Previous StoryNext StoryTable of Contents



The PreText Interview









































































































































































































































































































































































































































































Frank Heart led the team that designed and built the ARPANET, the network of linked research computers that later developed into the Internet of today. ARPANET was a project of ARPA, the Advanced Research Projects Agency at the Defense Department. The point man at ARPA, overseeing the entire project, was Larry Roberts.

Heart and his team at Bolt Beranek and Newman (BBN), a private company in Cambridge, Massachusetts, won the contract to actually build the system. They designed and installed the machines called Interface Message Processors (or IMPs) that allowed the computers at the university sites to communicate with one another. The first node on the Net, at UCLA, was hooked up after nine hectic months in 1969. The first computer link was established between Stanford and UCLA. The University of Utah and UC Santa Barbara were connected soon after.

Heart worked directly to extend and debug the new network for more than ten years.

Heart, 68, retired from BBN in 1995. He spoke by phone with PreText Magazine editor Dominic Gates


***


The Building of ARPANET
The Military and ARPANET
The State of American Scientific Research
How the Net Changed
Jon Postel and Management of the Root Zone
On CyberCulture
Where Credit is Due
Retirement


***


The Building of ARPANET

Gates: What was your personal role?

Heart: I was the principal investigator at BBN for the IMP contract, and I stayed on as manager for 10 or 12 years.

Gates: BBN's job was what, to build the IMPs?

Heart: We really did a good deal more than that. The specific contract was to build a four-node network, including those interface message processors, and to permit it to expand beyond that. Indeed our major job was to decide what to use, to buy, to modify, to deliver, and to write software for those machines.

We had to [deal] with the phone company. We had to [deal with] all the terminal equipment that connected the phone company to the computers. We also had to play a large role in the discussion with the hosts about the software in the host computers.

So we really viewed ourselves as the systems engineers for the net. But we had a lot of ancillary responsibilities that were in some ways equally important; that is, if they hadn't been worked on, it wouldn't have run.

Gates: It wasn't until 1973 that ARPANET developed into the Internet. In what way was the ARPANET the progenitor of the Internet?

Heart: I don't think there is any question about that. The ARPANET was the true progenitor. At first the ARPANET was really viewed as a very experimental thing, and no one paid an enormous amount of attention right away. After a while, the ARPANET began to be a tour stop on everybody's visit around the world and there began to be other networks built all over the country, and all over the world. I don't think there is any date when the Internet came into being. I think the Internet was a thing that grew like topsy, without any real beginning middle or end. I don't think it has even really begun to grow yet in some senses.

So I think that what happened was a number of other nets got built; some with government support, and some non-US government-supported nets. And then it became obvious to everyone that it would be nice to connect those networks.They began to get connected, first in ad hoc ways, and then in a more orderly way. And that became the Internet. But I don't think there is a date when the Internet started.

Gates: I was identifying the start of the Internet with the writing of the TCP/IP protocols that allowed those various networks to communicate with one another.

Heart: That's certainly fair. I think that certainly was a major, major boost to the growth of the Internet; but there were connections to networks even before that.

Gates: Which ideas would you single out as being the most important and where would you give the intellectual credit?

Heart: That's a tough one. There were clearly several major seminal changes between the ARPANET and all prior communication networks. The one that was the most obvious was that it was a packet switching network. That is to say, that it broke messages into pieces, and sent them down different directions and different paths. To my knowledge, that had never been before in any network that was really used by anybody.

As you probably know, contemporaneously, there was a tiny in-the-building network at the National Physical Laboratory in London, which had some characteristics that were similar, in that it was breaking messages up into packets. But it never got to the size, at least not at that time, where it was used by anyone. It was an experimental thing, in a building.

Packet switching was certainly the single largest new idea. But the project developed many, many things that were new. I think a lot of the remote debugging, and remote management of the network, had not been done before in anything like the same way. The development of protocols to use the network hadn't been necessary ever before; so they were certainly new. There were many new things that came out of that development.

The single one that isolated it from all previous work, though, was certainly the packet switching notion.

Gates: You are a technical person. [Heart has degrees in electrical engineering from MIT.] So, as well as overseeing everybody's work, were you doing nitty gritty work?

Heart: Absolutely. And in addition, I came to the project with some special background; having worked at Lincoln, and having built many computer communications systems before the ARPANET at Lincoln -- for seismic arrays and for directing antennas and all kinds of other thing -- I came with a feeling for reliability in those systems, and for certain kinds of design approaches that weren't necessarily present in any of the other people. Sure, I had a very direct technical involvement, certainly in the early stages.

One thing I'd emphasize is, the project paid attention to reliability to a very unusual extent, and it was very very important that it did so. Because as the network grew, you're talking about an unmanned network of dozens of sites all interconnected, which had to keep working. Which became like the telephone system or the electrical system; it became a utility. So there was enormous energy put into reliability, and I was the prime mover in pushing for that degree of attention.

Gates: These IMPs were what, the forerunners of routers?

Heart: I think that's right. I think that their job was to take messages from hosts, to break those up into pieces to get them to destinations, and to put them back together again. The current terminology would be closest perhaps to routers. The analogies aren't exact, but yes that's as close as you can get, probably.

The original machine wasn't fully militarized but it was hardened, which meant it had much better RF rejection and power-fail handling . . .

Gates: What do you mean, "militarized"?

Heart: Well I mean it was built to military specifications, nearly. It was a big steel cabinet with I-bolts, so that you could put it on a submarine or an aircraft carrier. And it was in a big thick steel cabinet, and had very few switches and lights; and it was built to have much better rejection of local electrical noise, RF noise; it had a very high quality soft power failure, so if the power went off the machine would come to a soft stop, and go right back on again when the power came on. I call it a "hardened" machine, it wasn't a fully mil-spec machine. But it was hardened.

And there was all kind of stuff put into it. For example one could remotely, from Cambridge, cross patch the connections to any one of the host or phone line ports.

One could do remote debugging of the machine from Cambridge: actually look in the running memory, as it was running, and do debugging. If the software bombed in some way, it had a little funny thing called a "watch dog timer." If the program didn't reset that time at intervals, the machine would go into a restart mode.

It was possible to reload the whole program remotely, from Cambridge.

And we did that for all the software releases. All the software releases in the system were done remotely over the phone lines. The machine introspected, that is, every few seconds it sent a little report back saying how it was feeling, and what it was doing, and some statistics about its performance.

There was tremendous effort put into that. Effort was similarly put into a monitoring system at BBN to watch over that network, and be able to debug that network. I think that attention to reliability was absolutely crucial in the system being able to run the way it did, as an unattended system all around the United States and even overseas. That was very unusual. I don't know of any system that had been built by that date that had that kind of attention to unmanned multiple computer operations.

Gates: Tell us more about that first IMP machine, the one you describe as not fully militarized but hardened?

Heart: The machine was based on a Honeywell 516. However it had a number of special interfaces, which BBN people designed and Honeywell built and put into the machine. BBN at the time didn't have hardware construction capabilities; so it purchased the machine from Honeywell and convinced Honeywell to take our designs for special interfaces and implement them in the machine. That entire process was done in nine months. It was nine months from the start of the contract to the delivery of the first machine to UCLA with special interfaces designed, implemented, and integrated into that machine.

The Military and ARPANET

Gates: Before you joined BBN you worked at Lincoln Laboratory on Air Force projects called "Sage" and "Whirlwind." What were those?

Heart: I had two jobs in my life. I worked at Lincoln Labs for 15 years and then I worked at BBN for 28. Lincoln Laboratory was established by MIT in response to an Airforce request to work on an air defense problem. That was Lincoln's reason for coming into existence. The air defense problem, when I first went to Whirlwind, in 1950 or something, I was a research assistant. Whirlwind was being used for two purposes. It was being used for a wide range of general civilian purposes, and there was a classified project at Whirlwind to build a thing called the Cape Cod system which was the progenitor of the Sage system. That was an Air Force-supported project, and Lincoln was heavily Air Force-supported for most of its life. Probably still is.

Gates: What was SAGE?

Heart: SAGE stood for Semi Automatic Ground Environment. It was specifically a system of computers and radar and phone lines and people and things to do air defense in the United States.

Gates: How important was military input into the ARPANET project, which was of course for the Defense Department?

Heart: I think the influence was close to zero. I really do. I believe that ARPA was supporting computer research in the United States for the general benefit of the military, and for the general benefit of the country; and that the people at ARPA saw those two things as almost synonymous. My understanding of the original basis for [ARPA program manager Robert] Taylor and others' interest was that they wanted to consider how to tie together their research sites, which were mostly at universities, partly to do resource sharing, partly to avoid buying a new computer for every university.

To the extent that there was a military influence, it was only the general military desirability that was behind a lot of ARPA's computer R & D. So I think the actual military sets of reasons where almost zero.

Gates: The discussion about this has arisen partly because it's known that when Paul Baran was preparing his early 1960s paper (suggesting some of the ideas behind packet switching) -- which originally did not gain attention -- he did talk to people in the military, and sold the idea of a distributed network partly on the basis of its robustness as a defense system.

Heart: At the very beginning I don't think the ARPA people knew Baran existed. When they began to talk about a network for resources sharing, and Taylor began becoming interested in not having 3 [disconnected] terminals in his office, I don't think those people knew Baran existed at that time.

Gates: Was there some inherent technical reason why the system became a distributed network, rather than a military reason?

Heart: Absolutely. There was a critical technical reason for doing that, and it had nothing to do with distribution for security purposes or in case of war.

If you have a phone conversation, as we're having, there are all kinds of pauses when the line isn't being used, and it's not an efficient way for computers to communicate. Computers would like to share the line so that they can intermix their messages, and take advantage of the bandwidth of the line. The phone company even does that to some extent with voice lines. It tries that same trick, but it isn't able to do it nearly as efficiently as you can with a packet switching network.

Gates: When BBN was pitching their proposal, and perhaps later when the project was under way, did you sometimes have to meet with military people?

Heart: No. There was almost no contact with any military personnel, for quite a while. In fact the first contacts with the military that BBN ever had was when a small number of military sites wanted to join the network. Those early military sites tended to be research sites. Although they were military in some sense, they weren't military in my normal use of the word. They were sites that were interested in research.

In other words, the first military sites to join the net wanted to play with it just as much as the universities did. It was only quite a bit later that military sites with no research interests wanted to join the net simply to use it. Or [some were] solicited to join the net, I guess partly because ARPA wanted their support in the net growing.

In the beginning there were just no military people involved whatever, except those that had been assigned to ARPA. ARPA gets its staff both from civilian and military sources. I don't actually remember anyone from the very beginning who was a military assignee.

Gates: One last question on this issue of military provenance. As ARPANET began, the Cold War was still on; and so much American science had grown out of the launch of Sputnik. All of NASA's achievements were impelled by the space race. Was there any sense of a race in your engineering field?

Heart: No. I think not; not in the way it felt to me. In that period the government, and in particular ARPA, was supporting the majority of decent computer research in the United States. They supported time-sharing; they supported the development of computers; they supported development of super computers; they supported network work. They were a major force in the United States computing community. That was in part due to the original reason for setting up ARPA, which was indeed a response to some of those pressures of the outside world, like Sputnik.

But the Information Processing Techniques Office, which Licklider first ran, was really taking the view that the support of computer research in the United States was an important national objective. For many purposes, including military purposes. It was part of that activity that led to the ARPANET. It was the support of computer research in the United States.

The State of American Scientific Research

Gates: Most of the progress in American 'big science' in this century has come from government, indeed mostly from Defense Department support. Elsewhere you've expressed criticism of the growth of bureaucracy since ARPA was created. Things like a culture where the lowest competitive bid is the one that's automatically accepted.

Heart: I am critical of some of that. As is usual, it has arisen with the best of intentions. The Congress, in its desire for fairness and equity, and all those good motherhood-and-apple-pie like things, simply has missed the critical point. Which is that if you want to get really good things done, you have to get really good people to do them. And you cant always decide that equity and fairness are the only good objectives. Performance and so forth, in my view as an engineer, I consider pretty important.

So I think there has been a very big change for the worse. You can see it all around you. The New York Times is constantly full of troubles with the air traffic control computer systems; troubles with the Internal Revenue Service computers; troubles with the Social Security computers, and so on and on. These are all examples of tremendous disasters in the computer field, and this is only the tip of the iceberg, that happens to get into The New York Times.

These big system problems are primarily political rather than technical problems. It's true there has been a very unfortunate change along those lines.

We might just take a moment to consider why ARPANET was so successful compared to other projects.

Let me list just a few things that I consider different or unusual. First there was a cost plus contract, instead of a fixed price contract. That meant that when the government wanted to change something, all they had to do was make a phone call. It didn't take weeks or months to go through the contracts office.

The project was entirely unclassified. The project had no access controls on people. Anybody who could get near it could log on. The project provided access as a free good; nobody had to make a tough cost-benefit analysis as to whether they wanted to try the network.

Now these things - unclassification, no access controls, and a free good - despite the fact that this was being done by the U.S. government and the Defense Department; it's really quite amazing in retrospect.

The system didn't have to be backwards compatible. We didn't have to live with 20 years of prior communication history. We could build the IMP the way we wanted to. We've already talked about it being a small bright team. But some of those things are unusual, if you look at other computer projects. And so I think it had some characteristics that it would be nice if one could emulate them in other government computer projects.

Gates: It seems like one very crucial link in everything was Larry Roberts.

Heart: No question about it. He not only was very smart, he was a very good manager, he knew how to let people go and when to get involved. He was technically extremely competent. Yes. If you had to pick one person who was the most crucial person, it was Larry Roberts.

Gates: Nowadays you have companies like Microsoft and Intel doing research. Do you think that private companies can fill in the gap that is left by the decline in government research?

Heart: I don't know that I would necessarily say that government research is declining. Just recently there has been a lot of play in the papers about how Clinton is pushing more R & D in this country. I think that private research tends to be different than government research in the sense that private research groups are usually working on relatively short-term product plans. They very seldom are taking on things that are much further out or much higher risk or have a much smaller potential market and so on. So no private organization is going to fix the IRS computer system or do the research that is going to fix the air traffic control system. They just aren't going to work on those kinds of problems. They are going to work on the problem that gets them a new browser or a new machine they can sell to millions of people

No private company is going to do the Genome Project.

Gates: You mentioned that at the National Physical Laboratory there was some research going on that was in some ways parallel, led by Donald Davies. How big was the British contribution? At one stage there was quite a bit of collaboration. How did the U.S. come to dominate so much?

Heart: I have perhaps a slightly odd view of British R&D.; I think the British have been incredibly clever at making use of extraordinarily limited resources. The National Physical Laboratory was building that network in my view almost with chewing gum and shoelaces. I mean their resources were just much, much less. But that didn't mean they were any less clever. They've always made interesting contributions to the computer field with seemingly minute resources. And I think that is probably true to this day.

Yet the work they were doing, in my view, did not contribute substantially to the ARPANET project. It was contemporaneous and it was interesting talking to them; we did collaborate and we did interact. But had they not existed, the ARPANET would not be any different, in my view. I don't believe there was a major contribution made there.

Gates: You've praised the dynamic of small teams of good people working together. It can't all have been so rosy.

Heart: The original BBN group that wrote the proposal was less than half a dozen people; and the group that did the software and hardware never did get to be more than about a dozen people. It was a tiny enough group that every one knew everything that was going on, and there was very little structure. I was in charge of this project; and there were people who specifically saw their role as software, and they knew a lot about hardware anyway; and the hardware people all could program. There wasn't a great deal of internal difficulty. There's always a little, I don't think you can get even 6 people together without having some kind of a problem, I can give you an example:

Bob Kahn was a theoretician, and was interested in having things as absolutely as good as they could be. Some of the other people knew they had to get something done by a given time, and they simply couldn't have it be as good as it could possibly be. So there was a certain amount of concern over whether people were just rushing ahead without getting it perfect. On the other hand, I understood that it had to get done by Labor Day.

So the people who were in charge of doing, sometimes won over the people who wanted it studied more and perfected.

On the other hand, it was amazingly nice. I look back on that period, and I think all those people look back on that period, as having the time of their lives. There was amazingly little argument or fighting. Also there was very strong management in Larry Roberts at ARPA.

Gates: Kahn left BBN and moved to ARPA. Do you think that environment suited him better?

Heart: Frankly, I think Bob did a magnificent job at ARPA. So in the end whatever we thought about that move, it was very successful. He became a strong advocate at ARPA for computer research in the United states and did a very good job there. To be candid, I thought he was more successful at that job than he had been with the local group.

On How the Net Changed

Gates: At some stage, the net became something different; it wasn't just a bunch of engineers and research scientists talking to one another. I have this image of West Coast hippies, who got interested in building community in cyberspace, and I'm wondering if there was a stage where the engineers, having built the thing handed it off to people who were more excited about what you could do with it, rather than what it was.

Heart: I don't think there was ever a handoff. The net as an engineering project really continued for quite long time. As the network grew, and as it began to appear in many more places, there were indeed other influences that began to be very important in the whole thing. There were people interested in research and how to use it, research in network protocols, research in using graphics over the net, and that kind of thing.

And there were other influences. There were people who began to get involved who couldn't care at all about how it worked. They literally wanted it like they wanted their lights to go on in their office, their telephone to work, or their oil burner to work. So there began to be many other pressures on the net, some from people as you say interested in research and using it, others interested just in having a utility for e-mail and other purposes. So indeed as the community grew, there were many competing interests for attention.

One of the most serious problems that was faced by our group at BBN was the fact that if you build a system primarily for R&D;, and then all of a sudden people really want to use it every day, 24 hours a day, you have a problem. Because how do you change it? How do you shut it down? How do you test it? How do you make a modification that might break it?

That became, for us, one of the terrible problems of success. I call it the success disaster. You're building this thing, which you view as a plaything, which you can change whenever you feel like it, and all of a sudden there are people out there going crazy when it stops working for half an hour. And that had never been planned for.

If ever there was a pressure of change, it was that pressure. The people who wanted to experiment with how to use it, they tended to be our buddies. The people who didn't care at all about it and wanted it to run, they were a problem.

Gates: And overseeing this growth to this day, there is the Internet Society and the Internet Engineering Task Force. Were you on either of those?

Heart: No, I was not. Overseeing is a strong word. I'm not sure there's been a similar example in our lifetime of a thing that is so little run by anybody. It's really unusual in that sense. It's kind of amazing that it keeps working under those circumstances. The Internet Society had very little operational impact. The Internet Engineering Task Force, and various boards there, they did have an important influence. They tended to focus more on the use-end things -- protocols and name spaces and domains and things that affected the host ends -- more than they did on the engineering of the net itself.

Jon Postel and Management of the Root Zone

Gates: Another clash that emerged recently was over the managing of the root zone. I don't know if you read about Jon Postel . . .

Heart: Yes, the "Postel Day" (laughs).

Gates: Yes, that little test he did, which upset some people. [See feature story in this issue: Culture Clash in the Root Zone] My reading of it is that it is a kind of culture clash between the old Internet technical guys and the new commercial interests.

Heart: Well, that's probably true. I have not been directly involved in any of that, but I think there you have it right there. There probably is a little culture clash there.

As the Net has become so incredibly popular and commercial interests have begun to be a major force, there's a culture clash there. But, you know, I look at that as quite trivial in the following sense: The Internet is still a tiny baby. We're talking about something that only started in 1969; it's what, 30 years old? That's hardly anything. It took that long to get the first jet airliner after the Wright Brothers. People who look at the Net now with alarm over this or that problem simply are not thinking through how really young it is. There'll be lots of problems. It will break. There'll be people who do the wrong thing. There won't be enough bandwidth. And then all of those problems will be fixed, and it will be fine again. I think people don't have the image that this is a tiny baby still, and has a lot of growing up to do.

Gates: You knew Jon Postel. Would you be happier to have the coordinating and management of the root zone in his hands than in the hands of some new private company run by some MBA? Or do you think the time has to come when the Internet must go out of the hands of technical people?

Heart: I don't think there really is a choice. I would prefer is to try very, very hard to get very good people involved in any activity that manages it. But whether you can stand having individuals with no responsibility to the rest of the world doing these things, that's probably not appropriate any more. So I'd try for a middle ground. I'd try for something that was more representative of the broader interests, and I'd try to get very good people involved in it. Like Postel, or others.

But there clearly has to be change as the thing gets to be a worldwide utility. It can't stay a toy of people.

On CyberCulture

Gates: What do you think of the cyberculture that has developed from your baby?

Heart: I think it's quite amazing. There's probably nobody who was involved in the early '70s who had any clue what it would become. A number of people, including myself, knew it was moderately important. We knew we were doing something that was going to have an impact, certainly in science. But I don't think there was a single soul who was able to predict this explosion that took place.

It still is doing miraculous and wonderful things for scientific research. I don't know if you follow the Genome Project. The fact that it is doing all these other things hasn't subtracted from the benefit it is accruing to the people who are using it for science. Fortunately it is big enough for both of those.

So I think it's quite amazing. I'm constantly amazed at it, and I'm delighted in my retirement to be able to use it and have it as a toy.

Gates: Do you use the Net a lot?

Heart: Sure, we use it a fair amount. As it happens, BBN was reasonably nice to people like myself and we still have access to their computers. I use it through BBN or GTE, or whatever it is these days.

Gates: Any ideas about what the future holds for cyberspace?

Heart: I think it is very, very young. I think it is going to continue to become a critical utility. I think it is going to continue to grow and become completely pervasive. I think it is a very big deal. I think it is one of the biggest deals of the century. It isn't the only big deal. Atomic energy was a big deal. Going to the moon was a big deal. But it is one of the big deals; and it is wonderful to have participated in one of the big deals.

Where Credit is Due

Gates: There are no Nobels in engineering. Do you feel adequate credit has been given to those who built the Net, including yourself?

Heart: I think that in any project of this size I often make the joke that failure is an orphan and success has many fathers. I think that various different groups all think they played very large roles in any given success and they have different views about the other people that were involved. I think that each of the groups that were involved in that probably wish they'd gotten more points out of it. That probably includes how I feel about the BBN group.

On the other hand I think that everybody that was involved had a wonderful time, the time of their lives. So it's not so terrible that credit isn't distributed perfectly fairly.

Gates: Some enormous fortunes have been made from the Internet. BBN failed to cash in as well as they might have.

Heart: I was never all that greedy. I did fine. I had a very good job at BBN, and BBN had a stock option plan. The fact that the company didn't become a Microsoft didn't mean that nobody had a reasonable reward structure there. I think people did okay, and they had good jobs. If they'd been more greedy people, they'd probably would have gone and done something else I guess. Isn't that really right? I mean don't people who really feel that way go and do something about it?

Retirement

Gates: What you are doing in your retirement and how you are enjoying it?

Heart: I think that how you like your retirement is partly a function of how it comes about. If you end up leaving unhappy or feeling that something bad has happened, you probably don't like it as much. I ended up on a high. I had a wonderful few years there. My piece of the company had made a lot of money. I had people that I was really fond of there. They were nice to me when I left. So because of all that, I came out of there pretty happy.

I've been following a few spin-offs; I've been putting a little energy into BBN/GTE; and I just got back from almost two weeks of skiing in Utah. I'm having a fine retirement, and I'm not doing anything in particular.

Gates: Do you wish you were still involved?

Heart: You know, I don't know how to answer. When I retired two and a half years ago it was the exactly perfect thing to do because BBN had a new chief executive and the chief executive had the audacity to want to run the company. While one can't blame him for that, I was kind of there with the second steering wheel and mine not fully connected to the axle. So it was a perfect time to leave, and I was 65 anyway. I don't have the slightest regret.

I had two enormous rockets that I ended up riding. Being at Whirlwind in the early 50s was also an extraordinary experience. So having that experience twice, having it once at Whirlwind and then again with the ARPANET, I don't think I have any complaints.


__________________________________________________


Send your e-mail feedback to
Dominic Gates
Selected comments may be published on our "E-mail to the Editor" page.




Previous StoryNext StoryTable of ContentsBack to Top
About PreText


    © 1998, PreText, Inc.

    PreText Magazine is produced by PreText, Inc., a digital media studio specializing in content, design, and technology for the Web. For more information visit the PreText Web site.