). This is what we do internally for locus.com and its subdomains. Now, what do you need to do concretely? There is no one true way to set this up, all I can do is give you one example, other schemas are possible, and unfortunately will require a fairly good understanding of sendmail's macros and rewrite rules. So, anyway, here goes... First define a macro for your forwarder: DRmy_internet_forwarder And the mailer to use to get there, this might be tcp or uucp depending... DMuucp Now, the second to the last rewrite rule in sendmail.cf looks as follows: R$*<@$+>$* $#tcp$@$2$:$1<@$2>$3 You want to change this to the following 2 rules: R$*<@$-.$D>$* $#tcp$@$2.$D$:$1<@$2.$D>$3 user@host.ourdomain # Send all other domain addresses to the relay host R$*<@$+>$* $#$M$@$R$:$1<@$2>$3 user@host.otherdomains You will need to make certain that any unqualified hostname, either in rewrite rules or via the nameserver gets a fully qualified form before reaching these rules! These rules also only cover the simple case where you have a single flat domain without subdomains, if you do have subdomains as locus.com does things get more complex. What I would do there is make the D macro into the base domain, i.e., in our case DDlocus.com. Next define a D class with all the legal subdomains, eg. CDla sd bos. Then the 2 rules above would then be changed into 3 rules as follows: # Handle any subdomain addresses... R$*<@$-.$=D.$D>$* $#tcp$@$2.$3.$D$:$1<@$2.$3.$D>$3 # Addresses in the base domain... R$*<@$-.$D>$* $#tcp$@$2.$D$:$1<@$2.$D>$3 user@host.ourdomain # Send all other domain addresses to the relay host R$*<@$+>$* $#$M$@$R$:$1<@$2>$3 user@host.otherdomains Making these changes should get you running, you might also take a look at the sendmail.cf on the other machines that are working properly to see how they implemented things. >I tried defining the $U macro like they say, but >that doesn't work. When I send to 'host!user' I get: > >No '@' foudn in SMTP recipient "host!user". Hmmm... From looking over the cf file I would have expected this would work. You should run sendmail in Address Test Mode and follow the ruleset flow to determine whats wrong here. Good Luck, and as always I am available via email for further questions. Also, you should try IBM support for help if this isn't enough to get you going. Disclaimer: I rule the mail, but I don't speak for the company :-}. -- Jack F. Vogel jackv@locus.com AIX370 Technical Support - or - Locus Computing Corp. jackv@turnkey.TCC.COM
paul@uxc.cso.uiuc.edu (Paul Pomes - UofIllinois CSO) (10/23/90)
dwatts@ki.UUCP (Dan Watts) writes: >Has anybody gotten a sendmail.cf to work on the RS/6000? >Specifically, I've got an SGI which is my UUCP link to the outside world. >Locally, I've got the RS/6000 and a Sun 4. On the Sun 4, I defined my >SGI as my mail router and all mail to 'user@host' or 'host!user' gets >forwarded on to the SGI for delivery. Of course, if the host is local, >then the Sun just sends the email via TCP/IP on the Ethernet. I found the sendmail/sendmail.cf system as supplied to be more than inadequate. We've replaced it with the IDA sendmail (V5.65+IDA-1.3.5). It doesn't work with frozen config files unless you supply a malloc that doesn't put critical values in the data segment. A getpwent.c file that hasn't been linked with the system malloc is required as well. What I wouldn't give for a straight 4.3 BSD-reno port for our 6K. AIX is one royal pain in the neck. Btw, the IDA sendmail can be found on uxc.cso.uiuc.edu in the pub directory. /pbp -- Paul Pomes I'm official (officially myself). Accept no substitutes unless they pay cash. UUCP: {att,iuvax,uunet}!uiucuxc!paul Internet, BITNET: paul@uxc.cso.uiuc.edu US Mail: UofIllinois, CSO, 1304 W Springfield Ave, Urbana, IL 61801-2910
jeffe@sandino.austin.ibm.com (Peter Jeffe 512.823.4091) (10/23/90)
In article <871@ki.UUCP> dwatts@ki.UUCP (Dan Watts) writes: >I tried defining the $U macro like they say, but >that doesn't work. When I send to 'host!user' I get: > >No '@' foudn in SMTP recipient "host!user". I tried posting this several days ago but it got eaten. So better late than never: this is a bug in gold sendmail that was fixed in the 9035 build (the September update?). Basically it complains about any address that resolves to an smtp mailer but doesn't have an '@' in it. After you apply this update, you should be able to use the U macro to route any uucp addresses (either user@host.uucp or host!user) via your uucp relay host. In article <1990Oct20.205159.2866@turnkey.tcc.com> jackv@turnkey.TCC.COM (Jack F. Vogel) writes: >I can appreciate your frustration Dan. The sendmail.cf file shipped with >AIX, and before any of you 6000 support types rise up in revolt this >applies to AIX370 and PS/2 as well :-}, is extremely simpleminded. It >has a chance of working OK on real internet-connected sites but that is >about it. If you are in an increasingly common configuration, where you >have a local domain and its network, but rely on a uucp-style connection >to an internet MX fowarder for all other domain-style addresses, it just >won't work for you as shipped. Basically, you need to do some surgery on >Ruleset Zero. I don't understand, Jack. Why don't you use the MX record to point you to the right host, instead of mucking about with ruleset 0 to kludge it up? I can understand if you want to get very granular in your routing, but for the scenario you describe the nameserver should tell you where to go. And if you do want to get fancier, what stock .cf file would do that for you? ------------------------------------------------------------------------------- Peter Jeffe ...uunet!cs.utexas.edu!ibmchs!auschs!sandino.austin.ibm.com!jeffe first they want a disclaimer, then they make you pee in a jar, then they come for you in the night
jeffe@sandino.austin.ibm.com (Peter Jeffe 512.823.4091) (10/23/90)
In article <1990Oct22.193236.18485@ux1.cso.uiuc.edu> paul@uxc.cso.uiuc.edu (Paul Pomes - UofIllinois CSO) writes: >I found the sendmail/sendmail.cf system as supplied to be more than inadequate. >We've replaced it with the IDA sendmail (V5.65+IDA-1.3.5). Hmmm, Paul, I'm sort of confused as to what you find so terrible about AIX3's sendmail? It's real close to 5.61 with several enhancements that I think are not too shabby. So please enlighten me as to what is in IDA's 5.65 that you like so much more, and we'll do our best to get it into the next release! ------------------------------------------------------------------------------- Peter Jeffe ...uunet!cs.utexas.edu!ibmchs!auschs!sandino.austin.ibm.com!jeffe first they want a disclaimer, then they make you pee in a jar, then they come for you in the night
jackv@turnkey.tcc.com (Jack F. Vogel) (10/24/90)
In article <3973@awdprime.UUCP> jeffe@sandino.austin.ibm.com (Peter Jeffe 512.823.4091) writes: [ with regard to my suggestion to change Ruleset Zero to use relayhost...] >I don't understand, Jack. Why don't you use the MX record to point you >to the right host, instead of mucking about with ruleset 0 to kludge it >up? I can understand if you want to get very granular in your routing, >but for the scenario you describe the nameserver should tell you where >to go. And if you do want to get fancier, what stock .cf file would do >that for you? Well, for one thing, what if a small network/domain doesn't use a nameserver? They might well be able to maintain a small host file. Also, maybe its just my lack of experience in configuring a non-internet nameserver (meaning one not talking to the real root nameservers), but is this sort of thing really doable? The configuration I was thinking of is where what we want to do is pass all domain addresses not in our domain via uucp to, say, uunet. Now how is the nameserver supposed to be set up to give you an MX to something that you don't talk SMTP to?? Should you just have some sort of wildcard for anything outside my SOA pointing at uunet and then when my nameserver fails to resolve that A record sendmail figures out its a uucp link?? As I said, maybe this is just due to my lack of expertise using the nameserver, please enlighten me, if this can be done we could change things for locus.com. Disclaimer: I speak for myself, not LCC or IBM. -- Jack F. Vogel jackv@locus.com AIX370 Technical Support - or - Locus Computing Corp. jackv@turnkey.TCC.COM
gs26@prism.gatech.EDU (Glenn R. Stone) (10/24/90)
In <1990Oct23.175522.10472@turnkey.tcc.com> jackv@turnkey.tcc.com (Jack F. Vogel) writes: >In article <3973@awdprime.UUCP> jeffe@sandino.austin.ibm.com (Peter Jeffe 512.823.4091) writes: > >[ with regard to my suggestion to change Ruleset Zero to use relayhost...] > >>I don't understand, Jack. Why don't you use the MX record to point you >>to the right host, instead of mucking about with ruleset 0 to kludge it >>up? >Well, for one thing, what if a small network/domain doesn't use a nameserver? .... among other things... I run a small cluster (soon to be its own subnet) of 6k's hung off gatech's broadband. I have found that it's much easier to just not fool with named at all, and just let remote nameservers do the work (if you have access to them) or dump mail off to an appropriate smart-host (if not). That way you only have one config file to muck with, ergo only one place things can go wrong. Occam's razor (aka the KISS principle {keep it simple, stupid}) applies. 'sides, there's plenty more people who mess with sendmail out here on the net that can help out than those who diddle with such vagaries as named... sendmail's even big enough to have it's own group. (pardon the grammar, after talking C and sendmail most of the day, it's hard to think like an ordinary critter :) -- Glenn Stone gs26@prism.gatech.edu, glenns@eas.gatech.edu
shawn@jdyx.UUCP (Shawn Hayes) (10/24/90)
Help!! I'm trying to bring an RS/6000 into an ethernet network of HP 9000s and X Terminals and I'm having lot's of problems. I managed to get the IBM on the network so that we could do rlogins and everything seemed fine. Then we had to move the HPs for a class. Now that everything is back in place the IBM won't talk anymore. I've added the various host names through SMIT, brought up TCP/IP and the various deamons but I'm still having troubles. When I try to do an rlogin from the IBM to an HP I get the message "No route to host". If I try to do an rlogin from the IBM to itself I get the message "Permission denied." And finally if I try to rlogin to the IBM through the loopback host it works fine. Anyone have any suggestions??? Shawn
paul@uxc.cso.uiuc.edu (Paul Pomes - UofIllinois CSO) (10/25/90)
jeffe@sandino.austin.ibm.com (Peter Jeffe 512.823.4091) writes: >Hmmm, Paul, I'm sort of confused as to what you find so terrible about AIX3's >sendmail? It's real close to 5.61 with several enhancements that I think are >not too shabby. So please enlighten me as to what is in IDA's 5.65 that you >like so much more, and we'll do our best to get it into the next release! I'd really like to follow up via email but things look to be very screwed up on your DNS server and mail machine: % mx sandino.austin.ibm.com sandino.austin.ibm.com preference = 10, mail exchanger = cs.utexas.edu.austin.ibm.com OK, let the contact for the austin.ibm.com zone know about it: Date: Wed, 24 Oct 90 14:37:40 CDT To: <paul@ux2.cso.uiuc.edu> From: MAILER-DAEMON@test.austin.ibm.com (Mail Delivery Subsystem) Subject: Returned mail: Service unavailable ----- Transcript of session follows ----- >>> HELO test.austin.ibm.com <<< 553 test.austin.ibm.com Do not communicate with self 554 <root@austin.ibm.com>... Service unavailable: No such file or directory ----- Unsent message follows ----- Received: by test.austin.ibm.com (AIX 2.1 2/4.03) id AA01576; Wed, 24 Oct 90 14:37:40 CDT Received: by ux2.cso.uiuc.edu id AA13474 (5.65+/IDA-1.3.5 for root@austin.ibm.com); Wed, 24 Oct 90 19:37:16 GMT Date: Wed, 24 Oct 90 19:37:16 GMT From: Paul Pomes <paul@ux2.cso.uiuc.edu> Message-Id: <9010241937.AA13474@ux2.cso.uiuc.edu> To: root@austin.ibm.com Subject: missing period in DNS records Try printing the MX record for sandino.austin.ibm.com . /pbp -- Paul Pomes UUCP: {att,iuvax,uunet}!uiucuxc!paul Internet, BITNET: paul@uxc.cso.uiuc.edu US Mail: UofIllinois, CSO, 1304 W Springfield Ave, Urbana, IL 61801-2910
jeffe@sandino.austin.ibm.com (Peter Jeffe 512.823.4091) (10/26/90)
In article <1990Oct23.175522.10472@turnkey.tcc.com> jackv@turnkey.TCC.COM (Jack F. Vogel) writes: >> [ my suggestion to use MX records rather than hacking ruleset 0 ] > >Well, for one thing, what if a small network/domain doesn't use a nameserver? Then they definitely can't query it for MX records :-) >The configuration I was thinking of is where what we want to do is >pass all domain addresses not in our domain via uucp to, say, uunet. Ahhh, so sorry. If you're trying to get out using uucp there's no way the MX can get you there directly; sendmail will try to connect() to whomever it's pointed to by the MX record. But there's still a way to do this if you designate a gateway host and point to it with the MX; then all he needs to do is understand how to deal with the mail. This may be a better solution in terms of maintenance, and there may well be a better solution than this, but I can't think of it. In any case, I will recommend that AIX include optional rules that pass unknown addresses to a relay host, since this seems to be a popular mail configuration, and we should have a way of doing this without the user having to do major rewrite rule hacking. ------------------------------------------------------------------------------- Peter Jeffe ...uunet!cs.utexas.edu!ibmchs!auschs!sandino.austin.ibm.com!jeffe first they want a disclaimer, then they make you pee in a jar, then they come for you in the night