[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[ih] When did "32" bits for IP register as "not enough"?

On Thu, Feb 14, 2019 at 2:58 PM Craig Partridge <craig at tereschau.net> wrote:

> Dennis' decision to adopt TCP/IP for NSFNET was critical.
Hmmm.  It certainly did not hurt and I agree was cleafrly an amplifier and
important.   But I'm not sure that was as important as the basic economics
of the time -> Moore's Law and cheap cycles (i.e. I don't think the NFSNET
choice, while helped the effect, was not the high order bit on the success

This was the time of the beginning of the workstation wars and the dominate
OS was UNIX and in particular BSD which came with a known working IP stack
(as I said, the USG paid to have that written by BBN).   Other stacks were
for BSD were available on the market and there were even OSI stack
implementations to be found.

But the code that was being written at the time assumed BSD sockets, which
most naturally mapped to (assumed) IP/TCP.  I think the >>user<< code does
not get enough of the credit here.  Because it was that code that provied
the services that we think of as the 'value.'   I think the NSFNET choice
was driven by the cheap Unix implementations not the other way around.  But
the HW and SW just used sockets (and TCP under the covers) to do real
work.   Once those services were there, and they were interconnected, the
protocol that provided it all took off.

If a different system had been as cheap to use and used OSI (or whatever),
I think that would have been the winner.  IP road the success wave of the
cheap cycles that the UNIX workstations created and provided by Moore's law.

I really believe that it Metcalfe's Law made those system more and more
valuable because they could be interconnected and the services agregated,
but the systems had to start to be in place to get the effect started.
That said, the two processes played on each other.   The Internet boom
absolutely needed cheap HW and SW to be there so that the Noel's
observation about the phone to be true.

Looking at it from a different and related technology.   AT&T made the
cheap phone in the 50s and 60s and Metcalfe's law made it useful (remember
AT&T spent $200 per American household in 1959 dollars to install that
black hand set.  But once it was there, then magic happenned).   The
Internet and IP/TCP needed the moral equiv of cheap cycles to be
ubiquitious before it would dominate.

  ... TCP/IP-based networking engineering expertise and products.  By
> 1988/1989, the market for TCP/IP products was much bigger (and visibly so)
> than OSI and we had (a small number of) thousands of network engineers who
> understood how to put together and operate TCP/IP networks.
That's because of Moore's law, UNIX the cheap workstation etc....

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://elists.isoc.org/pipermail/internet-history/attachments/20190215/0403b98a/attachment.html>