Errors-to: owner-drums@CS.UTK.EDU
Message-ID: <sjnlbfm00WBwE9xsIM@andrew.cmu.edu>
Date: Fri,  2 Jun 1995 10:02:19 -0400 (EDT)
From: John Gardiner Myers <jgm+@CMU.EDU>
To: drums@CS.UTK.EDU
Subject: Re: address syntax
In-Reply-To: <199506020938.LAA12390@dale.uninett.no>
References: <199506020938.LAA12390@dale.uninett.no>
Beak: Is

Harald.T.Alvestrand@uninett.no writes:
> I think you mean that the construct
> 
> @[ip:v:6]:user@[ip:v:6]
> 
> will be awfully hard to get right?
> If the BNF says that [] are quoting the construct, I think it should
> not be impossible (but of course it breaks existing heuristic parsers...)

It breaks existing correctly written deterministic parsers.

> A mailer SHOULD NOT attempt to "clean" the address syntax of nonlocal
> addresses; such behaviour will often make worse problems than leaving
> them alone."

This directly contradicts the directive in section 5.3.3. of RFC 1123:

	If the address is an explicit source route, it
        SHOULD be stripped down to its final hop.

There are delivery agents widely deployed in the internet which
implement this directive by searching for the first colon in a
route-addr (when it starts with "<@") and treating the part after the
first colon as an addr-spec.  This is correct because a domain (and
thus a route) cannot contain a colon.

If the syntax of domain is loosened up to permit colons, such
previously correctly written agents will then turn route-addrs with
colons in the route into chopped liver.  This is completely, flat-out
unacceptable.

-- 
_.John G. Myers		Internet: jgm+@CMU.EDU
			LoseNet:  ...!seismo!ihnp4!wiscvm.wisc.edu!give!up