Interview: Robert Metcalfe, recipient, National Medal of Technology

Bruce Taylor spoke with Robert Metcalfe on the occasion of his receiving the National Medal of Technology for 2003 on March 14, 2005. This is an edited transcript of that conversation. You may also listen to the original interview here.


On March 14, Dr. Robert M. Metcalfe will receive the National Medal of Technology for 2003, presented by President Bush in ceremonies at the White House. The award is for his pioneering work while at the Xerox's Palo Alto Research Center, or PARC as it is commonly known, that in 1973 resulted in the invention of Ethernet. Ethernet is the local area networking standard on which he shares four patents. In 2003, Ethernet's 30th year, 184 million new Ethernet connections were shipped globally for $12.5 billion in sales value.

Bob is an MIT graduate and received his Ph.D. in Computer Science from Harvard in 1969. Today, Bob is a general partner in the venture capital firm Polaris Ventures. He specializes in Boston area based information technology startups. Prior to becoming a VC, Bob enjoyed three quite distinctly different careers. From 1965 - 1979, Bob was a research scientist and engineer. He was the founder of networking company 3Com where he served as Chairman and CEO. For the decade ending in the year 2000, Bob took a turn in the IT publishing world as both a publisher and an industry pundit, first as CEO of InfoWorld, and then for 8 years as an Internet column writer. He is co-founder and continues to be actively involved with one of the foremost global conferences on the intersection of people and technology, PopTech in Camden, Maine. He also serves on the boards of IDG, IDC, and MIT's Technology Review magazine.

Bruce Taylor: Bob, welcome to the program. And for all of us here at ITworld, congratulations.

Robert Metcalfe: Thank you very much.

Taylor: You're someone over the years who's been honored in so many ways for so many contributions that it may appear to be that this is just one more. But I promise you I don't see it that way. And so for you, what does it mean to receive the National Medal of Technology from the President?

Metcalfe: It means that my mom and dad get to go with me to the White House. That's actually the biggest thing it means to me. To my parents who have lived the American dream and worked hard to send their boy to college and now it's sort of coming full circle for them.

Taylor: The National Medal, according to the press announcement, and as we know, is to recognized innovation that has advanced the nation's global competitiveness. Have you ever tried to imagine the scale of what Ethernet has made technologically and economically possible? Has anyone?

Metcalfe: Well, I think I have been through 10 or 15 ah-ha experiences where I realized that Ethernet was really catching on. Like last year for example I learned recently from IDC that there were 200 million new Ethernet switch ports shipped worldwide, 200 million. And that's a really big number. I have never conceived of numbers that big. Even the last time I had an ah-ha experience. And Ethernet doesn't seem to be tapering off; it seems to be continuing to proliferate and evolve.

Taylor: And has anyone ever tried to put an economic value to Ethernet?

Metcalfe: I've not seen it. I know that the revenue number associated with the 200 million ports is on the order of $12 or $13 billion, but that's just for the ports. And that's just in one year. So I don't know. I'm not sure how you'd go about assessing the economic impact of something like Ethernet.

Taylor: So depending on what birth date one chooses to use, Ethernet is now 30 years old, is that, is it 31 or 29?

Metcalfe: The birth date that I have chosen for many years is May 23, 1973, which would make it 31 years old.

Taylor: 31 years old. So now I have to ask you, casting back to those days at Xerox PARC, and perhaps even before, what in the world were you thinking? How'd this happen?

Metcalfe: Well, when I graduated from MIT and went up the river to Harvard to graduate school, I was a new grad student. And what we grad students do is when we arrive we start looking for how we're going to leave, what our dissertation research is going to be about. And as it turns out in 1969 when I arrived at Harvard, the big thing in university research was a thing called the ARPA Computer Network, the ARPANET, which I think of as Internet 1.0. And so it was obvious that I should get involved in that somehow.

So I started building hardware and software as a grad student to contribute to the early evolution of the, let's call it the Internet. Although that's disputable. There's a big argument about whether the ARPANET is the Internet or vice-versa. I think of ARPANET as an Internet 1.0. So that got me into the networking business. And then when I went to Xerox PARC in '72 I was the networking guy. And my second grand stroke of luck was to be the networking guy at Xerox PARC in 1972 because that was the first time ever that anyone had been ever given the job of how do you network a computer on every desk because in 1973 that didn't occur in the real world, a computer on every desk.

Taylor: And David Boggs was your close partner on that.

Metcalfe: Absolutely. He and I built the first Ethernet together. And I tended toward the theoretical, mathematical, computer programming, some hardware. And David sort of started from the software and went into the hardware into the analog. We were a perfect team.

Taylor: At the time that you went to Xerox PARC we were still living in a world where most enterprises, if you could call them enterprises at the time, had perhaps several large computing facilities scattered around the world, disconnected. Three, four, five mainframe computers maybe a PDP-11. And what Ethernet essentially opened up was the possibility for something that hadn't happened yet, and that was a very large number of computers able to be stringed together to do work. When did it occur to you or how did it occur to you that that was the opportunity?

Metcalfe: Well remember, I had worked for a couple of years already, three or four years on the ARPANET, which was connecting mainframe computers at a distance, across the country. And incidentally, we did that in order, ARPA called it resource sharing, the Advanced Research Projects Agency, the Department of Defense funded all that work so that they thought they were going to share many computers across the country so they wouldn't have to buy so many for the research scientists. But that didn't turn out to be what ARPANET was used for. By then it was clear that email was the killer app, that it is, it was communication rather than shared computation that was the motivation.

When I got to PARC, we were planning to put, as I mentioned earlier, a computer on every desk. So you could think of the building as the continent and the goal is to connect things together. But we had a very real and different kind of incentive. We were building what I think is the world's first laser printer. It was a page per second 500 dot per inch laser printer made out of a Xerox copier. And it printed gorgeous documents and ran at a page per second. So the question is how would all these PCs around the building share that printer? How would we keep the printer busy it was so fast? We couldn't imagine carrying our disks down there, because in those days disks were two feet across. So the first motivation for the local area network connecting all those PCs in the building was to share that laser printer. The second motive was to hook up to the Internet so that we could send email to each other.

Taylor: And you saw that then. You saw email as the killer app.

Metcalfe: Definitely. Yeah. I've been an email user since 1970, maybe even before, it depends on what you count as email. And even then I was hooked.

Taylor: Who were some of the great contributors to not just your own work, but other work in the networking field, heralded or unheralded that you can recall?

Metcalfe: Well, the granddaddy of the Internet, or perhaps the great granddaddy, and I don't want to get into a fight about this, was J. C. R. Licklighter at MIT, who was at ARPA in the early days. And then there was Bob Taylor, who had been at ARPA and then went to Xerox PARC. And then there was Barry Roberts, who actually managed the construction of the ARPANET through Frank Hart at BB&N. And then there were the people I worked with at MIT, Licklighter, my mentor there was a guy named Al Vessa, Bob Bressler, who later became the CTO of Sun. Tom Knight at MIT, who helped me get my hardware to work. And then at Xerox PARC of course there was a bunch of computer science stars, Butler Lamps and Chuck Thacker, Alan Kay. Then my partner, we were called the Bobsy Twins there for a while, but Dave Boggs and I hang out together for a few years building the first Ethernets. What I'd like to say about this is that a lot of people have been involved with Ethernet over the years, and I like to say that by 1981 there were people buying Ethernet whom I had not met. And by 1985 there were people inventing the Ethernet whom I had not met. The thing really broke out and started proliferating and attracting a great deal of intellectual energy. So I could go on for hours naming people who have contributed to the success of Ethernet.

Taylor: So now if we go back to that point in time where you just said there were people who were buying Ethernet who you had not met. That was also at about the time where the raging discussion on the street was between Token Ring and Ethernet. In hindsight, was it a technological reason or a cost and market forces reason that caused Ethernet to win?

Metcalfe: Well, we could if we wanted have a whole discussion about how Ethernet was superior to Token Ring technically. It was simpler and cheaper because Ethernet understood its place in the protocol hierarchy, and so it only did what it had to do. It didn't replicate a lot of functions that were needed elsewhere in the hierarchy, which was one of the mistakes made by Token Ring. But that's not really the answer to your question because I think what caused Ethernet to win was its business model, not it's technology.

So, Ethernet was organized around a de jure standard made by IEEE Project 802, and that was Project 802 before the dots. There were no dots in those days. Project 802 was just standardized LANs. So a de jure standard. And IBM in those days was the Microsoft of its time, the dominant anti-competitive monopoly that it was. So, it didn't really believe in that model. And although it was eventually forced to make Token Ring into an 802 standard, 802.5 as opposed to Ethernet, which was 802.3, IBM's heart wasn't in it. And for years afterwards they didn't behave consistently with the idea that Token Ring would be an open industry standard. Then the Ethernet model was based on fierce competition among contending suppliers. And of course soon there were hundreds of Ethernet suppliers. But for many years IBM was 90% market share at Token Ring, and that notion of fierce competition driving the perfection and evolution of the standard was not as big a factor in Token Ring as it was in Ethernet.

Taylor: So in that fierce competition, there had to be several major tipping points where the various generations of Ethernet and the 802 standard altered to create a whole new level of performance. What were some of those? And who lead those besides 3Com?

Metcalfe: There were many people involved, not just me and 3Com. The first Ethernet was at PARC. It ran at 2.9 four megabits per second and David Boggs and I built it. The second Ethernet was a 20-megabit per second Ethernet that we built in the systems division of Xerox. Ron Crane, my buddy there helped build that. But then by 1979, as I was leaving Xerox, we found DEC and Intel to join us in making Ethernet a standard. And in order for Intel to implement the standard on its chips we slowed Ethernet down to 10 megabits per second to get it on the chip. And then it came out as IEEE 802.3 with a thick cable, almost a half-inch diameter coaxial cable as the standard ether down the middle of the corridor. But by 1982, 3Com was shipping a version of Ethernet that was exactly compatible with everything else, except it ran on a quarter inch thick piece of coaxial cable and didn't use big heavy connectors, but small tiny little connectors.

And by then, Sun was using 3Com cards to get started, so 3Com and Sun sort of took Ethernet in its initial days and started projecting it. And then we got into the IBM PC with our thin cables. But then other companies starting interfering, excuse me, intervening, competing with us on Ethernet, which was really annoying. Competition usually is annoying. And so guys over at, it became Nortel, before that it was Bay and before that it was Synoptics came out with a twisted pair version of Ethernet in the mid '80s. And that was the death knell of Token Ring because until the twisted pair telephone wiring version of Ethernet came out, Ethernet was a shared coax system and the Token Ring method of a star of twisted pairs was found to be superior. So Token Ring, both because of its wiring and because of IBM's support was even in the mid to late '80s was used as the inexorable industry standard that would eventually kill Ethernet. But then when Ethernet switched to twisted pair and then got hubbed and eventually switched, then the remaining features of Token Ring, benefits of Token Ring went away and then Ethernet inexorably killed Token Ring. Hooray.

Taylor: So we have become network speed and capacity junkies. Is there any end in site?

Metcalfe: Nope. So the Ethernet was, I mentioned 2.94, then 20, then 10. Briefly when it was being reinvented with twisted pair it went from 10 to one megabit per second. But that didn't last long. So it popped back up to 10. And then came 100 meg, fast Ethernet, 100 megabits per second. And that would have been in the '90s. And then we went from 100 to gigabit Ethernet, so my McIntosh here has an RJ45 port on it that will either be a 10, 100, or 1000 megabit per second Ethernet, depending on what hub you plug it into, what switch you plug it into. And 10-gig Ethernet is now in volume being used for long haul and back offices and so on. And we're now wondering about the next step. It will either be 40 or 100 gigabits per second. There is no end in sight. But speed of course is not the only evolution of Ethernet, but it's certainly in the LAN, now we routinely use 100 megabits and gigabit in the office.

Taylor: So when you say it's not the only evolution, what other major evolutions can you point to that really mattered in terms of performance?

Metcalfe: Going wireless is one of the evolutions of Ethernet. And we're seeing that in many ways, not just WiFi. And whether it's Ethernet or something else that's akin to Ethernet, the whole movement of packet Internet connectivity, which sometimes is synonymous with Ethernet and sometimes it isn't, that's proliferating. Then you see Ethernet is now going downstream below PCs. So, embedded microprocessors are now being networked with schemes that vaguely resemble Ethernet, vaguely or not so vaguely resemble Ethernet. The so-called ZigBee Standard, 802.15.4 for embedded mesh wireless networking is a direction. There are eight billion microprocessors shipped every single year, about 98% of them aren't networked yet. And that's what ZigBee is about, it's about networking them. So that's another direction that Ethernet is going.

Taylor: In your opinion, what's missing today, whether in the United States or elsewhere, in terms of research? If you had one systemic problem or issue that you could wave a wand and fix, what would it be?

Metcalfe: I think, I don't often think in these terms, but of course getting the National Medal of Technology sort of provokes you to think big like that. So I've had one thought, which is that we're in transition now on research. Until now, and for a very long time, especially in the computer area, we've relied on monopolies, business monopolies to fund our research because only monopolies can afford to do research. So you had the AT&T telephone monopoly supporting Bell Labs. You had the IBM computer monopoly supporting Watson Research. You had Xerox's copier monopoly supporting Xerox PARC. Today you have Microsoft's Windows monopoly supporting its fantastic research activities. But I don't think monopolies are worth it. Nor do I think they're the best way to proceed with research. So I see a transition going on away from relying on monopolies. In fact, we should have fewer monopolies. They're very hurtful to our economy and to innovation.

Taylor: Including in research, the government monopoly.

Metcalfe: Well, the government is the ultimate monopoly. And I claim that research has to be funded by the government. And it is a monopoly, and they can afford it because it's a monopoly. And that the best way for us to get research done now is no longer relying on a monopoly, corporate monopoly research labs, but relying on research universities funded by government to do research. So, if I could have one thing it would be a blossoming of government support. Not for government labs and not for monopoly support of corporate labs, but government support of research universities, with the goal of not jobs program. It's a problem with government funding is it often becomes a jobs program. But I mean government money focused on supporting research universities to do the most urgent research on our agenda. And why are research universities so good at this? And the answer is very simple. It is the job of universities to graduate people. Students who then graduate, they are the best vehicle for the transfer of technology from the lab into companies and markets. So I'd do that if I had my druthers.

Taylor: Bob, as always it's a delight. Thank you so much for the time today. And again, our great congratulations.