Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!ucbvax!VENERA.ISI.EDU!pvm
From: pvm@VENERA.ISI.EDU (Paul Mockapetris)
Newsgroups: comp.protocols.tcp-ip
Subject: Re: RFC 1101
Message-ID: <8905021943.AA21368@venera.isi.edu>
Date: 2 May 89 19:43:27 GMT
References: <364@belltec.UUCP>
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 28

In general, the DNS philosophy for associating new forms of "host
addresses" with hosts names is to enter new data (in the form of RRs)
either with existing IN class information or in a new class.  For
example, the Chaos class was defined to use existing host names and a
new address format in Chaos class RRs.

For example, if we wanted to be able to look up phone numbers given host
names, we might define them as being carried in a TXT RR, or define a
new RR type of, say, PHONE, analogous to the A RR (these RRs would be
stored at the host name).  If we wanted to look up host names from a
phone number, we would need something like the IN-ADDR.ARPA tree (Which
we might be able to get along without, just as we get along without such
a directory in the real world).  From experience, there's a fair amount
of work involved in standardizing the data (e.g. include area code?,
international code?, extensions?), especially if you want an Internet
standard.

RFC 1101 doesn't specifically address (pardon me) this isssue, except to
provide an solution for a different problem and general thoughts.  RFC
1101 deals with the specific question of mapping between network names
and IP network numbers (part of every IP address), and some general
thoughs about mapping between arbitrary things (relevant to the reverse
mapping, a la in-addr.arpa).

Namedroppers is probably a better list for this, and I'm sure there
is a lot of discussion about similar issues in its archives.

paul