[ih] Internet History - Commercialization (was Re: When did "32" bits for IP register as "not enough"?)

[Jack Haverty]

[Thread renamed to better reflect topic...]

I agree with all of Vint's thread, but I'd like to add to it.  I think
there were several other situations and activities that I think also
drove commercialization of "The Internet"

1) The Proliferation of Private internets (aka Intranets)

2) Parallel Creation of Other internets (aka Othernets)

3) Deployment of Multi-Protocol internets (aka Multinets)

4) Establishment of a Internet Technician pipeline

5) Adoption of Web technology as a Client/Server standard

Intranets

While "The Internet" was growing, and steered by decisions by
organizations such as NSF, NIST, etc., the same hardware and software
was used to build private internets, totally disjoint from "The
Internet".  This probably occurred first in governmental arenas, where
it was straightforward for a non-research group to simply buy some
routers and hook them together as a private intranet.  The same was soon
true in commercial environments.  I personally encountered this in
1989-90 while consulting for a large Wall Street investment house, who
were struggling to deploy their own TCP-based intranet connecting LANs
and Sun workstations in New York, London, and Tokyo.  But there were
many types of Intranets, not based on TCP, because of Othernets.

Othernets

While "The Internet" was evolving, other organizations were building and
promoting their own technology.  Some of these were driven by "standards
bodies" like ISO or CCITT.  Others were driven by corporations who each
had their own version of a "global network".  IBM, DEC, Xerox, Wang,
Novell, Banyan, Apple, and probably others I've forgotten all had their
own technology for creating a global network.  All of these would work
great in a global installation, at least according to Marketing,
especially if you only used their hardware and software.

Multinets

With so many choices, people made different decisions.   Engineering
seemed to like DECNET.  Finance was true-blue with IBM SNA.  Marketing
often had Appletalk.  Departments with lots of PCs had multiple choices,
e.g., Novell, Banyan, etc.  The weird guys in the Lab had something
called TCP (which, if you were in the UK, was a mouthwash).  There were
X.25 networks in use, and perhaps someone was talking about ISO for the
long-term solution, but they were still waiting for something they could
install.

Multi-protocol routers, bridges, and other such pieces of the Tinker Toy
networking kit allowed you to run as many different Othernets as you
liked, all over the same physical wires and circuits.  Instead of
picking one of the choices as your corporate standard, you could just
have them all.  You could have your own private Multinet.

The Technician Pipeline

With all of these different network technologies and products, staffing
an IT department was a formidable challenge.   You could hire people,
and them send them to the various vendors' training to learn how to use
each particular technology and products.  Of course everyone would need
retraining frequently as all the technologies were evolved.   And then
the best people might disappear, lured off to some other IT department.

The exception was the TCP world.  Somehow it seemed that new graduates
were continually emerging from Academia already trained and experienced
in using TCP-based systems.  There seemed to be a pipeline through the
schools around the world that was pumping TCP-qualified potential IT
staff out into the world, who could get a TCP-based system up and
running while other technologies advocates were still in networking school.

Web Technology as Client-server Standard

By the early 90s, the networking started by the ARPANET had evolved for
20 years.  But over those decades, the search for the next "killer app"
hadn't produced much.  File transfer, remote terminal access, and
electronic mail were still the primary "apps" used broadly, and had been
for 20 years.

With the advent of The Web, that changed in the early/mid 90s.  Suddenly
there was a new "protocol on the wire" technology that enabled people to
create and consume content, without the requirement to have the
identical word-processing, graphics, or other "user interface" software
on every machine.  The first new "killer app" in 20 years had appeared.

In the general business world, people wrote and read documents, but they
also did a lot of other tasks with their IT systems.  They did inventory
control.  They ran analyses, in financial, sales, marketing,
engineering, and every other business discipline.  They modelled
physical systems such as weather or economics.   They did everything
associated with their business operations, using computers.

Although the Web technology began as a way to collaborate on documents,
it had the requisite "hooks" to enable "documents" to be generated on
the fly, tailored by instructions gathered from the human user who would
be viewing the result.   This enabled many of the broader business tasks
to be done using Web technology.   Instead of viewing a Quarterly
Analysis using an application on your workstation, you could view that
same analysis as a "web page" that had just been created by a program as
you just specified.

In effect, the Web technology brought a wire-protocol standard for
"Client-Server" interactions to the TCP technology arsenal.   That was a
capability none of the "Othernet" players had.

======================================================

All of the above reflects only what I personally experienced.  So
there's probably a lot more that was going as well that should be
included when looking at the Commercialization aspect of Internet History.

Note that everything I just mentioned happened outside of the topics
that are most often discussed here as "Internet History" - RFCs, IENs,
IETF, IAB, ISO, NIST.   Those were all important, but I think there was
lots of other things going on, like the ones I listed, that also
strongly influenced the growth of the Internet.

It seems to me that there's quite a lot of material recorded in things
like the RFCs and even academic papers, but very, very little about the
myriad "Othernet" technologies, or about what was going on in the "User"
(i.e., Commercial) business world as it struggled to actually use all of
this stuff. 

Back in the early 90s, Oracle had recruited me as "Internet Architect"
with the rather vague direction to "get us into the Internet
business".   I recall one meeting, circa 1991-2 (pre-Web), where I
finally realized that TCP had won. 

We had a "Customer Council" which was composed of senior staff (CIO,
CTO, etc.) of a diverse group of customers.  They were all from
non-techology companies, e.g., Finance, Manufacturing, Transportation,
etc., and from different countries and continents.  Their common mission
was to figure out how to use IT technology in their business operations.

One of our meetings focussed on networking.  We went around the room,
and everyone briefly described their existing networking.  There were
"IBM shops", "DEC shops", and the like, but they all had smatterings of
other stuff, e.g., PC-based LANs of some ilk, etc.  They were all
becoming de facto Multinet IT departments.   More accurately, they were
struggling to get it all to work.

We went around the room again, and got everyone to describe their
corporate strategy - their vision of where they wanted to get to.

I was absolutely blown away by the response.  Several dozen large
corporations, in many industries, across many geographies ... every one
of them said their future target was TCP.  It might take them years, but
that was the destination.

I had to ask "Why TCP?"....

There were a few common themes in the answers:

- This multi-technology stuff is driving us nuts. 

- TCP works.  We can see it running in the Internet (e.g., at Interop). 
We have it running in our lab.  We can see other people using it
successfully.  It's here.  We can use it.

- TCP is under our control.  [Bet you thought IETF was in charge...] TCP
is an open de facto "standard", unencumbered by secrecy, patents, or
vendors' business strategies and changing plans.  We can buy products
from many vendors, and we could, if we really had to, even build missing
pieces ourselves.

- TCP is well-supported.  In particular, we can hire people, even just
out of college, who already know about TCP and often already have
several years operational experience with it.  They "hit the ground
running" in our IT department, saving us lots of time, and money.

No one brought up anything resembling a technical issue.  No "the XYZ
routing protocol is better than the "JKL" one, or anything like that.

These themes tie back to events in the technology world.  For example,
DARPA's choice to make TCP technology "open", and to create free
implementations, coupled with NSF, CSNET, and other decisions to inject
TCP into the academic machinery were likely crucial in creating that
"pipeline" of qualified IT staff.

Similarly, the technology industry's decision to proliferate competing
technologies, and never coalesce into a standard, put the user
communities in the position of having to deal with Multinets, which were
a world-class arena for vendor finger-pointing.   If you think building
an operating a TCP Intranet is challenging, think about a Multinet (been
there, done that...)

This led to the User community taking the lead, if inadvertently and
reluctantly. E.g., the Web came from a User environment, driven by the
need to share documents and data.

===================

There was really a lot that happened outside of the more visible
activities of the technical and standards worlds of IETF et al, much of
which had significant effects on the growth and evolution of the
Internet.  I only saw a small part of course, and haven't run across
much else written about the networking experience in the "real world"
back in the 80s/90s and even today.  Maybe I really should write down
some more of at least what I remember.....

/Jack Haverty


On 2/15/19 1:29 PM, Vint Cerf wrote:
> Commercialization of the equipment and software arose from a different
> thread:
>
> 1. IBM, Digital and HP all implemented TCP/IP on their commercial
> operating systems - but it was their Research Groups
> who did that (I encouraged this).
> 2. I also encouraged the UNIX TCP/IP development at 3COM for which
> Metcalfe never forgave me because after
> he implemented that offering, the Berkeley release came out for free
> 3. INTEROP made Internet visible to a much larger, non-academic
> audience, had training sessions and allowed
> a lot of product vendors to demonstration the interoperability of
> their software/hardware - a major sales point 
> when you are trying to decide what to buy. 
> 4. there were very few implementations of OSI and none that I know of
> were commercially successful
> 5. In 1992, NIST was persuaded to do an analysis of TCP/IP and OSI and
> concluded that it was OK
> for government users to procure TCP/IP despite previous guidance to
> use OSI according to the
> Government OSI Profile (GOSIP).  
> 6. MOSAIC hits about 1993 followed by Netscape Communications and its IPO.
> 7. Cisco, Proteon and later, Juniper, produce commercial routers. Sun
> Microsystems produces Work stations and they
> all use TCP/IP.
> 8. Novell tries to use IPX and XNS but just doesn't have the traction. 
>