Errors-to: owner-drums@CS.UTK.EDU
Message-Id: <199505311820.OAA00828@augusta.math.psu.edu>
To: drums@CS.UTK.EDU
Subject: mail name rules
In-reply-to: Your message of "Tue, 30 May 1995 15:42:14 +0200."
             <199505301342.PAA06159@dale.uninett.no> 
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Date: Wed, 31 May 1995 14:20:48 -0400
From: Dave Barr <barr@math.psu.edu>

In message <199505301342.PAA06159@dale.uninett.no>, Harald.T.Alvestrand@uninett
.no writes:
>About the hostname stuff: I think the new docs should clearly identify
>that hostnames are legal DNS names, and leave it at that; it might
>copy a few of the rules for reference (like the charset limitations),

There's actually quite a difference between 822 and 821.
822 is very liberal.  The characters ``!#$%^&*_/\'' are legal in
an RFC 822 address according to the BNF.  So in theory if you
weren't using SMTP you could have some pretty bizzare addresses.
Or if the addresses you were using were referring to Internet hosts,
you'd fall under the Internet hosts naming requirements in 1035
(as amended).

>but should otherwise refer to the DNS documents for the definition.
>(the under_score allowed by BIND *is* a bug, IMHO.)

Actually I think the reverse is true.

It's not BIND's job to conform to hostname standards.  BIND's job
is to conform to DNS domain name standards, of which there are very few
restrictions.  (except for the case-insensitivity factor, you can
have virtually anything in a domain name)  While the vast majority of
people out there use BIND for _hostname_ translations, it is used for
other purposes.

Now you can argue that the LHS of an MX should conform to 821/822,
for example, because that is a clear application of BIND to refer
to a mail name.  However if we put such checks into BIND it makes
things very hard to change them later on.  I'd have no problem with
someone modifying BIND to log warnings about illegal hostnames/mail
names, however.

This ``what's in a name?'' confusion really has to be clarified,
and IMHO relaxed.  It apparently was the intention from the start
to work towards relaxing the naming requirements.  Unfortunately
some big name companies write software which is really anal to the
standards which makes such relaxations very hard.  One of the
tenets of the ``be liberal in what you accept...'' philosophy is that
it opens the way for relaxation of the rules.  (that's one reason
why the leading digit rule was able to be changed, for example)

--Dave