Xref: utzoo comp.mail.misc:2666 comp.mail.sendmail:1188
Path: utzoo!attcan!uunet!wuarchive!cs.utexas.edu!jarvis.csri.toronto.edu!neat.cs.toronto.edu!rayan
From: rayan@cs.toronto.edu (Rayan Zachariassen)
Newsgroups: comp.mail.misc,comp.mail.sendmail
Subject: Re: How serious is the at->@ problem?
Message-ID: <89Nov25.204944est.2233@neat.cs.toronto.edu>
Date: 26 Nov 89 01:50:53 GMT
References: <3747@ccnysci.UUCP> <7867@ditmela.oz> <3945@amelia.nas.nasa.gov>
Lines: 29

>>You are not allowed to register a domain name that is already used
>>as a higher-level domain name. If the .at domain was registered in the
>>British name scheme (it isn't - sigh), 'at.man.ac.uk' would be illegal.

>It does pose an interesting dilemma to DNS software run in the 
>offending domains. My guess is that it will only be a problem if/when 
>.ca.us and .ca both achieve large Internet populations.

As I recall, it used to be DNS policy (stated in one of the early RFCs)
that within a host one could default domain names by a sliding match (first
common subdomain) on one's own domain name.  For example, if your host was
foo.sf.ca.us you could use glob.la.ca as a shorthand for glob.la.ca.us.
It was always stated that all names leaving a system should be fully qualified.

This policy was one of the major arguments in the CA vs. CA.US battle.

About the time the ISO codes got popular and country-level domains were
getting registered, Mockapetris denounced this mechanism.  There is no
requirement that toplevel domains don't appear at lower levels, its just
good practise to avoid it when possible.

The scheme lives on in the present DNS resolvers, but the sliding window
matching isn't quite the same and it usually stops at the organizational
boundary anyway.

As for ' at ', dealing with that shouldn't be the job of an RFC822 mailer,
it should be done by a protocol translation program.

rayan