Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!samsung!uakari.primate.wisc.edu!ames!eos!shelby!portia!jessica!morgan
From: morgan@jessica.Stanford.EDU (RL "Bob" Morgan)
Newsgroups: comp.protocols.tcp-ip
Subject: Re:  tcp port numbers
Message-ID: <8338@portia.Stanford.EDU>
Date: 17 Jan 90 20:12:18 GMT
References: <9001161022.aa27296@SEM.BRL.MIL>
Sender: USENET News System <news@Portia.stanford.edu>
Reply-To: morgan@jessica.Stanford.EDU (RL "Bob" Morgan)
Organization: Stanford University
Lines: 31


Terry asks:

> Personally, I don't see why the existing registration procedures can't
> be applied to all port numbers.  Are there any significant problems in
> doing this?

As various other postings about inconsistent use of port numbers on
different systems should make clear, we have got to the point where
maintaining a universal static list mapping service names to port
numbers is as unappealing and difficult as maintaining a universal
static list mapping host names to IP addresses.  It's an inevitable
consequence of the growth in size and diversity of the Internet.  The
Numbers Czars seem to have stated their limited interest in
maintaining such a list by limiting the "official" range to 0-256.
(Maybe they're thinking of expanding this?)

Clearly, what's needed is a standard mechanism for hosts to access
services on other hosts without knowing in advance what ports those
services are on.  One such mechanism is proposed in RFC-1078, TCPMUX,
but unfortunately it only applies to TCP-based services.  Sun's
portmapper provides this function for services using their RPC
protocol.  My impression of Hesiod, MIT-Athena's extension to domain
name service, is that it performs this function among others.  I
imagine that such a mechanism could be based on ISO Directory services
whenever they become widely available.  In the meantime, we'll suffer
along as usual.

 - RL "Bob" Morgan
   Networking Systems
   Stanford