Robert Elz <> Wed, 04 January 1995 13:58 UTC

Received: from by IETF.CNRI.Reston.VA.US id aa01819; 4 Jan 95 8:58 EST
Received: from CNRI.Reston.VA.US by IETF.CNRI.Reston.VA.US id aa01815; 4 Jan 95 8:58 EST
Received: from by CNRI.Reston.VA.US id aa06175; 4 Jan 95 8:58 EST
Received: from munnari.OZ.AU by (5.65c/5.61+local-20) id <AA27913>; Wed, 4 Jan 1995 05:41:15 -0800
Received: by (5.83--+1.3.1+0.50) id AA02714; Thu, 5 Jan 1995 00:41:04 +1100 (from kre)
Date: Thu, 5 Jan 1995 00:41:04 +1100
Sender: ietf-archive-request@IETF.CNRI.Reston.VA.US
From: Robert Elz <>
Message-Id: <>
Subject: Re:

    or when a function may be implemented, but if it is implemented, it
    must be done in a certain way for proper operation.

Yes, I realised that one after I had sent my message - I had a may qualifying
a nust or must not (or should), the obvious converse is a must or must not
(or should) qualifying a may, which is your one.

    in addition, we might wish to use r.r. to reinforce the necessary
    requirements of that function.

I'm not sure what you mean here.  Is this kind of similar to Scott's "hint"?
If so, MAY simply doesn't cut it - read the definition of MAY in the draft
(which is the same as in HR).  No-one could reasonably interpret a MAY as
a suggestion to do something if they actually read the doc.

If we want to suggest that a DNS resolver is a good thing then we either
have to go along with mo's suggestion and make it a SHOULD, or invent
new terminology to cover it (or perhaps just explicitly say "it is a good
idea to ..." with no capital letters at all).   If we are going to do that,
then we also need consider whether we actually want to say "implement a
DNS resolver", or "provide a mechanism to translate names to addresses
and vice versa using the DNS", the latter of which does not necessarily
require a resolver actually running in the router.  If we are to mandate
a resolver, then we need consider whether it has to be a full blown resolver,
capable of working in the presence of servers none of which support recursive
queries, or whether it can be a BIND style stub resolver that requires a
co-operative recursive server somewhere on the net.   We may also need to
say something about how it should be configured to know, or learn, the
root servers, and perhaps about DNS security.