Oral-History:Robert Metcalfe

From ETHW

About Robert Metcalfe

Robert M. Metcalfe

This interview is part of the Marconi Foundation 30th Anniversary Commemoration.

Robert Metcalfe has made important contributions to the development of the Ethernet. He studied at the Massachusetts Institute of Technology, MIT, where he received a degree in industrial management and a degree in electrical engineering. He earned a Ph.D. from Harvard University. He founded 3Com Corporation. He is recipient of the Marconi International Fellowship and the IEEE Medal of Honor.

Robert Metcalfe starts this interview with his family background and early interests. He talks about his training at MIT and Harvard and his jobs at XEROX and MIT. He discusses a major problem with his dissertation and how he solved it. He discusses at length the beginnings and development of Ethernet. He talks about the struggle to adopt Ethernet as a joint standard. He mentions several attacks to the Ethernet and the fierce competition among companies. He discusses the benefits of Ethernet in various areas and fields. He concludes the interview with his intention to start a supercomputing company and with comments about the Marconi International Fellowship ceremony.

See Vinton Cerf Oral History and Robert Kahn Oral History for further discussion of Metcalfe's influence, the development of ARPANET, and the invention of Ethernet.

About the Interview

ROBERT METCALFE: An Interview Conducted by Robert Colburn, IEEE History Center, 19 February 2004

Interview # 433 for the IEEE History Center, The Institute of Electrical and Electronics Engineers, Inc.

Copyright Statement

This manuscript is being made available for research purposes only. All literary rights in the manuscript, including the right to publish, are reserved to the IEEE History Center. No part of the manuscript may be quoted for publication without the written permission of the Director of IEEE History Center.

Request for permission to quote for publication should be addressed to the IEEE History Center Oral History Program, IEEE History Center, 445 Hoes Lane, Piscataway, NJ 08854 USA or ieee-history@ieee.org. It should include identification of the specific passages to be quoted, anticipated use of the passages, and identification of the user.

It is recommended that this oral history be cited as follows:

Robert Metcalfe, an oral history conducted in 2004 by Robert Colburn, IEEE History Center, Piscataway, NJ, USA.

Interview

Interview: Robert Metcalfe

Interviewer: Robert Colburn

Date: 19 February 2004

Place: Waltham, Massachusetts, USA

Family and educational background

Colburn:

I would like to ask about your early education, your early interests and what led you down the educational and intellectual paths you went down. I will also ask you to describe, as much as you can, the creative process behind the development of Ethernet and other things on which you have worked that you think might be of interest to the history of telecommunications.

Metcalfe:

My four grandparents arrived separately in Manhattan around 1900. Then they met each other, married and moved to Brooklyn – both sets. They had children and their two children met – my parents. They had both been raised in Brooklyn. They married and had me in 1946. We moved out of Brooklyn to Levittown around 1950 and moved a year later to Bay Shore, New York on Long Island where I went to school, kindergarten through 12th grade. I was recently inducted into the Bay Shore High School Hall of Fame; I got to invite my parents to go, and it was very nice. My parents did not go to college. My father was a technician and my mother was a housewife until she became a secretary at the Bay Shore High School, after I graduated. My father worked on gyroscopic platforms, including airplanes, submarines and missiles. My father had a little workshop in our basement, and one year he started a company called BAM Electronics – Bailey, Abrahamson and Metcalfe (BAM) – to repair televisions, which were pretty new in those days. The company did not last very long because the partners did not get along. Repairing a television in those days was basically just changing the tubes. His basement was full of all sorts of doodads among televisions in the midst of being repaired. He and I built a model railroad set on a 4x8 piece of plywood with switches and things. Somewhere around the 4th grade, I had to write a book report. I am a little vague on this. It is one of those things in my head that I don't know if it represents the actual literal truth, but I believe it is the truth.

Colburn:

True or not, it is now going to be history.

Metcalfe:

Maybe. I had to write a book report for school, and I had not read a book. I went down into the basement where my dad had his shop. He had some books down there. One of the books on the wall, incidentally, was a manual on gyroscopic guidance for missiles. It taught the fundamentals of missiles and guidance. Another book was a textbook in electrical engineering written by a couple of professors from the Massachusetts Institute of Technology. It was late at night, the book report was due the next day, and I went through it. I can only imagine that it must have been incomprehensible to me, but I went through it anyway. I wrote a book report like, "I read this book, Introduction to Electrical Engineering by these professors at MIT, and all in all it was an average book. It had its high points and it had its low points, but in the end I felt on balance it was probably average." I tacked on a gratuitous closing sentence, probably calculated to kiss up to the teacher, Mrs. Sharp, to make her more receptive and maybe give me a better grade. The closing sentence said, "And someday I plan to go to the Massachusetts Institute of Technology and get a degree in electrical engineering." From that day on, that was my goal. That was 4th grade as I recall, so that meant I had to go through 5th, 6th, 7th, 8th and all the way to 12th. In 8th grade, I took apart our model railroad set and built from toggle switches, relays, and neon lights what my teacher called a “computer” – it added any number from one to three to any number from one to three to give the correct sum between two and six.

Undergraduate studies at MIT and Ph.D. studies at Harvard

Metcalfe:

In 11th grade I applied and was accepted to MIT. I stayed at my high school another year because everyone told me that would be wise. Maybe they thought I was too young. I entered the class of 1968 in the fall of 1964, fulfilling that pledge that had been made gratuitously in 4th grade. I was sort of destined to be an engineer because of my father. He was a technician, but after thirty years in the union at this one company, Arma Corporation – an aerospace company that is now long gone – his title had become engineer. He did not get it the easy way like I did by going to college. He worked for thirty years as a technician. He was a Tech I, then Tech II, Tech III and Tech IV. Finally he got to be an electrical engineer, without leaving the union. He never wanted to leave his union. That is an alien thought to me, since I despise unions generally, associating them with organized crime, which, by the way, my mother’s mother, Anna Sorensen Myhre Berg, fought daily along New York’s docks during her long career at the Waterfront Commission. I guess I was destined to be an engineer, just by being surrounded with that and proud of it and liking it. I graduated MIT with two degrees, one in industrial management and one in electrical engineering. I qualified to get the management degree first, but my dad said, "I did not send you to MIT to get a management degree." Therefore I stayed a fifth year and got two B.S. degrees including my EE degree. I hasten to add that was in 1968-1969, so the war in Vietnam was raging. It was convenient to be there a fifth year. At the end of the fifth year I reported for induction services and was disqualified because I have asthma. Had I been dodging the draft, it would have been moot because of my asthma. I had not realized that asthma would disqualify me. I should add the detail that in 1968, I married MIT's draft counselor. We have since divorced and she has since died. It was a strange time. In that time I applied to graduate school. I had done a computer modeling and applied mathematics. I was a computer guy, but they did not have computer departments then. One could not get a computer science degree in 1968 at MIT, but instead got an electrical engineering degree. At Harvard one could not get a C.S. degree. An applied math degree was the closest thing they had. I applied to graduate school and got accepted at Harvard. Skipping some details there, I eventually ended up in a Ph.D. program. I consider my Ph.D. to be in computer science, though I am not sure that my degree says that. I think I am allowed to say that, but I think it was recorded as a degree in applied math. That was where all the computing was, however.

Shortly after I arrived at Harvard I hated it. I was an MIT guy and Harvard was completely different. I did not like it. It was like oil and water. I had to find an area to do my Ph.D. research, and the big new DARPA research project in those days was called ARPANET. It was not DARPA yet; it was still ARPA. It went from ARPA to DARPA to ARPA, and I think it is called DARPA again now. It had something to do with the war. They slapped the "D" on as a result of the Mansfield Amendment, which was related to the Vietnam War. Anyway, it was ARPA and the ARPANET. I had just finished taking a bunch of courses at MIT in digital design and computer programming, so I made a proposal to Harvard. Harvard had a computer called PDP-10. It was a DEC computer. All the universities had PDP-10s. I said, "I am going to build an interface for that PDP-10 to put it on the ARPANET. That will be my starting project toward a Ph.D. related to the ARPANET." Harvard said, "No. You are a first year grad student. We cannot let you do that. We have to have a company do it." They hired BBN to do it, the inventors of the ARPANET, who turned around and hired a fellow graduate student to build it. Ben Barker was his name. There was some irony in that. While that was happening and my disaffection with Harvard was growing, I went down the street to MIT with the same proposal. I went to J. C. R. (“Lick”) Licklider at MIT’s Project MAC. Lick is, not Al Gore, the Father of the internet.

Project MAC at MIT

Colburn:

Yes.

Metcalfe:


Audio File
MP3 Audio
(433 - metcalfe - clip 1.mp3)


Your being in the history business, you may have heard that theory. You can go back to Vinton Cerf and Robert Kahn or you can go back to Lawrence Roberts or you can go back to Paul Baran or if you want you can go back all the way back to Joseph Licklider. I went to Licklider's group at MIT and met with a man named Al Vezza, who was Lick's right hand man. I said, "I want to build a PDP-10 interface to the ARPANET." They said, "Well, we need one of those. Clearly you know how to do it, because you took 6.01, 6.02, 6.70 and 6711 and 6.721. You already know how to build stuff," and they gave me a job to build it. I became a member of the research staff at Project MAC at MIT. Since I was not a grad student at MIT I was paid more than my professor at Harvard. Had I done this at Harvard, I would have done it for three cents a year or whatever they were paying grad students, but I had stumbled into an arrangement where I was a grad student at Harvard while a member of the research staff at MIT. It was a perfect arrangement. Anyway, I built the interface, which I still have at my home. A picture of that interface appeared in the IEEE Spectrum article. I am on the cover getting the IEEE medal, which is prominently displayed on my wall at home. That was my first high-speed network interface. It was really rip snorting at 300 kilobits per second. I wrote my dissertation about the construction of the ARPANET, and submitted it to Harvard, anticipating that I would receive my Ph.D. in June of 1972. I went out job hunting and got nine job offers. I did not get the one I wanted. I wanted to be an assistant professor at MIT. I would have taken that in a heartbeat. I was offered other jobs at MIT, but not that one. Sadly, I had to accept a job as a member of the research staff in the computer science laboratory at the Xerox Palo Alto Research Center (PARC). I was paid substantially more, had no teaching responsibilities, was flown first class, was surrounded by geniuses and got all the money I needed for equipment. When I showed up in May of 1972 to undertake my thesis defense with my committee, I went in and described it and failed my thesis defense. Have you ever heard of such a thing? It is very rare. I have been told repeatedly that it is the university's fault when that happens. My committee should have told me that I was not ready prior to my defense. However you see I was working at MIT, and the people on my committee were all Harvard people. I had very little contact with them. Had I been more sophisticated in social processes and so on I might have noticed that I had a problem in that I had not carried water for any of those professors.

Colburn:

Quid pro quo?

Metcalfe:

Well, they were not properly socialized into giving me my Ph.D. so they failed me. I had to tell my parents not to come to Harvard Yard two weeks later.

Completion of dissertation on ARPANET and ALOHANET while working at Xerox PARC

Metcalfe:

I called up Xerox PARC. Bob Taylor, who was the head of the lab, answered the phone. I explained to him that even though he had offered me this position assuming that I was going to be a Post-Doc, I was not going to have my doctorate. Bob Taylor said, "Oh, that's too bad. Well, come anyway. You can finish your dissertation while you are here." I moved to Palo Alto in June of 1972. The first thing I did at Palo Alto was build a second ARPANET interface for connecting a PDP-10 clone machine that we were building there. I did for Xerox PARC what I had just finished doing at MIT. I became the "Network Guy" at Xerox PARC. While I was there in that first year, in late 1972, I was building this hardware, and also working to finish my dissertation. I changed thesis advisor, which I needed to do, to Jeff Buzen. He subsequently founded a company called BGS Systems here, which was quite successful. I have not seen him since by the way. I should go look up Jeff Buzen, because I owe him a lot. In our first meeting he said, "All right, let's figure out what we have to do to get you out of here." Until that moment I had been harboring the belief that I was writing the best Ph.D. dissertation that had ever been written, but he had a more practical bent. I needed something theoretical. What I had written up until then was a description of how the ARPANET was built, and how I had built stuff, and how I had written code, and some early experiments with protocols, and so on. The committee's consensus, with which I cannot disagree in retrospect, was that it was really yeoman engineering; it was not an original scientific contribution of the kind required to qualify one for a Ph.D. from Harvard University. I needed to find something theoretical. ARPA, although I was working for Xerox, asked me to be an ARPANET facilitator. Now that the ARPANET had been built – or the internet 1.0 had been built – they wanted two or five facilitators to go around the country and describe it to people. I did that under the auspices of Steve Crocker, who was a program manager at ARPA at the time. I visited Steve in Washington frequently and took to staying at his house rather than a hotel. One night after I had just flown in from the west coast and was sleeping on his couch, on the shelf next to the bed was a blue book. It was AFIPS Conference Proceedings 1970. Perhaps you remember AFIPS. Is there still an AFIPS?

Colburn:

I believe there is.

Metcalfe:

The American Federation of Information Processing Societies, a member of IFIP, the International Federation for Information Processing. Their fall joint 1970 conference proceedings were on the shelf. I was a little tired and needed to read to help me to sleep. There was a paper about a thing called the ALOHANET by Norm Abramson at the University of Hawaii. It was a packet radio network he was building there on the Manoa campus. He had in this paper a queuing theory model of how the ALOHANET would perform under increasing traffic loads. I had studied queuing theory and mathematical modeling and Markov processes and Poisson processes and so on at MIT and a little bit at Harvard. I understood Abramson’s model, but more than that, I disagreed with it. There were two assumptions in the model that I found irritating, and they were assumptions made for mathematical tractability reasons. There were theorems and stuff in queuing theory that if a model was of this kind [pointing to an example] then this [pointing] was the answer. He had concocted this ALOHANET model in a very straightforward way in the way queuing theorists did, assuming Poisson processes everywhere, and infinite populations. He made two assumptions that I viewed to be unrealistic. One was that the ALOHANET had an infinite number of users. Well, the fact is they would be lucky to have ten users. Ten is a lot less than infinite, so I figured that was not a really good assumption. Remember this was 1970. People were typing on what amounted to cardpunches, hitting enter, and sending card images to the mainframe. His other assumption was that people would keep typing when they received no acknowledgement.

Colburn:

Oh.

Metcalfe:

The way this worked was that the image would be sent in and a little light would go off. When the acknowledgement came back saying that the mainframe had received a card image, the little light would go back on, and then one could type again. This is how they actually worked. However, Abramson’s model assumed that people would just keep typing, even if there was no green light. I did the math. I assumed there were N users where N could be one, two, three, ten, or a million, and that it was a blocking system. That meant one would type, stop until an acknowledgement was received, and then type again. Therefore if it took a really long time to get back the acknowledgement, one would not continue typing. To me, that is closer to human behavior. I wrote this up, and this paper won an award. It was published by SIGCOMM. By the way, this is my dissertation, which was eventually accepted. It has been republished. I will give this to you – others can buy it at Amazon.com.

Colburn:

Thank you.

Metcalfe:

I am wondering if it actually mentions the SIGCOMM paper. Not everyone's Ph.D. dissertation is currently available on Amazon.com.

Colburn:

No, relatively few.

Metcalfe:

The chapter in here on the ALOHANET originally appeared as a paper.

Colburn:

"Analysis of Broadcast Packet Communication" on page seventy five?

Metcalfe:

No, no, no. This came too soon. I am getting a little tangled up in my history here. SIGCOMM could be IEEE or ACM. SIGCOMM is ACM by nomenclature, right?

Colburn:

Sounds like—

Metcalfe:

Looking at the ACM Digital Library, I see that it was reprinted in SIGCOMM Review in January of 1975, "Steady State Analysis of a Slotted and Controlled Aloha System with Blocking." I caused a bit of a scandal here. I basically redid Abramson's analysis of the Aloha System, and it appears in that book. My Ph.D. was about the ARPANET and the ALOHANET, and its principal contribution had to do with observing the following: Abramson's model of ALOHANET had no parameter related to how long a station would wait before retrying to send. In other words, if it failed to receive an acknowledgement for a packet that it had sent in, it would try again, and Abramson's insight was it should try again after a random time. He said that if there were two stations that happened to send at the same time and interfered with one another, then they would not try again at the same time, but would randomize. One would go, and then the other would go, and a collision would be avoided, and the packet would finally get through. That was his insight that Ethernet leveraged. However, I noticed that the period of randomization was not mentioned in Abramson’s model. From that, I built a notion of controlling that variable. A lightly loaded network should randomize over a very short interval and try again very soon, but as a network got more heavily loaded it should take back off and take more time. The throughput graph for ALOHANET showed that the busier it got, the more successful it would be until some point where suddenly everyone's retransmissions collided with one another, and the whole thing collapsed. I added a back-off algorithm that said if a collision is detected, it should be taken as evidence that the channel is busier than anticipated, and increase the interval over which the randomizing is done. If there are successive collisions, that interval will continue to be increased, backing off. Where uncontrolled ALOHANET throughput peaked at 17 percent Ethernet could peak at 99 percent plus. There it was. In examining his model and comparing it to my model, he did not have a parameter for that variable, and I did. With that, I observed that -- by backing off that parameter -- one could stabilize the channel. That got me my Ph.D. In that same moment – literally the same month – Xerox PARC set out to build what is arguably the world's first personal computer. I do not want to get into that debate so I will call it an early personal computer.

Development of Ethernet for early personal computers at Xerox PARC

Colburn:

‘Pioneering’ is a good word.

Metcalfe:

An early pioneering personal computer. It was called the Alto. It looked a lot like the machine that is on my desk today – bitmap display, mouse, keyboard, removable media, storage – except that was in 1973. It was thirty-some years ago. I was the network guy, so they said, "Bob, you've got to build us a network for this machine. Build a card that plugs into the computer that hooks it up to the network so that it can communicate, and it needs to communicate with the ARPANET," and they said, "We are going to build a laser printer that does 500 dots per inch, a page per second, so we want for all the personal computers we build to be able to print their documents on this printer." If you do the math on that printer, that is a lot of bits per second. I haven't done the math recently, but 500 times 500 times 8 1/2 times 11 per second, so that's a big number. That is 25 million bits per second. RS-232 was then the standard for connecting terminals together, and it frequently ran below 19.2 kilobits per second. It was not even close to what was needed.

Colburn:

I remember having to write some code to slow down our computers when we networked printers to them because they were too fast for the printers. We had to give the computer a task that would slow it down.

Metcalfe:

Yes. However in my case, the network had to be fast enough to keep the printer busy. Therefore, I set out to design a network to connect these personal computers we were building at the same time. I had just finished building two versions of an ARPANET interface that ran at 300 kilobits per second. I had also just finished submitting my thesis and studying the ALOHANET. Distributed computing, spreading things out, was our theme at PARC – not centralizing. I wrote a memo on May 22, 1973 in which I named Ethernet and laid out its principles. Basically it was to be an ALOHANET like Norm Abramson was building, but it would not be radio. Radio would be too slow. The ALOHANET ran at 4800 or 9600 bits per second – which was just too slowly. Since we were within a building, we did not have to use a big radio modem, so we chose wires. We chose coaxial cables for a very special reason. We wanted to be able to add and subtract nodes from the network without bringing it down. Therefore we did not want to have to cut wires to insert nodes.

David Boggs and I were working on this, and David Liddle down the hall reminded us that there was thing called a cable television tap, which allows one to tap into a coax without cutting it. Therefore we chose coax as our means of communication. In this memo I described the principles of operation – very distributed, no central control, a single piece of “ether.” I did not call that cable “the coax.” There was to be a coax that ran down the middle of the hall, and every computer would run a cable up and tap into it so that all the cables did not have to go back to a room somewhere. They just had to get to the ether. The memo said, "Let's call it ‘the ether’." Ethernet was media-independent from the beginning. In retrospect, that was an important insight because Ethernet has been on every medium since. By calling it “the ether” we opened up the possibility that ethernet could be coax, twisted pair optical fibers, and eventually Wi-Fi, a sort of wireless ethernet. Wi-Fi, or 802.11, used to be called wireless Ethernet, which is where Ethernet started, wireless, as ALOHANET. They sexed up the name a bit and called it Wi-Fi. I wrote that memo in 1973 and recruited Dave Boggs, a grad student at Stanford, and he and I built the first 100-node Ethernet. We chose to run it at 2.94 megabits per second. That speed was fast enough to feed the laser printer and easy to send through the coax. There was no electrical problem. The attenuations were such that it could go 1000 feet, which was our goal. Was it 1000 feet or was it a mile? I have forgotten. In any case, 2.94 megabits per second came about in an odd way. We were designing the first ethernet high-speed network interface card, not unlike the two I had built before. However, the card was smaller than the previous cards. It ran a lot faster, so it needed more buffering. David and I ran out of space. Boggs, to his considerable credit, found a way to add an extra chip, just on the edge, that was not in the plan, so he could put cyclic redundancy checks among these packets to see if they were successfully transmitted. We did not have room for a clock, but the computer had a clock. That clock ticked every 170 nanoseconds, so we used that. We used 2 ticks per bit. That is every 340 nanoseconds and that works out to be 2.94 megabits per second. Over the years, many people have urged me to round that off to 3 megabits, but I insist on calling it 2.94 for the following reason. In 1973 the internet [sic; ARPANET], on a good day, ran at 50 kilobits per second. The circuits were that fast. If 2.94 megabits is rounded up to 3, the rounding error is bigger than 50 kilobits per second. Therefore, as a matter of emphasis, I have always refused to round it up. Ethernet’s round-off error was bigger than then internet. That is how fast the Ethernet was running. Of course, now the ethernet runs 10 meg, 100 meg, a gig, 10 gig, and 40 gig is coming. By comparison 2.94 meg seems kind of slow. That was thirty years ago, and in 2003 184 million Ethernet ports were sold and shipped for $12.5 billion. That is Ethernet ports. That is not counting PCs with Ethernets built in, and not counting the PCs themselves. That is just ports. It caught on, so I got the Marconi Award – and the IEEE Medal of Honor before that, which I take as the highest award an electrical engineer can win. There is no Nobel Prize for electrical engineering. Now if only one would get a million dollars with it. If you really want to make that medal the Nobel Prize of electrical engineering you have got to put a million bucks per year behind it.

Colburn:

I am sure they would like to do that. They work very hard at getting companies to sponsor it.

Metcalfe:

I am sure. I am just needling you. ACM thinks the Turing Prize is the Nobel Prize for computer science. It probably is, but I think it is only $20,000. It is not going to be the Nobel Prize until it is a million dollars, or whatever the Nobel Prize is at that time. Anyway, the Ethernet has caught on. Do you want to go back over any of that?

Influences on the invention of the internet and Ethernet

Colburn:

Actually that did a very good job at answering a lot of the questions I wanted to ask. Here is one of the questions the Marconi people want me to ask each fellow: Marconi was fond of using Newton's image of standing on the shoulders of giants, except that Marconi would say that there were many giants standing on each other's shoulders. Could you perhaps elaborate that sense of where in the process or how in that process of communications your work fitted in, whose shoulders were you standing on and who is standing on your shoulders in turn?

Metcalfe:

In that story I just told, I left out how Ethernet became a standard, and I left out the company I founded to commercialize Ethernet, 3Com Corporation. Those are big parts of the story. Regarding the point about shoulders: JCR. Licklider, Norm Abramson, Larry Roberts, Vint Cerf and Bob Kahn, David Boggs… Marconi. The ALOHANET was a radio network originally. Wi-Fi, which is the latest hot version of Ethernet, is once again a radio network. It has come full circle back to Marconi.

Colburn:

Yes, definitely.

Metcalfe:

I was reminded by the radio this morning that there are two Marconi Awards. Don Imus has won several Marconi Awards. Don Imus is on WFAN Radio in New Jersey. He is the morning shock jock and one of the most famous morning disk jockeys. I listen to him every morning. Don Imus noted that he had won several Marconis. There is a Marconi Award given for broadcast excellence to disk jockeys and so on, so that is different from the Marconi International Fellowship that we are talking about. However it all has to do with radio. It would be hard to make a complete list of the giants because it would be so long. The people who built the ARPANET and the people who built the ALOHANET are certainly people on whose shoulders, if not toes, I have stood. One of the things I have observed late in life, is what happens when something succeeds. Everyone shows up as having invented it. It gets pretty ugly. As you know, Al Gore at one time was reported to have invented the internet, and that was laughable. The Ethernet had many inventors too. I have always been inclusive about that. It is such a big enterprise now. My joke about this is that, by 1982 there were people buying Ethernet from my company, 3Com Corporation, whom I did not know. However, by 1986, there were people inventing Ethernet whom I did not know. The whole twisted pair Ethernet, the whole switched Ethernet. There were people at Xerox and Intel and DEC and HP who are giants in the history of Ethernet, some of whom I am only meeting today. Bill Seifert, a fellow venture capitalist now here in town, is a giant in the history of Ethernet. I barely know him because he worked for Interlan, which was the arch competitor of 3Com. He and I competed, but I never really knew him because we were competing and not friendly. The list is much bigger of people who have contributed at one point or another to the proliferation of the internet. The internet and Ethernet are pretty closely linked. The Ethernet is basically the low-level plumbing for the internet. My first cut estimate is that every internet packet goes through a number of Ethernets before it gets from its source to its destination. I think the average number is probably a dozen, but I do not know that for sure. I am looking for the answer to that. In our machine room, there are probably several Ethernet segments that connect to a front-end Ethernet that goes off to the telephone company and they probably have a router that goes through an Ethernet and so on. I am definitely not the inventor of the internet, despite the fact that they are very closely related. I was at Harvard in 1969, which is the year that Al Gore graduated. I can assure you he was nowhere to be seen.

Colburn:

[laughing] He was not writing packet switching papers?

Metcalfe:

I have a joke about that too. Al Gore may not have invented the internet, but it does run on many Al-Gore-rhythms, so he has left his mark.

Colburn:

If nothing else, from that quote, for which I do not think he will ever be forgiven.

Metcalfe:

Buddies of mine who are his defenders – including Vin Cerf, I might add – went out of their way to defend that outrageous claim. The peak of the defense was, "He never claimed to have invented the internet. Do you know what he actually said? The speech said he “created the internet." I say, "Oh, that's much better." The other way they handle it is that they draw the distinction between the ARPANET and the internet. I do not. I view the ARPANET as internet 1.0, and then 2.0 started in 1984. They pull the “internet” forwards from 1969 by saying the internet protocol was not installed until 1984. However, we were in fact working on those protocols at Stanford the summer of 1973. That was the big year in both the internet’s TCP/IP and Ethernet were invented. They were both invented in the summer of 1973 in Palo Alto, California. Of course TCP/IP was not installed on the internet until 1984. Al Gore was in the Senate in the 1980s. Back to the shoulders of giants, the guy who taught me queuing theory, Alvin Drake, is still at MIT. I took probability and statistics from him, and that led to queuing theory. The Ethernet is essentially a queuing theoretic invention with some hardware and software derived from that. Professor Donald Troxel at MIT taught me how to build digital electronics. He is a giant. He is still there. That led me to build my first high-speed interface, which became a $5 billion company and a much larger industry. Who am I forgetting? It depends on how far you want to go back. More giants are Ken Olsen at DEC, Bob Noyce at Intel and Steve Jobs at Apple. Steve Jobs was the first one to build Ethernet onto the motherboard of his computer. No, no. Sun was before him for that, but he put a version of Ethernet called AppleTalk in the Macintosh in 1984 and built it onto the motherboard. Steve Jobs is a giant who helped get this all going. He is hard to take, but thank God for Steve Jobs. Gordon Bell, who was the head of engineering at DEC; David Liddle, who was the head of the Office Systems Division at Xerox when Ethernet needed promotion.

IEEE project 802, 3Com, and efforts to make Ethernet the industry standard

Metcalfe:

We started the IEEE project 802 around 1980 in order to get Ethernet adopted as an open industry standard. Then it got out of control and became a huge standards bureaucracy of unbelievable impact.

Colburn:

Was that a bit of a battle?

Metcalfe:

It was ugly. The battle was waged mostly between January of 1980 and December of 1982. It was three years of horrible, ugly infighting. It could not be avoided. It is essential, but not very pleasant when you are in it. The focus of the battlefield was IEEE project 802. Have you heard of DEC?

Colburn:

Yes.

Metcalfe:

I persuaded DEC, Intel and Xerox to adopt Ethernet as a joint standard. However, I could not get them to meet to discuss this standard because they were afraid that they would violate antitrust law. In those days IBM was being sued for antitrust. Howard Charney went to law school at Santa Clara after graduating from MIT and sued IBM on behalf of Memorex – or Memorex on behalf of IBM, I have forgotten which. Charney was an antitrust expert, so I called him up. I said, "Howard, I can't get these three companies to meet," and he said, "Oh, that's easy. They are just not allowed to do this, this, this or this and then they can meet." One of the conditions was that marketing people not be in the room, because they would be tempted to divide territories or set prices. The meeting must be open, and a representative of the federal government should be there to be sure nothing untoward happens. The goal of the effort should be an open industry standard. There were a couple other conditions. Anyhow, I called up DEC, Xerox and Intel and told them, "All you have to do is these five things." They checked with their attorneys and their attorneys said that was right, so then they started meeting. We formed a DEC-Intel-Xerox consortium to develop an Ethernet standard, the specs for which appeared in a blue book on September 30, 1980. Then we submitted that book to the IEEE 802 Committee, which formed in December of that year. Then IBM and General Motors both stepped in and said, "We do not think Ethernet is the right standard for local area networking" and a battle ensued. After a year of hell, the IEEE committee divided itself into three committees – 802.3, 802.4 and 802.5. The 802.3 committee was for the Ethernet, 802.4 for the General Motors cable television-based token bus local area network and 802.5 for the IBM token ring. The IEEE made the decision to make three standards instead of one. It was the way out of the deadlock.

Colburn:

Definitely. Yes.

Metcalfe:


Audio File
MP3 Audio
(433 - metcalfe - clip 2.mp3)


The 802.4 went on to waste millions and millions of dollars and then crashed and burned immediately. The token ring was bit more successful, but eventually succumbed. In the end, the Ethernet won as a result of all that. During that time I learned that professors could be bought. IBM and General Motors needed to show people that Ethernet did not work, so they paid professors money to write papers that were total baloney, proving that Ethernet did not work. We had hundreds of them in operation and working just fine, but when a professor from a prestigious university writes a paper, it can have an effect. That professor, a very well-respected man, was paid by IBM to write papers to attack my baby, and I will never forgive him for it. He was not the only professor that was paid to write bogus papers attacking Ethernet. There were many. No getting around it, it was hardscrabble competition, and the best man won, even if I do say so myself. Ethernet won out in the end. That became the standard. In 1979, I left Xerox to pursue entrepreneurial ambitions. I spent 1979 as a consultant to MIT. There I fell into a nest full of token ring people. MIT was a token ring shop. As a result of my wanderings and consultancies related to all that, I managed to put DEC, Intel and Xerox together and get that standards effort going. As the standards effort got going I went to the venture capital community in Silicon Valley and said, "You should fund me to start a company to exploit the new standard that is certain to result from this collaboration among DEC, Intel and Xerox through Project 802." I got the funding and 3Com was founded on June 4, 1979. I stayed there through 1990. There were a flock of companies founded in 1979 related to local area networking. 3Com was the first, but days later SciTech was founded, and a month later Ungermann-Bass was founded, and Interlan a little while after that. Those were the first, and many others came after that. We ended up fighting, and it was really fierce. There is a book about this. It is a book about the history of Ethernet. I don't have it here, but the beauty of this book is that it very accurately retells the story of the fierce competition among all the companies that cropped up around Ethernet, of which 3Com was only one, but 3Com prospered. I wish I could remember the name of this book, but it should be easy to find. Do you know that we had this 30th Anniversary celebration at Xerox PARC?

Colburn:

I don't know that I did.

Metcalfe:

On the actual day, May 23, 2003, we celebrated the 30th Anniversary of the invention of Ethernet. All the old timers came out and many of the giants showed up, including Paul Baran, Larry Roberts, Bob Kahn and Vint Cerf were there. That book was present at the meeting – the book that described this hard- fought battle through the 1980s of the many companies organized to promote the ethernet and internet. Here is a story. In 1981 or 1982, we came across a company called Seeq Technology that started up making chips. In our conversations with them we figured we could build an Ethernet chip. Intel was trying to build an Ethernet chip, but there was no sign of their being finished. We were a little company and ambitious, so SEQ said, "Why don't you help us design this chip since you are the Ethernet experts? Then you can sell it." We helped them design the chip and they went off designing it, and then we asked ourselves, "What are we going to do with this chip when we get it? What product are we going to bring to market?" We went to three big vendors. We went to Apple and said, "You need Ethernet for your personal computers" and they said, "You're right. Here's a contract. We would like you to build thirty of them." It was a big deal for us, maybe $3 million. DEC said, "We're an Ethernet company, and you, 3Com, are a dirtball little company, so we are going to have to think about it for a really, really long time." IBM said, "Screw you. We're doing the token ring." We set out to build the Apple thing and eventually did build them and we sold them to Apple. They are all in the bottom of the Bay in San Francisco right now because no one ever bought them. It was a stupid idea. We did exactly what Apple wanted, which was for us to build a box to which they connect Apple IIs and Apple IIIs. They did not have Macintoshes yet. And the box was about the same size as an Apple II or an Apple III. It was not a good proposition. They should have taken our advice and had us build them a card. The customer is not always right. The vendor who said yes was a dead end. DEC jerked us around for two years, during which time they taught us how to manufacture things. They kept sending armies of guys to see if we could manufacture this card of sufficient volume, quantity, quality, repeatability, serviceability and everything they needed in order to have a product. In essence they taught us how to run a factory. They would say, "We would like to meet your statistical quality control engineer," so we would say, "Howard, go hire a statistical quality control engineer." We built a great factory in response. In the end DEC decided to build their own cards, and they eventually did – two years after we had ours ready to go. They lost two years in the marketplace, and that was the beginning of the end for them. It took another decade before their end really came. IBM had told us to go away, so we were free to do whatever we wanted for IBM PCs. We built a card to go in the IBM PC, and that was the one. They announced the IBM PC in August of 1981 and we shipped the first EtherLink -- the first Ethernet for a personal computer -- in September of 1982 and it took off. Until then we had been selling hundreds of boards per month to Sun Microsystem. Hundreds. I think 3Com makes several million EtherLinks per month now. It is an exhilarating time when people want what you've got. It is really fun. We went public in March of 1984 and kept growing. By 1990 it was time for me to retire. I became a journalist. I was a journalist for ten years, and then I became venture capitalist, and here I am now. That completes that story.

Benefits of telecommunications for society and business

Colburn:

One of the themes of the Marconi Foundation is the benefits that come through telecommunications – the idea that where communications go, prosperity follows.

Metcalfe:

Absolutely.

Colburn:

You have already in a sense given a very good backing for that argument. Obviously the internet and the information it makes available to people and the empowerment that it gives to people has done incredible good all over the world. Are there particular aspects of that which you look back on fondly and point to as being especially rewarding or gratifying ways that Ethernet has been used? Are there things where you look back and say, "I am glad that someone thought of using it that way" or ways that it has been used that have been unexpected?

Metcalfe:

A series of little stories might do. I remember peddling Ethernet cards to a bank vice president. I do not remember the bank. This was when 3Com was little dirtball startup company. I was explaining to him how by plugging this card into PCs then they could share printers. Printers were much more expensive in those days. The PCs could also share disks. In 1982 the IBM XT came out which had 10 megabytes of disk, and it was very expensive so it paid to share it. I was convincing him that the printer and disk could be shared. He got it. He understood it and said, "I want to buy some of that." I wondered why this guy would take a chance on me, because I was not IBM and he was a banker. We had dinner and in a roundabout way he told me why he was making this investment. His strategy for advancement in the bank was to use technology to achieve things his competitors in the bank could not do because they were too conservative. It was his personal career strategy was to advance the bank by delivering better service to the bank's customers by using advanced technology that is out on the bleeding edge. Wouldn't you know, years later he was running the bank. It worked. We helped each other. Son of a gun.

Another event was the company picnics in the late 1980s got really, really big. People who worked at the company had spouses and children, and I could see that basically I was going to send all these kids to college. We had created twelve thousand jobs. More importantly, those twelve thousand jobs represented people with roughly twelve thousand spouses and twenty-four thousand children. My God, look at the economic impact just here. Then I remember the first time I was on a military naval vessel. Ethernet was running the length of this naval vessel. I said, "Aha. My baby is out there keeping the world safe for democracy. Son of a gun. Look at that." It is interesting. There is a series of stories involved with going to visit customers. It is not just the initial part when you sell stuff to them, or you are dealing with adventurous propositions, but years afterwards and after the chickens have come home to roost, and the bleeding edge is no longer the bleeding edge. The story has played out and the applications have been developed.

I still do it today. I was at the Leahy Clinic the other day getting my medical checkup. There is a computer in every room and I always look at the back, and son of a gun, there is an Ethernet on the back of it. I do not suffer that much, but I have asthma. For the time ever I saw my asthma on a computer screen. They are very proud of their brand new computer connected to a breath measurer. I took the test and they showed me, "Here's the plot for a normal human being, and here is the plot for you." Then they gave me an inhaler and I inhaled and they did the test again. "Ah, here is the plot for you now with this inhaler." That clinic is health machine, and there are thousands of people stream through that clinic getting fixed. Every one of those computers they are using for appointments, finance and MRIs are all on the Ethernet. It is gratifying to see that. I have always had an inferiority complex about being a plumber. Steve Jobs makes movies at Pixar [Animation Studios] and doctors cure diseases using asthma curves, but my life is about cables and electrical signals, so I always feel I am down in plumbing, and those guys are up there have all the fun. I expressed this sentiment to Steve Jobs when he had just come out with his first movie. I went to the opening of his movie. I said, "You guys have all the fun with computers." Jobs turned to me and said, "I want you to know that every pixel in that movie was carried over your Ethernet. Every single pixel." Thanks, Steve. It feels much better now. We start life sciences companies here at Polaris. These companies are going to cure everything – asthma, cancer, diabetes, etc. They are working on them. They sit at computers all day, and those computers spend all their time sending packets over the Ethernet all day long. Even though I am not going to be the actual doctor to cure malaria, AIDS and brain cancer, the people working on it are all using Ethernet. I think, "All right, that's part of my little contribution to the world." Marconi was right about the giants and he was right about making the world a better place through communications. Communications is associated with prosperity. There is a related theory I like where someone plotted over time the number of telephones per capita in a country. I wish I had this paper. Maybe it does not exist and it was a talk someone gave. There is a theory that, at a certain per capita number, like seventy or ninety, totalitarianism is impossible. They plotted the growth per capita of telephones in the Soviet Union and showed that where perestroika occurred was right on the number. The way the talk went, seventy was the number and here was the chart of Russia. Seventy – boom, perestroika. They have similar charts for China, which is under a murderous totalitarian communist regime. The hope is that by increasing the level of communication generally, and telephone in particular, and the internet, that communication makes totalitarianism difficult if not impossible.

Colburn:

Along those lines, have you ever seen the file of email messages that was collected by a history professor during the first attempt at the 1991 coup in Russia? When the security forces went through they smashed the fax machines and the telephones. They did not realize that computers could be used for communication. Those first three days of the coup the primary communication with the outside world was over the internet. Someone has gone through and collected the messages from journalists and the people and students who got messages out through their buddies in Finland and other universities. It is absolutely gripping reading, because it is from inside. From the styles in which they are written you can almost hear the different voices and recognize some of the people. There was one moment where they are crouching down under the windowsills in the offices of TASS because the tanks were rolling through, and a woman is writing back to someone in the West saying, "Oh, do not say that. I can hear the tanks going by outside." It was the very low point of the coup. Then it all turned out well. The reference is because the internet was the communications tool that was left in place. It was very important in successfully breaking up the hardliners in the Soviet Union. I will try to track that down for you. I think you would very much enjoy reading it. That is the proof of what you were saying. The internet there made totalitarianism impossible.

Metcalfe:

I need a high-minded motive for what I do. I am a venture capitalist now so I have a slightly different story, but when I was a guy building Ethernet, the story was that communication breeds freedom and prosperity, therefore the more of it the better. Shine a light. Want to get rid of the cockroaches? Just shine a bright light on them and they will skitter away. I sort of feel the same way about entrepreneurship and more generally in our particular role as venture capitalists.I am trying right now – I do not know that I will, but I'm trying – to start a supercomputing company. It is a pretty daunting task to start a computer company – especially when you are out to make money. I am excited about this project, and one of the things I have determined through the hundred due diligence calls I have made, is that progress in the world on many fronts is limited by the amount of computer power than can be delivered. In the basic sciences – physics, astronomy, chemistry and biology – they are all trying to build Linux clusters to compute the solutions to their problems. Now, in medicine and disease, it is all computation. I am exaggerating, but biology is now nearly all computation now. Everyone is a computational biologist – doctors, DNA analysis, genomic analysis, sugar synthesis, disease exploration, molecule databases – it is all computation. I want to start this supercomputing company because it is going to deliver teraflops of computation cheaply and conveniently, and therefore move back all the frontiers. All of them are dependent on teraflops. I have determined by talking to people, that those who are on the cutting edge use lots of teraflops. Petaflops is what they actually want. They are only getting fractional teraflops at the moment. They have a budget, and every year they spend that budget on as many teraflops as they can afford. If we start this company and improve that situation by a factor of ten or a hundred – which is what we are planning to do – we will make a lot of money, which is our job. We will push back the frontiers of all those areas. Automobile design, physics, chemistry, biology, history and the arts are all teraflop-hungry. That is pretty motivating, and it is going to be fun if we ever get around to starting this company to go sell it, go visit the National Security Agency and the National Science Foundation and the oil patch in Texas and the general hospitals and the other places where they are trying to cobble together as many teraflops as they can afford. We'll say, "You can have a hundred times that for half price." Whoa. Maybe we can cure cancer tomorrow, instead of the day after. That's pretty motivating.

Colburn:

Yes indeed.

Metcalfe:

We have a $20 million term sheet. I was over there at the company today presenting our term sheet. The problem is here.

Colburn:

That is your number one?

Metcalfe:

We need two more investors before this gets going. That is our problem. As I said, starting a computer company is daunting.

Colburn:

Yes. Maybe they will do that instead of sending people to Mars, which may not be quite as immediately useful as something like this.

Metcalfe:

That investment will help send people to Mars.

Colburn:

Exactly.

Metcalfe:

I am a fan of space exploration, but I do not know enough about it to enter that particular argument. I liked when Kennedy wanted us to go to the Moon. It annoys me that we have not been to the Moon recently. We've been there, done that, so what's next? Mars. Let's go there. MIT is full of enthusiasm for going to Mars. Kids are doing projects trying to get to Mars. I like the effect it has on people. NASA is a big customer for teraflops.

Colburn:

Yes, I am sure they are.

Metcalfe:

They are an old bureaucracy and need some shaking up. This program to go to Mars could only be accomplished by some housecleaning at NASA. Eventually government labs become jobs programs and more than mission programs. I think that may have happened at NASA. It is just natural. It happens at all agencies at all times. They become jobs programs and people forget the mission. Their job becomes to have a job rather than to get to Mars in the best possible way. The mission is not to provide jobs; it is to get to Mars. That is the job. Providing jobs is not the job. I am never going to go to Mars myself, but maybe my kid will. Tell me again about your project.

Colburn:

The project is to celebrate what is being called the Marconi Century. It is actually going to be more than a century, from the transatlantic transmission and his work prior to that, and celebrate the century of telecommunications. The IEEE History Center is writing a book on it. We are going to make what we hope will be a very elegant coffee table type of book.

Metcalfe:

Is the IEEE History Center a building somewhere?

Colburn:

It is a building on Rutgers campus in New Brunswick, New Jersey. It is close to the Edison Papers, which are also at Rutgers. There are six of us who are researchers, historians and archivists whose job it is to be both the corporate archives for the IEEE and IEEE history and also to research and preserve history of electrical and computer technologies. The Marconi Foundation has given us a grant and asked us to work with them to produce this book and collect the oral histories of the Marconi Fellows.

The Marconi International Fellowship

Metcalfe:

I just learned yesterday who the next Marconi Fellows will be.

Colburn:

I have not yet heard. Probably the email is back in my office.

Metcalfe:

You want me to tell you?

Colburn:

Sure, if it’s not confidential.

Metcalfe:

The two founders of Google. Sergey Brin and Larry Page.

Colburn:

That is very interesting.

Metcalfe:

Clearly communications.

Colburn:

Right. They are the ones who can find the stuff for you.

Metcalfe:

Anywhere in the world. It is amazing. Anyway, it is a great choice and very timely and deserving.

Colburn:

Yes.

Metcalfe:

I am going to this event. The theory is that you return the year after you are given the Fellowship to report on the work that you have done as a result of the Fellowship. As far as I can tell, no one actually alters his or her life in any way as a result of winning the Fellowship. One does not set aside the year to work on one's Marconi Fellowship, although I think that may have been the founding principle of it. I will go back and give a speech in Italy in the fall at which the Google boys will get their Marconi. The guy who spoke when I was getting my Marconi was Tim [Timothy] Berners-Lee. He did the nicest thing. It was at the Waldorf Astoria in New York. As I mentioned earlier, my grandparents arrived around 1900 in Manhattan. They lived within walking distance of the Waldorf Astoria, so it's been in our family thought since they arrived a hundred years ago. I invited my mom, dad and sister to come to the Marconi event. My sister is a nurse. My father and mother are retired, never went to college and have no idea the meaning or significance of Marconi, and barely understand the concept of the Ethernet. They do not know. They went to this black tie event at the Waldorf Astoria where their son was getting an award, so they had a great time. They were, and still are, clueless about the significance all these different people. They are from a different world. Tim Berners-Lee got up to give his Marconi address and spent 15 minutes of his half hour talking about me. In a way – although I am sure he did not plan it – he was explaining to my parents what I did. It was like the closing of a huge loop. When my parents got married they had two goals: retire and send their children to college. They achieved those goals. To have this distinguished gentleman stand up and gave this exposition on what their son had done was like a life validation. It was just wonderful. He is a great guy.

Colburn:

I am hoping to get to interview him. I know he is very busy.

Metcalfe:

He is very busy, very much in demand and very hard to reach.

Colburn:

Yes, I'm sure.

Metcalfe:

He will do it. He is a nice person at the core and just busy as hell I'm sure. Celebrity really screws things up. Having been knighted, he is a first class celebrity now. I hasten to point out that he occupies the 3Com Founders Chair at MIT. That was endowed to the tune of $3 million by the founders of 3Com. We are proud of this, and it is the first chair ever at MIT that can be held – and is preferentially held – by someone who is not tenure track faculty member. Amazingly, Tim Berners Lee fits that bill. He is a senior research scientist. He is not a member of the faculty. He gets all these awards. We came up with a chair for him, and had to fight because MIT did not want it with this type of stipulation. For $3 million they took it. There is a stipulation that they can only give the chair to a faculty member if they cannot find a deserving member of the scientific staff. They hate that. I am anti-tenure, so that is part of my motivation there. Maybe that was why he gave that 15-minute speech. I will have to find out whether the chair has done him any good. He has had it for a couple of years now. Did it make any difference? Would you please ask him?

Colburn:

I certainly will. Thank you so much for seeing me and for your time. This has been a wonderful interview.