BBN JOB #41540
3-1
"BBN Job #41540"
Bob Kahn Interview
August 10, 1994, Tape Three
-- CARBINER GROUP
QUESTION
BOB KAHN
Okay, I'm Bob Kahn,
I'm currently the president
of the Corporation for National Research
Initiatives. Uh,-4969, Ivas uh, the senior
scientist at.oprack(?) and Neumann(.,and i
had, uh, just begun working on the Arpinet
Project.
QUESTION
BOB KAHN
I'm going to go back to your original question.
QUESTION
BOB KAHN
My name is Bob Kahn. i'm currently President of
the Corporation for National Research Initiatives,
-----------------------------------------------------------
BBN JOB 1141540 3-2
QUESTION
BOB KAHN
You mean, the camera, it's no good?
QUESTION
BOB KAHN
Okay. Okay.
QUESTION
BOB KAHN
Okay, let's do it again.
QUESTION
BOB KAHN
All right.
QUESTION
BOB KAHN
Oh, I thought you wanted it on the record, on
the tape.
QUESTION
BOB KAHN
Oh, so ...
-----------------------------------------------------------
BBN JOB 1141540 3-3
QUESTION
BOB KAHN
Oh, so, we don't need to go back over that
again.
QUESTION
BOB KAHN
All right, so, then, let's forget it.
QUESTION
BOB KAHN
I can't tell.
QUESTION
BOB KAHN
Uh, the story of the Open Net(?)?
QUESTION
BOB KAHN
Well, uh, I joined BB&N in 1966. And at the
time I'd been a professor at MIT in the Electrical
Engineering Department. And I'd taken a leave
of absence with the idea that I would get
involved in some practical kinds of activities.
-----------------------------------------------------------
BBN JOB 1141540 3-4
And the one I happened to choose to work on
was computer networking, probably with the
encouragement of several of the people at BB&N.
I'd started to develop some notions of what
computer networking might be like, and sort of
to document it. And, uh, by the time we had
actually gotten the Arpinet(?) Award, I had
decided to actually get involved in the
implementation of that as well. I was principally
involved in the system design aspects before
that, but I eventually decided that, uh, being
actually part of the project itself was probably
the best way to get this practical experience.
QUESTION
BOB KAHN
Uh, yes, in a very material way. Cybelle(?)
Weinstein(?) and I actually sat down and wrote
the technical part of the proposal that went in.
The proposal was packaged as a much larger
proposal, and it involved many other people at
-----------------------------------------------------------
BBN JOB II4 1540 3-5
BB&N. It was done as part of a group that, uh,
Frank Hart(?) led. But Cybelle and I actually put
together the technical part of the proposal. I
actually have the handwritten version of it that
he and I wrote during an all-nighter session at his
house.
QUESTION
BOB KAHN
Uh, I thought the proposal that we put together
was extremely good. And I thought that it had a
really good chance of winning. Uh, but at the
time that I wrote it, I don't actually think that,
uh, I had understood that I... I really wanted to
be part of the implementation of that activity. I
was so thinking seriously about staying with
research and in fact going back to MIT to
become a professor there again. But, uh, in the
final analysis, I thought that it really needed
somebody to stay with the system design, and
so I did.
-----------------------------------------------------------
BBN JOB #41540 3-6
QUESTION
BOB KAHN
Yeah, there was a tremendous amount of
excitement in the group when the award was
won. I was sort of 180 degrees out of phase
with it. I ... I was really excited when the
proposal was written, because we got our ideas
down on paper. And when the contract actually
came in, I was kind of in a low, because, uh, I
knew at the 'point the work actually had to be
done.
QUESTION
BOB KAHN
You mean after the contract was awarded?
QUESTION
BOB KAHN
No comment on that.
QUESTION
BOB KAHN
The Washington Hilton?
-----------------------------------------------------------
BBN JOB #4 1540 3- 7
QUESTION
BOB KAHN
Uh, in many ways, the unsung hero of that 1972
demonstration is AI Buzza(?) from MIT. AI never
really got the visible recogni.tion, but he and I
worked very, very closely in putting the
demonstration together. It involved bringing
together, oh, some 40 terminal manufacturers,
establishing an Arpinet(?) note in the ... in the
Hilton. And in fact, the timetable on which We
actually had to do the installation was something
like six hours to get the phone lines in, to get the
machines up before the thing actually opened up.
And AI just did a tour de force in managing the
whole crew to do that. Uh, it took us about, I
would say ... I would say, close to a year to
actually plan and carry out that demonstration.
And the significance of the demonstration was
that it was the first really forceful push to make
the Arpinet useful to the end users. Uh, in
-----------------------------------------------------------
BBN JOB #4 1540 3-8
1969, we installed the first net, the first net
node(?) on the Arpinet at UCLA. And by
December of that year we had the first four
nodes in place. And Dave Walden and I went
out to demonstrate the workings of the network,
or not, as the case may be, and shortly
thereafter. But by the mid-197½', o. ne....timeframe,
there was still not very much use of the net. It
could move packets around, but there weren't
machines connected doing useful things. And
that project was set up to really force the utility
of the Arpinet to occur to the end users. And by
that 1972 October demonstration at the
Washington Hilton, we had to shoot (Inaudible).
QUESTION
BOB KAHN
No. I think, uh, it's fair to say that we were
relatively confident that this could be done, that
the community would follow suit. I menton AI
particularly because he is a ... really a figure that
-----------------------------------------------------------
BBN JOB II4 1540 3-9
needs the recognition for his efforts there. But it
was really a community-wide effort. We had at
least, uh, 30 or 40 significant major people from
the field, from the academic and the research
community that contributed to making that a
success. People whose names are household
words today were the workers in the field. It
was just a marvelous team effort all around.
QUESTION
BOB KAHN
I think most of the attendees that had been part
of the Arpinet experience at that time obviously
knew what this was about. But it was exciting
because it was the first time the whole
community had shown up in one place at one
time. If somebody had dropped a bomb on the
Washington Hilton, it would've destroyed almost
all of the networking community in the United
States at that point. On the other hand, the
people who attended that knew very little about
-----------------------------------------------------------
BBN JOB 1141540 3-10
networking Arpinet, Ithink, in many cases, were
astonished about, not only what it could do, but
the fact that it worked at all. I really put packet-
switching on the map, in my opinion.
QUESTION
BOB KAHN
Well, when I first started working on the whole
concept of packet-switching, which was really in
the 1966 timeframe, my view was that this was
an interesting way to do computer
communications. I think by the time the Arpinet
project was underway, it was pretty clear to me
that this was actually a practical, viable
alternative for the marketplace. But whether it
would accept it or not, I did not know. I think
after the demonstration at the Washington
Hilton, it was pretty clear to me that this was
gonna fly commercially as well. Of course,
Steve Levy(?) and I were very active in putting
together the early plans for Telenet, the first
-----------------------------------------------------------
BBN JOB #41540 3-1 I
commercial packet net that, uh, made it in the
marketplace. And, uh, I think that was one more
piece of evidence that this was likely to fly,
when, in fact, Telenet made it.
QUESTION
BOB KAHN
Well, the Arpinet was really fairly tightly
controlled as a network activity. It was run,
policy-wise, out of ARPA(?). It was really run
technically out of ARPA from the early days as
well, with BB&N playing the key contractor role
for the subnet. Uh, but I think the thing you're
referring to really came about because of another
development, which was really brought upon by
the existence of other networking technologies,
not just the Arpinet, that led to the creation of
what is now known as the Immernet(?). And
that's what's had the broad reach around the
world. !t was not the Arpinet ...
QUESTION
-----------------------------------------------------------
BBN JOB II41540 3-12
BOB KAHN
I mean, it's gonna be the notion of someone
asking the questions?
QUESTION
BOB KAHN
Okay, right.
QUESTION
BOB KAHN
Okay.
QUESTION
BOB KAHN
Well, the Arpinet itself was a network that was
fairly tightly managed by ARPA, certainly during
its early years and subsequent to that. In the
early 1970s, however, I was involved in the, uh,
development of several other networking
technologies, based upon the same ideas that
went into the Arpinet. One of these was a
ground based packet radio system that had all of
its nodes completely mobile, using
-----------------------------------------------------------
BBN JOB 1141540 3-13
microcomputers and spread spectrum. Uh, the
second and ... second network that we were
involved with was a network that eventually was
implemented on the Intel Sec(?) IV satellite.
Called Satnap(?). And those three networks
seemed to me, uh, were going to need to be
interconnected if we needed to get ... let's try
this one again, I'm not happy ...
QUESTION
BOB KAHN
That's... yeah, let's go back. l actually had a
different train of thought I wanted to ...
QUESTION
BOB KAHN
Yep. Uh, the Arpinet was, uh, a relatively small-
scale network by comparison with, let's say, the
Internet. It was controlled by the,Ab,,vance
Research Projects Agency in a fe/tight way for
most of its existence. In order to get on that
net, you really needed to either be a defense
-----------------------------------------------------------
BBN JOB II41540 3- 14
contractor, or be approved ... the government
paid for all the connections. When you compare
that with the Internet, which had a very broad
reach ultimately, and reached around the world, 1
think the real implications of this are being felt
more from the Internet than they are from the
Arpinet per se. Afar as the origins of the
Internet, it really goes back to the early 1970s,
uh, when, in addition to the Arpinet I was in the
process of designing two other networks, one of
which was a ground-base packet radio network
that embodied many of the notions of the
Arpinet package switching concept that applied
to small, mobile nodes that, uh, used radio
propagation. Uh, the other was a satellite-based
net called Satnet, that used Inte IV. It
turned out that BB&N was involved in ... in both
of those, but in the ground radio network, there
were approximately 17 contractors that were
involved, and it was managed out of ARPA,
-----------------------------------------------------------
BBN JOB II41540 3- 15
under my auspices. Uh, in the case of the
satellite net, that actually started under Larry
Roberts' tenure at ARPA, and I'd gotten involved
in it. Uh, and it turned out to be a network that
involved multiple countries in Europe and the
United State on the Intel Sat IV satellite. I knew
that for computers on one network to talk to
computers on another network, we would have
to understand how to link the networks together
and how to make it possible for computers on
both ends to talk to each other. That meant
developing a methodology to do that, because
these networks were all different, they had
different data rates, they had different packet
formats, they had different error structures, error
controls, different, uh, methodologies. And
that's what led to the develop of the TCPIP
protocol suite as a method for that. This is an
effort that, uh, Vint Surf(?) and I collaborated on.
Vint had been originally involved in the early NCP
-----------------------------------------------------------
BBN JOB 1141540 3-16
efforts, along with, uh, Steve Crocker(?) and ...
and a number of other people. And so Vint was
very knowledgeable about both the original
Protocol(?) development for the Arpinet, and the
requirements of the machines to house it. I had
a pretty good idea of how this ought to work,
but I was looking at it from the communications
point of view and not necessarily from the needs
of the host machines. And Vint and I got
together, pooled our talents and the result was
the TCPIP Protocol Suite, which in essence
formed the framework for the Internet.
QUESTION
BOB KAHN
I don't think there's any single characterization of
what it was like to be a pioneer in developing the
Arpinet. In fact, there were so many people
involved in it that I remember one article where
people said that Arpa didn't even know what it
was doing, because one article said it was trying
-----------------------------------------------------------
BBN JOB 1141540 3- 17
to develop packet switching, and another article
said it was trying to do computer resource
sharing and couldn't make up its own mind. In
fact, the ... the essential element of a really good
ARPA program was it was solving multiple
problems, achieving multiple objectives, often
simultaneously. This was an example where the
community that was involved was in fact very
large. It involved people across the country in
various disciplines, all contributing in their own
little way. And I think if you were to write the
history of the Arpinet, it would be pretty hard to
do a complete job referencing every single person
who made a contribution to it.
QUESTION
BOB KAHN
Well, it depends on ... the essence of the
Arpinet development really depends on what you
think of as the Arpinet. To some people, they
think about it as just the communication pieces
-----------------------------------------------------------
BBN JOB #41540 3-18
of the network. Other pieces (sic) think about,
you know, the use of Jabet(?). Other people
think about the strategy and the structure and all
of that sort of stuff. And Ithink those were all
different communities, all interacting in
interesting ways, but there was no quintessential
Arpinet team when you look at it broadly.
QUESTION
BOB KAHN
There was no question in my mind that when
they installed that in, that it would do what we
expected to do, which was, one, end function,
and send packets out the lines and bring packets
in. You remember, one node network is about as
useful as one telephone system. So the best you
could do was loop some lines and make sure that
what went out a line came back in the same line.
Or that the software didn't crash, or that the
internal activities were functioning properly.
There was not a major effort to try and check
-----------------------------------------------------------
BBN JOB #41540 3- 19
out that installation when the first node was
installed, because there wasn't a whole lot to do,
other than to be sure the software ran. It was
after the fourth node was installed ... we actually
had a four-node subnet on the West Coast, that
Dave Walden and I went out to the West Coast
to actually do the field testing and debugging of
the system.
QUESTION
BOB KAHN
Nope.
QUESTION
BOB KAHN
What's the if?
QUESTION
BOB KAHN
Well, of course, the Arpinet was
decommissioned by ARPA in 1990. And the
Internet is gradually becoming more and more
embraced by commercial industry. I think the
-----------------------------------------------------------
BBN JOB #41540 3-20
basic notion of the Arpinet was fundamentally
sound, in terms of the packet-switching approach
and the ability to break things into small trunks
and send them, I think the fundamental notion of
the Internet was sound, in terms of linking
independent systems without requiring central
and overriding the administration of everything.
--985, I left ARPA specifically to work on
fostering the development of the National
Information Infrastructure in the United States.
At the time, this was not a well-known term, and
Ithink many people probably viewed it as
equivalent to harvesting the clouds. But over the
course of the last eight years, there's been a
growing awareness of the importance of this
whole activity, and the importance not only to
this country, but perhaps to the ... to the whole
world. The Internet plays a very important role
in that it currently is as close to an infrastructural
base for the global infrastructure as I think exists
-----------------------------------------------------------
BBN JOB #41540 3-21
in the world today. And the credit for that goes
to a lot of the people that m_ade that possible
d. Butl---
around the worl think l--the--f-u-t-u[e
really going to see the whole notion of packe-
switching embodied in these national
infrastructures, and the global infrastructure.
And what we'll see is better, more rapid access
to the information that we need, more economic
ways to communicate with people, better ways
to integrate data and other kinds of information
in ys_that can be_nefit all of society.
QUESTION
BOB KAHN
Well, I'm an optimist when it comes to
technology and society. Ithink that almost any
technology can be put to positive use. On the
converse, almost any technology can be misused
and even technologies that have inherently
positive uses can be used by small fractions of
the population. So Ithink we have here an
-----------------------------------------------------------
BBN JOB #41540 3-22
opportunity, and Ithink it's one that we're going
to capitalize on and it's one that, uh, you know,
I'm basically very positive about.
QUESTION
BOB KAHN
Well, you know, I think the most memorable
aspects of that whole experience for me were
the facility when it actually got into existence,
partly in the activity that Dave Walden and I had
when we went out to try to bug the first instants
of it. The feel of satisfaction ... I saw ... actually
the first public demonstration of it at the
Washington Hilton. And Ithink if I had to pick
any singular event during that period, it would
probably be the efforts that Len Klein, Arquel(?),
Frank and I had in actually writing the first paper
on lessons learned in the Arpinet experience.
That was, uh ... it was a wonderful interchange
between three people who were then and are still
very good friends. But actually brought
-----------------------------------------------------------
BBN JOB #41540 3-23
completely different points of view to the same
common experience we had all been through.
Len had been approaching this whole activity
from the point of view of the theorist. He had an
analytical notion. I mean, his view was that, you
know, we can build networks up to about 40 to
60 nodes. And Howard and I both shared that
view. But Len's argument was that the
denominator of some equation went to zero.
And Howard would say, from a simulation point
of view, the simulations don't work beyond that
point. And I would say, there's something in the
implementation, like the routing cables, that we
got too big on the lines, or something beyond
that ... that size. So the interplay of our three
different approaches was something that was
really fun to work through and to actually sit
down and write.
QUESTION
BOB KAHN
-----------------------------------------------------------
BBN JOB #4 1540 3-24
Well, Howard ... Howard and I actually got
together once, I recall, up in Boston. And it's
possible that Len and I actually chatted on one or
more occasions at his office at UCLA, although I
have no specific recollection of that. Uh, I do
have a very specific recollection that we did not
all get together at any one point at any one time,
so there surely was no pizza that we all shared
simultaneously. But maybe Len and Howard will
contradict that.
QUESTION
BOB KAHN
Remember, it was 20 years ago.
QUESTION
BOB KAHN
(END OF TAPE 3)
-----------------------------------------------------------
