Newsgroups: comp.protocols.tcp-ip
Path: utzoo!henry
From: henry@utzoo.uucp (Henry Spencer)
Subject: Re: ToasterNet (was Re: Running out of Internet addresses?)
Message-ID: <1988Dec2.191340.8180@utzoo.uucp>
Organization: U of Toronto Zoology
References: <8811281821.AA00300@bel.isi.edu> <207@logicon.arpa> <1010@asylum.sf.ca.us> <729@lts.UUCP>
Date: Fri, 2 Dec 88 19:13:40 GMT

In article <729@lts.UUCP> amanda@lts.UUCP (Amanda Walker) writes:
>... He described Ethernet addresses as having
>been "designed to be intergalactically unique."
>
>The biggest problem, I think, is that 32 bits (or 48, or whatever) is
>certainly big enough to serve as a *physical* addressing scheme, but
>we keep chopping up addresses so that we can have a *logical* addressing
>scheme...

Another thing that Xerox arguably did right with Ethernet:  the 48-bit
addresses are *not* chopped up this way.  They are in fact divvied up
by manufacturer, but manufacturers are supposed to use their part of
the space completely before getting another one.  The Xerox notion is
that the 48-bit address provides a unique identifier for the destination,
and any information needed to efficiently locate said destination should
be supplied separately.  (As I recall, XNS adds a 16-bit network number
as a routing hint.)  The paper some years ago in SIGCOMM [grr, I just
realized I don't have an exact reference handy] on the design of Ethernet
addressing should be required reading for anyone thinking about this.
-- 
SunOSish, adj:  requiring      |     Henry Spencer at U of Toronto Zoology
32-bit bug numbers.            | uunet!attcan!utzoo!henry henry@zoo.toronto.edu