[ih] XEROX/PUP and Commercialization (was Re: FYI - Gordon Crovitz/WSJ on "Who Really Invented the Internet?")

[Jack Haverty]

Hi Bill,

We probably ran into each other in some hallway at PARC.  Yes, there
was a lot of cross-fertilization back then - and not just in the
research communities.

I think the government played a role in "internet technology" that was
much more than "fund research and get out of the way".  It proactively
constructed the way, filling holes and removing obstacles to create
"the way" which then was the easiest (and most profitable) path for
others to take.   I've always wondered if that decades-long activity
was orchestrated or just happened by a collection of smaller random
quantum events.

I experienced a few of these government efforts.  No doubt there were
many many others.  Here's some I can remember.

The government:
- funded the creation of the Arpanet as a research tool, and the
deployment of multiple Arpanet clones in operational environments
- funded software development of TCP implementations for a wide
variety of computers, making sure there were equipment choices
available that could use the Internet.
- established TCP/IP as an official DoD Standard, much to the chagrin
of many researchers (guilty!) who didn't think it was ready (there was
always a never-shrinking list of issues to be researched)
- directed work at government labs into the TCP-based technology.  I
recall discussing the IP "source quench" with someone working on TCP
at the Army's Ballistic Research Lab, and the difficulty of getting a
host to actually stop sending traffic.  We joked about "Ballistic
Source Quench" as a possible technique using some ordnance they had
lying around.
- applied procurement rules so that contracts for any project using
computers had to be able to use TCP.  In other words, the government
created a market, into which the Internet industry could grow, and
guaranteed at least one huge customer.  It thereby motivated the
broader technical industry to figure out how to incorporate TCP into
their products and services, or risk losing future government
contracts.  It removed the ability for different parts of the
government (DoD) to make different choices of technology; if it
communicated, it had to be TCP.
- adopted the Internet TCP/IP technology as the basis for its own
operational networks.  In marketing lingo, this is "eating your own
dog food", and shows that the technology not only works, but is
trusted for non-research uses.
- modified the Arpanet, in January 1983, so that only computers using
TCP would be able to communicate, after a long warning period; funded
development of tools and services to assist government systems
contractors in preparing their existing operational systems to survive
the transition
- scuttled the pre-existing plans to implement the next generation DoD
communications system called Autodin-II, and contracted instead to
build the Defense Data Network, using proven Arpanet technology, which
made a clear path to internet evolution
- funded the definition and development of operational rules,
procedures, mechanisms, and techniques to enable the reliable
operation of non-research networks by people who weren't the ones who
wrote the code.  This was crucial to scaling the Internet.
- proactively marketed the Internet technology within the government
into non-research organizations, who had been using their own "Arpanet
clone" networks and faced the same challenges of communicating across
multiple networks.  I recall one meeting where Vint presented the
Internet slideshow to the guy in charge of a very large existing
non-research network with many hosts.  When the slide showing the
current Internet map - the standard boxes-and-lines graphic - was
shown, his reaction was "That's it!  I want one of those."  His aides,
sitting behind him, were taking notes.  He got his own TCP-based
internet.  There were other such vignettes.  Such marketing created a
growing demand for commercial TCP-based products and services and
helped establish an Internet industry.
- marketed TCP into other governments, especially in Europe, through
collaboration and joint projects
- by instilling TCP technology into a wide range of educational
institutions, e.g., through NSFNet and CSNet, created a large and
growing pool of skilled labor, emerging from school for employment by
the technology industry as well as institutions using of TCP-based
systems.  These new employees usually didn't know how to use other
internet technologies, but were well-versed in TCP.
- by establishing the Internet as a visible, reliable, and available
networking approach, created demand from non-governmental early
adopters.  I did some work in the late 80s with companies on Wall
Street, who were feverishly building their own Internet systems.  They
chose TCP because it worked, did not lock them into any single vendor,
and most importantly it allowed them to make more money.

At that point, the government didn't have to "step out of the way".
It just became one of the many, many customers.

IMHO, none of the activities I've listed above would be categorized
typically as "funding research".   They are more like the steps a
company follows to create a successful product-line and crush the
competition.   With 2 billion users now, and all the competitors gone,
it seems they did a hell of a job too.  I still wonder if this was
orchestrated and if so by whom.

If Xerox, or IBM, or Novell, CCITT/ISO, or any of the other
contemporary competitors building internet technology had done the
same things, The Internet today might be running on X.25/X.75, or
carrying PUPs, or using SPX/IPX, instead of TCP/IP.

But they didn't.  The government(s) did.

/Jack Haverty





On Sun, Jul 29, 2012 at 10:07 AM, Bill Nowicki <winowicki at yahoo.com> wrote:
> Most of this has been discussed much already, but I just wanted to add my
> chorus to those who point out how much cross-fertilization there was in
> those days. Remember Xerox PARC is on Stanford land, and not only did many
> of us work as summer interns, they gave us "no fee consulting" agreements so
> that we could bicycle up the hill to attend "dealers" that sounded
> interesting and collaborate on other projects as Xerox people would also
> listen in on seminars. People like Dave Boggs (Ethernet co-inventor) and
> John Shoch were Stanford graduate students, and many others were students of
> CMU, MIT, Berkeley, etc. which were all ARPA contractors. The ARPA-funded
> (along with other government agencies) projects at SRI like NLS were also
> clearly connected by some of the same people.  Xerox even designed and built
> a computer (jokingly called "MAXC" which makes another great anecdote)
> specifically to run the TENEX operating system of BBN and put it on the
> ARPANET.
>
> So as usual real history is more complicated that pundits like to convey.
>
> It might be safe to say that PUP was the first working corporate internet
> ("intranet" was not coined yet). It did have multiple different computer
> architectures and network technology from the start, but did not have the
> multi-organizational issues of the ARPA/netInternet. To follow PUP is also
> interesting. I was hired by Xerox briefly as a consultant to help a few
> researchers run a version of PUP we did for BSD Unix. I do  not think 3Com
> ever did sell PUP. Xerox tried to come up with a new generation protocol
> which they called Xerox Network Systems which is probably what was meant.
> XNS used 48-bit addresses, while PUP used only 16 bits and TCP/IP used 32
> bits. The irony was that Novell tried to implement XNS from the
> specifications, but made a few byte-swapping errors. This Novell variant was
> arguably the most commercially sucessful internet protocol as part of their
> NetWare product during its short heyday in the 1980s.
>
> So spinning this history to either the political left or right is clearly
> misleading. Internet technology is a very good example of what government's
> role should be: fund the research, then get out of the way and let the
> market competitive forces drive products. The more interesting debate is
> what the government role is later after the technology matures. Do they
> subsidize the entrenched interests who can afford lobbyists (the phone
> company and cable TV model), or act as referee to attempt to keep the
> marketplace competitive (at the expense of perhaps hampering the
> innovation). But those are opinions creeping in again.