Errors-to: owner-drums@CS.UTK.EDU
Message-ID: <8jp6oeC00WBw09xmxd@andrew.cmu.edu>
Date: Tue,  6 Jun 1995 10:58:50 -0400 (EDT)
From: John Gardiner Myers <jgm+@CMU.EDU>
To: drums@CS.UTK.EDU
Subject: Re: address syntax
In-Reply-To: <2337.802412153@munnari.OZ.AU>
References: <2337.802412153@munnari.OZ.AU>
Beak: Is

The 822 domain-literal syntax is indeed very general, it permits such
things as:

[\1\2\8\.\2\.\1\0\.\1\0\1]

[3(4@\]:,a"$!\\;]

However, I seriously doubt that much, if any, of the installed base
can correctly parse such things.  There are enough problems getting
implementations to correctly parse group syntax.

I'm about as much of an RFC lawyer/syntax weenie as they get, but I
think here's where we need to get a good dose of pragmatism.  This is
a dark corner of the syntax that hardly anybody gets right.  The spec
is way out of line with respect to reality.


With respect to the idea of allowing IPv6 domain-literals in
addr-specs but not routes, this would tend to complicate things by
requiring two definitions of domain.  Just like many implementations
miss the fact that unquoted dots aren't permitted in a phrase, I
suspect many will miss the fact that IPv6 domain-literals are not
permitted in a route.  It will become yet another dark corner of 822
address syntax.

There's also the fact that the colons are likely to confuse parsers
looking for RFC 822 group syntax.  I'm unsure how much of a problem
this really is.

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