Convergence:
Past, Present, and Future
Paul Baran Addresses CableLabs®
Winter Conference
When CableLabs invited me to speak at its winter conference, I wasn’t sure what the audience might be interested in hearing. So, I asked a few people, "What should I talk about?" One asked, "Where in hell did you get the nutty idea of breaking a data stream into little pieces?" (now called packets). Someone else asked, "Whatever became of the old 1985 packet cable system?" And, CableLabs suggested talking about "convergence: past, present, and future." So, I am going to try to cover all three topics, since there is a tenuous string relating them together.
The 1961 Epiphany
The first time the "light bulb" went on, or my epiphany on the matter of convergence occurred, was 38 years ago. That’s a longer time than the age of many in this room. I was at the RAND Corporation in Santa Monica. RAND is an interesting place: a not-for-profit organization set up to preserve the operations research capability that the Air Force developed during World War II.
Parenthetically, RAND also has a place in CableLabs’ history. About a dozen years ago, Dick Leghorn came up with the idea of CableLabs. He personally funded a study at RAND on the detailed creation of what is now CableLabs, and used the RAND report to help recruit Dick Green. And as we all know, CableLabs has gone on to become a great success, with major achievements advancing the cable industry.
Paul Baran, board chairman of Com21, Inc., addresses the CableLabs conference
audience.
Cold War Fears
But, back to the early 1960s’ story. For those around at the time, you might recall that this was a period of Cold War fears. Kids were taught to duck under their desks for safety when they saw a bright flash. Both the US and the USSR were fearful that the other side might be tempted to pull off a Pearl Harbor-type attack, but this time with thermonuclear weapons. This fear was amplified by the realization that whichever side went first, it could essentially wipe out the retaliatory capability of the other…"two scorpions in a bottle."
To cool down this hair-trigger state of affairs would require hardening and disbursing the strategic forces so that enough of the retaliatory capability could survive. But, there was a serious problem communicating to the disbursed forces at that time. The two main communications channels of the military were high-frequency (short-wave) radio and the telephone plant. A single high-altitude nuclear burst takes out the sky wave of high-frequency communications for hours, and the EM pulse can be quite bothersome.
The RAND computer simulation of attacks that were directed solely against the US strategic forces showed that these attacks also take out the nation’s long-distance telephone system, by collateral damage alone. So the Achilles’ Heel was communications.
The bombing simulations indicated that only a small portion of the nation’s telephone systems plant would be physically damaged. But, the highly centralized nature of the long-distance switching meant that even relatively slight collateral damage essentially would block all long distance communications.
Building a Survivable Network
I found that if a network could be built in the form of a fishnet, then most of the surviving elements would theoretically remain in communication with the other surviving elements. And, to achieve this level of robustness took only about three times the minimum number of elements needed to build a conventional network.
The next issue was how to do it. All long-distance links used analog transmission. The telephone switching system had a design limit of no more than five tandem spans, beyond which the signals are considered to be unusable.
This meant that we would have to use digital modulation, as you can continuously amplify and clip digital signals to prevent cumulative noise and distortion buildup. The next challenge was how to route traffic from any point to any other point through a torn fishnet. Chopping each digital stream into 1024-bit message blocks, or what are now called "packets," and adding housekeeping information could allow these packets to find their way through a network under attack. The proposed concept of chopping up data streams into small pieces and then sending them by various routes, was regarded as the "dumbest idea" they ever heard, by my more polite friends at the telephone company. I’ll not repeat some of the other remarks.
Chopping the Digital Stream
Yes, of course it did seem like a nutty scheme at that time – combining voice and data on the same network. But if successful, the payoffs would be significant. All sorts of wonderful new system properties became possible. It allowed easy implementation of error detection and repeat transmission, so that if no acknowledgment were received, the packet would be sent again over a different route. This meant that essentially error-free transmission could become possible. Then, there was the trade-off in reliability. As it didn’t make any difference whether a link failed because a bomb hit it or because one was using cheap, low-reliability links, very inexpensive, very reliable high-data rate networks could be built. The overall system became far more reliable than any of its pieces. And, there were also lots of other goodies. For example, one’s address was no longer locked to a physical location in the network.
This was my first encounter with "convergence," the now obvious concept that all signals should be converted to bits: teletypewriter signals, data, and even voice could be transmitted on the same network.
The Poetry of Bits
The underlying realization is that a bit is a bit is a bit. So, instead of long serial connections, it is more useful to think of future communications systems as fast-bit transportation systems. Once doing so, it becomes possible to consider implementation of networks with decimal orders of magnitude lower costs and of far greater reliability, than analog networks.
This work was described in 1964 in an IEEE paper with pointers to a series of a dozen unclassified RAND memoranda. For the curious, these are now available on the RAND website. (Note: The URL is www.rand.org/publication/RM/baran.list.html.)
ARPANET and Me
However, I didn’t want to write a novel. It was required only to address all the many objections raised at the time. On occasion, I get accused of creating the notion of the ARPANET. Not guilty! The ARPANET was an experimental system for sharing remote computer facilities. Because packet switching was chosen as the communications network for the ARPANET, it became intertwined with the ARPANET. The resource-sharing communications function could have been done with a highly centralized system as well. My work was done several years prior to the creation of the ARPANET, but not all the concepts were incorporated into the ARPANET. Most of the new ideas, for which I am not at all responsible, were added as well.
The Internet, an outgrowth of the ARPANET, also uses packet switching, and represents the ideas of thousands of people around the world working cooperatively and generally without particular remuneration. It stands as a tribute to how many people can work together to create something that simply could not be done by any single individual, group, company, or perhaps even country. It is a successful progression of early governmental support, first primarily by the US Defense Department and later by the US National Science Foundation. After the National Science Foundation very slowly and reluctantly allowed its connection into the commercial world, industry and individuals from all walks of life took over and made it what it is today.
Second Convergence/the First Packet Cable
The second time the light bulb turned on relative to convergence occurred about two decades after my work on packet switching. It was at a company called Packet Cable. Name sound familiar? I received a nice letter from Dorothy Raymond, CableLabs’ general counsel, about a year or two ago noting that since the name wasn’t being used at the time, would I have any problem with CableLabs using that name? I told her that, of course, I was delighted and honored that the name packet cable would be recycled by CableLabs.
Packet Cable, Inc., later renamed Packet Technologies, Inc., was set up in the early 1980s during the cable franchise wars. Cable started in the rural areas as an extension to a cable TV antenna. When the economics of cable allowed extending cable to the cities, there was a bidding war for the franchises. All sorts of games were played at the time, including rent-a-citizen, giving out cheap stock to bribe local political figures, etc. Then, there were the escalating promises to win the franchises – more channels, more local origination, and two-way data transmission for videotex, alarm services, computer connections, and even telephony.
New Digital Services in 1985
The concept behind Packet Cable, Inc., was that, to keep all those franchise promises made to the cities, there would be a need for new equipment. By 1985, we had two systems operating. One was at the United Cable’s system in Cupertino, CA, and the other in the Hearst Cable system in Los Gatos, CA. A number of people in this conference room visited us and saw the system working at the time. It was a multimegabit, asymmetrical system. We supported impulse pay TV channels and generated shared videotex in an off-premises unit. The data path originally ran at 2 Mb/s in a 6-MHz channel. We later shifted to using two 1.54 Mb/s in the same bandwidth to be able to support locally generated videotex and packetized voice transmission. Telephone voice quality was very important, so we used short 192-bit packets corresponding to the T1 frame timing. We suppressed silence periods in voice transmission and modified 64 Kb/s voice to 32 Kb/s ADPCM. As a result, we were able to pack 96 channels of excellent quality voice on a T1 circuit normally limited to 24 channels.
Michigan Bell liked the economics of the system, but didn’t need the cable transmission portion of the system. They were interested in a product to be used solely with twisted-pair copper, to improve the efficiency of their T1 circuits and to switch time slots saving manual labor in nailing up fixed circuits. This product was called the PacketDax as it provided the telco’s DACS functions. The product worked and it was clear that the convergence of quality telephony and data on the TV cable was feasible.
Investor Relations
Packet Cable’s major investor was a large oil company. When the price of oil dropped to $10 per barrel, they stopped their sponsorship. "No problem," we thought, as there were other companies who wanted to come in. But, greed got the best of that oil company, that was able to convert its equity to secured debt. Then it refused to let any new money come into the company, except at an unreasonable price. They seemed to want to bankrupt the company, and the losers in such a move would be the unsecured creditors. Packet Cable’s management felt this was unfair and worked without salaries to block this move. The company was allowed to perform a leveraged buyout of the telephone technology to the engineers working on that project. As the oil company saw the future to be in the cable products, and not in the telephone product, it blocked such a move for the cable products.
Over time, Packet Cable eventually was able to buy out the oil company’s position and pay off the creditors. After the dust settled, a little paper stock left over in the leveraged-buyout company, now called Stratacom, was distributed to the Packet Cable investors and employees. The rest of the story is that a few years ago Stratacom was sold to a large company for $4.4 billion in stock. Today, the value of this stock is over $16 billion. So, in our second brush with convergence, Packet Cable did well financially by taking a side road, but it never met its long-term objective of providing digital telephony, data, and video on cable TV systems.
Convergence III
Today, a third wave of convergence is underway with lots of players and lots of action. It is moving toward the interconnection of all forms of communications into a network of systems, or perhaps a system of networks. A major driving force is CableLabs’ work on DOCSIS 1.1 and on the PacketCable™ project, self-defined as "a fast-track effort aimed at developing interoperable interface specifications for delivering advanced, real-time multimedia services over two-way cable."
History should never be written while it is happening. So we are going to have to wait until the action sorts itself out. But, I think, given CableLabs’ initiative, we are really going to get there this time. And, again, I am delighted and honored that you have chosen the name, PacketCable.
His first job after school was as a technician working on the first commercial computer, the original Univac in 1949. Later he designed and installed the first telemetering equipment at Cape Canaveral in the early 1950s. In 1955, he joined Hughes Aircraft where he worked on radar data processing systems and in vulnerability studies and analysis. He joined the RAND Corporation in 1959 where his early work defined what is now called packet switching. While at RAND, he also conceived the doorway gun detector and recommended its use for airport screening. And, he was the first computer scientist to testify to Congress about the upcoming problem of computers and personal privacy. He left RAND in 1968 to cofound the Institute for the Future (which recently celebrated its 30th year) to improve longer-range planning technology.
Starting in 1972, he founded or cofounded a series of seven private companies in the communications field, of which five later became public companies. His most recent endeavor is Com21, Inc., in Milpitas, California, where he serves as its Chairman of the Board.
He is a Life Fellow of the Institute of Electronic and Electrical Engineers, a Fellow of the American Association for the Advancement of Science, a Member of the U.S. National Academy of Engineering, and a Member of SCTE.
His awards include: The IEEE Alexander Graham Bell Medal, The Marconi International Fellowship Award, The IEEE Edwin H. Armstrong Communications Society Award, the Drexel University Centennial 100 Medal, The Electronic Frontier Foundation Pioneer Award, and the Japanese Computers and Communications Foundation Award (together with Vinton Cerf and Tim Behner-Lee).