ARPANET, as I Recall
by David Retz, Comware International, Inc.
Last updated: 2010-04-05

I was a graduate student at UCSB, studying under Dr. Glen Culler - the Pied Piper of computer creativity. The Culler-Fried online system was chosen by ARPA to be one of the first four sites on the nascent ARPANET. One of our system programmers, Helen Smith, wondered what this "ARPANET" thing was all about - supposedly creating a national computing resource. (In retrospect, you have to admit, the original concept was "cloud computing"!). She was writing simulation programs on our 360/75 mainframe using GPSS. (Helen was one of those programmers of the Ramo-Woolridge system that could enter a program to manipulate a matrix in octal (machine language) and have it work the first time.)

Glen Culler left the University to form a new company, taking Helen and a few engineers with him. I was assigned the new ARPANET project, working with Roland F. Bryan to develop software "drivers" (we didn't know they were called that back then) for the IBM MVS operating system - talking to a new specially-designed hardware interface for the IBM 360. The box talked "Bus+Tag" specifications to the 360, and talked "BBN Report 1822" to the ARPANET via a special minicomputer called an IMP, or "Interface Message Processor".

At that time, I was handed a 3/4" set of notes from previous meetings. Here is an interesting one by Larry Roberts, dated November 7, 1967, kicking off the ARPANET concept to the Principal Investigators at some of the ARPA research sites. This is interesting reading, since it puts into perspective the emphasis on cost of digital communications at the time, especially on a nationwide basis.

[Note: the original IMP, a Honeywell 516 16-bit minicomputer) had hooks on the top to allow it to be lifted by a crane, looked like it could withstand a nuclear attack quite well, and had a rudimentary ASR-33 teletype as the operator console. It was the equivalent of a modern small Cisco router, only the cost was supposedly about $50k to the government, as opposed to less than $1k today. Note also: Ethernet did not exist at this point, so the I/O interface to the IMP had to be invented on-the-spot. This was a bit-wise handshake interface that could talk to any computer that existed at the time, invented by Jerry Burchfiel at BBN.]

One of the few photos, courtesy of BBN by Frank Heart (with skinny tie and crew cut),
and the IMP developers.
To give a perspective on events at the time: there were regular "Network Working Group" meetings held at the primary locations of UCLA, UCSB, SRI, and UTAH (the latter always in ski season!). I returned from one trip at UCLA to Santa Barbara, only to get my hand stamped on entry to my home town. The sheriff told me I had 15 minutes to make it to my apartment or I would be arrested. The town was under lockdown after a William Kunstler speech in Isla Vista which gave way to riots, overturning of cop cars, and burning of the Bank of America. The ARPANET development was concomitant with these Vietnam war protests, an ironic punctuation to the fact that the original motivation for ARPANET was to provide a packet-switched communication network that could survive a nuclear attack - the cold war post-Sputnik result from think tanks such as RAND corporation - with innovative concepts "packet switching" from people like Paul Baran.

There were very few pictures taken of the development teams (at the various sites) of the ARPANET - primarily because no one saw the tremendous impact it would have on the future of communications. This famous picture (one of several taken by Frank Heart, Group Director of the IMP development team at BBN. the ties shown in the picture are reminiscent of those seen in early NASA shots ... somewhat Rod Serlingesque as if from the Twilight Zone.

I graduated from UCSB and worked for a small research company in Santa Barbara called Speech Communications Research Laboratory (SCRL), which focused on basic research on speech understanding and communications. At the same time, ARPA was working on computer understanding of speech - and SCRL was chosen to be a contributor to the project because of its extensive phonemic dictionary of sounds.

At the time (around 1970), the computer industry was making its transition to a new techology: the minicomputer, and the predominant player was Digital Equipment Corporation (DEC), which marketed computers under the alter-ego of "Peripheral Data Processors". Our database ran on the DEC PDP-8 - a microscopic computer by any standards, and the upgrade was to the DEC PDP-11 - a general-purpose electronic box that could connect to about anything. We were located about 10 miles away from the nearest ARPANET location, and needed to implement a new "Very Distant Host" interface to the PDP-11. This was another first - connecting via a high-speed serial port to the nearest IMP, allowing many remote smaller computers to connect to the ARPANET. At the same time, we worked as part of the ARPA-sponsored "Speech Understanding Project" and "Speech Compression Project" with other sites such as Carnegie Mellon University (a href= target=_blank>Raj Raddy), Lincoln Labs (Jim Forgie), and USC Information Sciences Institute (Danny Cohen). We used our ARPANET connection to remotely access the larger-scale DEC system, the PDP-10, which somewhat ubiquitously used the time-sharing software developed by another team at BBN: Tenex.

Somewhere in the middle of that timeframe (around 1971), Ray Tomlinson at BBN had put together a simple program that would allow you to put a message in another user's folder - by simply appending to their MESSAGE.TXT file. On login, you could check for updates to the file - and there soon abounded plenty of cute programs that would scan the file for new updates. One of them, called "bananaread" was a macro that ran as part of the TECO text editor of the DEC-10. By 1972, email was the thing to use, and mail reader programs abounded. Soon afterward, forwarding capability was implemented to allow sending mail between host systems, using a convention of emailaddress@hostname. [Note: this preceded the invention of domain names. Believe it or not, all the hosts were maintained by a sweetheart named Jake Feiner at SRI International - who published the ARPANET addresses for everyone every month or so.]

Around 1974, I moved to SRI to work for Doug Engelbart in the Augmentation Science Center (ARC) and several months after that I became part of the Packet Radio group, under the auspices of a href= target=_blank>Don Nielsen and Ron Kunzelman. Bob Kahn (from BBN) had by that time gone to work as a Program Manager for ARPA - soon changing its name to DARPA - and was in charge of this new network that was radio-based, rather than landlines. This fit in well with the original plan of an indestructible communications network using mobile devices using wireless. The primary participants were: SRI International, as testbed coordinator, BBN, as software developer, Rockwell-Collins Radio, as hardware provider, UCLA, for Network Measurement, and Network Associates for network simulation.

This posed an interesting problem: how would addressing be performed of various nodes and devices on the network - then the Packet Radio Network (PRNet) was parallel to the ARPANET? (Coincidently, not too far away from SRI, was a commercial development by Xerox PARC that was proving successful for connecting nodes within a local area: Ethernet.). DARPA (under Bob Kahn) had a relatively small contract with Professor Vinton G. Cerf at Stanford to explore the problem of interconnection of multiple packet-switched networks. The result was a paper called "A Protocol for Packet Network Intercommunication". It became the basis for the "Cerf-Kahn Protocol", or the "Kahn-Cerf Protocol". Initially, the replacement for the original ARPANET Host-Host protocol was called "Transmission Control Protocol" (TCP), but the revisions to the documents included a more-generic way of addressing things in interconnected networks. The resulting address scheme became known as "Internet Protocol" (IP), which we now embrace as "TCP/IP".

The Packet Radio network was the first guinea pig to use TCP/IP, with a DEC PDP-10 version written by Jerry Burchfiel of BBN, and a DEC PDP-11 version written by Jim Mathis. During the Nixon Administration, there was an increased motivation to show relevance of this research to National Security. The DARPA sponsored work received a demonstration of Packet Radio technology commuicating from a mobile van developed by SRI, driving down the Bayshore freeway in link with repeater units on the mountain tops, through a gateway located at SRI to the ARPANET, then through an international satellite gateway to a host computer in England. Interactive protocol used at the time was simply telnet. (Yes, this was highly insecure, but since it was the only internet in existence, no one gave this security issue any thought.)

But the final stop for the mobile van was chosen by Ron Kunzelman of SRI to be a "hostile environment" - in keeping with relevance to military application. This was the parking lot of Rossatti's biker bar in Palo Alto, still well in reach of the repeater units at Mt. Umunum and Mission Ridge - and with good supply of local bikers who gave the appearance of hostility after the requisite number of beers.


Why the ARPANET was Built
An Interview with Keith Uncapher


Special thanks to all of those who helped make ARPANET possible (partial list!):

BBN: Frank Heart, Jerry Burchfiel, Ray Tomlinson, Ginny Strazisar, Dave Waldon, Wil Crowther
Lincoln Labs: J.C.R. Licklider, Jim Forgie
ARPA: Bob Taylor, Larry Roberts, Bob Kahn
UCLA: Steve Crocker, Jon Postel, Vint Cerf, Len Kleinrock, Charlie Klein
USC-ISI: Danny Cohen, Paul Raveling, Keith Uncapher, Tom Ellis
UCSB: Dave Harris, Glen Culler, Jim White, John Pickens, John McAfee, Art Bergreen, Ron Stoughton, Roland Bryan, George Gregg
CMU: Raj Reddy
SRI: Doug Engelbart, Dan Lynch, Don Nielson, Don Cone, Ron Kunzelman, Jim Mathis
SCRL: Jon Miller, Jim McClurg, Bob Dolan, Bruce Schaefer, John Markel

  © Copyright 1989-2014, Comware International, Inc. All Rights Reserved.