Reed's views, was [e2e] Cannara's views

Cannara cannara at
Wed Apr 25 09:02:19 PDT 2001


I imagine most of us have read "A Brief History of the Internet", "Scaling the
Internet During the T3 NSFNET Years", etcetera.  The focus of many items, like
addressing, becomes clearer after reading as many of these pieces as
I agree with your last paragraph, because zero-cost inclusion of any product
within a system bought in large volumes, for wide dissemination, is an
excellent way to market anything.  Having a free protocol stack on any Unix
machine from any vendor is very different from committing to DECnet and a
single system vendor.  This formidable marketing event (assuming no strategy),
drove TCP/IP into the commercial world, as it did in the academic one.

However, every one of the Sun (etc.) boxes that came with free IP also had at
least one Ethernet interface, with a 48-bit DLC address.  Choosing not to use
this as part of a truely-unique Net-layer address, was a surprising oversight
-- an oversight that more than a decade ago had not been made by Novell and
others.  When teaching classes that cover networking over the last 30 years,
students raise interesting questions about how the most common networking
protocol today could have such an inadequate address space, despite years of
warning, all the while enjoying both public subsidy and industry-product


John Day wrote:
> One other thing I was going to mention last night and forgot.
> As to the "short-sightedness" of 32 bit addresses and why would we be
> so stupid when we had Ethernet as an example.  The answer is quite
> simple.  Ethernet was already a commercial effort.  The Internet
> wasn't and was far from it.  If you had suggested to anyone in 1976
> or so that we would be able to take a military/academic research
> network into the commercial space untouched, everyone would have
> figured you were smokin' something you shouldn't have.  All of the
> examples we had to that time of taking things out of research into
> the commercial world involved building a duplicate from scratch.  The
> idea that we could transition the network as it was to the commercial
> world seemed a remote highly unlikely possibility.  So given the
> perceived scope of the Internet at the time 32 bits were chosen, 32
> bits was overkill.  In fact, if you want to get an indication of the
> thinking at the time, go to the IANA sight and look at who has Class
> A's assigned to them, and think how we might do it now.  VERY
> different ideas there.
> In fact, even Ethernet reflects the state of our understanding at the
> time.  Then we didn't understand that addresses only had to have the
> scope of the layer in which they were used.  I really doubt that
> there will ever be an Ethernet segment or even a bridged ethernet
> with 2**48 devices or anything remotely close to it.  16 bits would
> have been more than enough.  What is interesting is that IEEE 802
> still hasn't figured it out.  Look at firewire addresses.
> And when it really comes down to it, do we really think that our
> brilliant technical insights had anything to do with the success of
> TCP/IP?  Or did it really have more to do with economics and politics
> and timing?  A hard look at the events would seem to indicate that
> things outside the technical had more to do with the current success
> of the Internet than our technical solutions.
> Take care,
> John

More information about the end2end-interest mailing list