Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!gatech!purdue!bu-cs!dartvax!eleazar.dartmouth.edu!steve
From: steve@eleazar.dartmouth.edu (Steve Campbell)
Newsgroups: comp.mail.sendmail
Subject: Re: Second shot after TCP fails?
Keywords: SMTP nameserver UUCP Smail
Message-ID: <14095@dartvax.Dartmouth.EDU>
Date: 26 Jun 89 23:09:31 GMT
References: <2313@ccnysci.UUCP>
Sender: news@dartvax.Dartmouth.EDU
Reply-To: steve@eleazar.dartmouth.edu (Steve Campbell)
Organization: Dartmouth College, Hanover, NH
Lines: 23

In article <2313@ccnysci.UUCP> jeffrey@ccnysci.UUCP (Jeffrey L Bromberger) writes:
>...Let's say I send mail to
>user@foo.bar.com, and sendmail's rulesets decide that this should go 
>via SMTP.  The mailer tries to resolve the name foo.bar.com by 
>talking to the appropriate nameservers on the network, but finds out 
>that foo.bar.com is not a "fully qualified domain name" by network 
>standards.  Is sendmail bright enough to try the UUCP approach (send 
>the name off to smail for resolution) after the SMTP connection fails?  

No.  Once ruleset zero resolves to a particular mailer - the TCP mailer
in your example - then there is no turning back.  If that mailer comes
up with a fatal error such as unknown host, then the message bounces.

While this may be a shortcoming in some instances, the particular situation
described here "should not happen."  This is because if foo.bar.com is
a legitimate, registered domain name, then there should be a mail forwarder
for it on the Internet as well as a name server on the Internet to put
forth MX records pointing to the forwarder.  This should be true even if
foo.bar.com is a UUCP-only site with no Internet connections.  If all this
is not the case, then people shouldn't be using foo.bar.com in email.

							Steve Campbell
							Dartmouth College