woods@ncar.ucar.edu (Greg Woods) (12/06/89)
We are about to split our domain up into subdomains, and I am trying to make the necessary changes in the sendmail configuration to account for this. In particular, I want to make it unnecessary for users in one subdomain to fully qualify every name when sending mail to a host in a different subdomain. I'd like it to work the way all the other applications (e.g. telnet, ftp, rlogin) do. I can say telnet host.sub (i.e. as opposed to telnet host.sub.ucar.edu) and the resolver happily finds the right host. So I would like to also be able to send mail to user@host.sub and have it work. By adding this rule in ruleset 3 R$+<@$+>$* $1<@$[$2$]>$3 the mail will get there. The problem is that the To: line still reads like To: user@host.sub (i.e., it doesn't get properly canonicalized). This is all well and good unless a message is sent with this kind of address and is also sent to someone outside our domain. If that outside person tries to reply to all recipients, the original user@host.sub cannot be resolved and the message bounces. What I *really* don't understand about this is that if I run sendmail in address test mode, and pass it through rulesets 2,R,4 (which are supposed to process the To: line), it DOES get properly changed into user@host.sub.ucar.edu since ruleset 3 is always called first. Summary: with a name server lookup in ruleset 3, the envelope sender is properly resolved. In address test mode, using the rulesets which supposedly process the To: line, it is properly resolved. However in practice when I actually send mail, the To: line is left unresolved. Anybody have any idea what I am missing or overlooking? I've been working on this for several days and have not been able to solve this problem. --Greg
ehrlich@cs.psu.edu (Daniel Ehrlich) (12/06/89)
In article <5565@ncar.ucar.edu> woods@ncar.ucar.edu (Greg Woods) writes:
Greg> We are about to split our domain up into subdomains, and I am trying to make
Greg> the necessary changes in the sendmail configuration to account for this. In
Greg> particular, I want to make it unnecessary for users in one subdomain to
Greg> fully qualify every name when sending mail to a host in a different subdomain.
Greg> I'd like it to work the way all the other applications (e.g. telnet, ftp,
Greg> rlogin) do. I can say
Greg> telnet host.sub
Greg> (i.e. as opposed to telnet host.sub.ucar.edu) and the resolver happily finds
Greg> the right host. So I would like to also be able to send mail to user@host.sub
Greg> and have it work. By adding this rule in ruleset 3
Greg> R$+<@$+>$* $1<@$[$2$]>$3
The easiest way I know of to do this is have CNAME records in the outer
domain UCAR.EDU that point into the subdomain. Something along the lines
of:
$ORIGIN ucar.edu.
host IN CNAME host.sub
in the name server files for ucar.edu would do. This means that even though
you have sub-domains the host names must still be unique across all of
ucar.edu.
--
Dan Ehrlich <ehrlich@cs.psu.edu>
Quote of the month:
"And I have quite a bit of experience on mainframes. I learned C, Pascal,
and Assembler on various mainframes, including PDPs, VAXen, an IBM AS/9000."
-- Charles M Hannum II <cmh117@psuvm.psu.edu>ecf_hap@jhunix.HCF.JHU.EDU (Andrew Poling) (12/07/89)
In article <5565@ncar.ucar.edu> woods@ncar.UCAR.EDU (Greg Woods) writes: [...] >By adding this rule in ruleset 3 > >R$+<@$+>$* $1<@$[$2$]>$3 > >the mail will get there. The problem is that the To: line still reads like >To: user@host.sub (i.e., it doesn't get properly canonicalized). This is all >well and good unless a message is sent with this kind of address and is also >sent to someone outside our domain. If that outside person tries to reply >to all recipients, the original user@host.sub cannot be resolved and the >message bounces. What I *really* don't understand about this is that if >I run sendmail in address test mode, and pass it through rulesets 2,R,4 >(which are supposed to process the To: line), it DOES get properly changed >into user@host.sub.ucar.edu since ruleset 3 is always called first. [...] >Anybody have any idea what I am missing or overlooking? I've been working on >this for several days and have not been able to solve this problem. Why not put this rule in ruleset 4 too. That way, every To: and From: address should be properly canonicalized. That's what ruleset 4 is for - to put addresses in external form (i.e. canonicalized). Hope this helps, -Andy -- Andy Poling Internet: andy@gollum.hcf.jhu.edu Network Services Group Bitnet: ANDY@JHUVMS Homewood Academic Computing Voice: (301)338-8096 Johns Hopkins University UUCP: mimsy!aplcen!jhunix!gollum!andy