Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!uwm.edu!uakari.primate.wisc.edu!ginosko!usc!ucsd!ucbvax!NSIPO.ARC.NASA.GOV!medin
From: medin@NSIPO.ARC.NASA.GOV ("Milo S. Medin", NASA ARC NSI Project Office)
Newsgroups: comp.protocols.tcp-ip
Subject: Re: Pilot Project: Hosts Table To Be Available Over Internet
Message-ID: <8910220212.AA28259@nsipo.arc.nasa.gov>
Date: 22 Oct 89 02:12:53 GMT
References: <360@nisca.ircc.ohio-state.edu>
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 37


This doesn't make sense.  Let's say you go out and map out the whole
Internet DNS space by walking the tree (let's ignore the problem of
servers who refuse zone XFER requests for now).  You may have the
equivalent of all the A records in the DNS, but you have missed out
on all the MX records out there that many domains depend upon for
proper mail transfer.

As an example, you can send mail to user@host.span.nasa.gov, and that
points to an Internet<->DECNET mail relay.  There are oodles of hosts
that are serviced by this relay, and the DNS doesn't know anything
about those DECNET machines out there, so you can't get that information.

There are also many sites who have MX records that point to a mail relay
machine for local internal distribution.  Even for hosts that are 
Internet capable, but may not have advanced or properly configured mailers.
In this case, you'll try and deliver directly rather than go through the
relay like a DNS capable system would have.  

The DNS does more than just hosttable functionality.  People even use this
functionality!  If the host doesn't support DNS functionality, don't
attempt to use it as a general purpose Internet host.  Certainly,
don't attempt to mail from it!  Most major hardware vendors support the DNS.
I can't think of any that have no support.  Even the PC's and Mac's support
it.  Just exactly what systems out there are you trying to build this for?
DNS support is required, not just recommended.  Building in resolver
support is trivial if you have UDP support.  You can use some nameserver
out there on a real machine if you can't run one yourself.

This seems like a very bad idea, not just because it goes against current
good practices, but that people may actually think they can get by
with this, because the tables puportedly have all the hosts in them.  We have
enough problems to debug in the Internet already.  We don't need needless
ones.

						Thanks,
						   Milo