Prick up your ears -- it's Silicon Valley Radio. Larry Roberts is the proud father of the Internet. Despite a humble infancy and an awkward adolescence, the Net, Roberts believes, has a promising future. However, he insists that without parental intervention, an impending meltdown will arrest the Net's development and possibly block its passage to maturity.
| back to interview |
Larry, what were your initial thoughts when creating this computer network?. Larry Roberts: I got the impression that this was really going to be the next step of evolution for the computer field. The concept was that this was a major step for mankind also, because evolution has gone from language to the printing press and now we're where we've got to have computers do the same thing -- they've got to be able to interact and communicate. It is just like people talking or the printing press -- you've got to be able to have the extensive communication that expands mankind's ability to operate.
So that was the concept. I worked for probably about seven or eight years to get the technology together and do the first experiments with two computers, which we did in '64 or '65. And then we built the ARPANET in '69, which was the major experiment to build a real network that would be functional at the speeds we needed.
The network was immediately successful in terms of the communications technology. That worked very well, more so than we had even imagined. The packet switching was far more effective than circuit switching by 15 to 1 or something, in terms of economics. That's why nobody can hurt the network. It's too reliable in terms of too many paths to get anywhere.
But the concept at the very beginning, when we first started, was to interlink all of the software and data so that we could get at anything anywhere and really be able to operate anything anywhere, and have that all available. It wasn't so much for people to communicate -- e-mail, voice, video, that sort of thing -- that was not the original thought. That happened within two years; by 1971 we were well into heavy e-mail.
SVR: You were at the forefront of e-mail? Roberts: Right. I wrote the first one and everybody copied the technique. And, in fact, the format hasn't changed in the last 30 years. SVR: Just out of curiosity, was this something you patented? Roberts: No, I was working for the government and the whole ARPANET project was done with government money, so there were no patents involved at all. SVR: You had mentioned that security might be the major issue holding back the Internet's development. What's your approach? Roberts: The approach should be to use the public key systems that have been developed. They are well developed, they are extremely powerful, and, given long enough keys and so on you can protect against virtually anybody. They can protect signature authority, they can protect that I am talking to you and I know that it's you. They can protect cash. They can protect access.
The issue of being able to do that in a way that really solves the problem of ensuring that everybody who is authorized is able to get in means that people will have to carry a card with that key on it.
SVR: When you were creating the Internet, what did you see as some worst-case scenarios that might come from having this new technology available to the public? Roberts: What we were worrying about at the time were technical problems, such as the Net going into lockup and crashing. In about '78 we had a lockup where the routing that we were using was making a decision on the which route to take on the basis of the traffic. It switched too much traffic when it made that decision. It decided that the northeast route was too full by 2 percent or something, so it switched all the traffic to the southern route. And that brought the whole Net down.
But I guess what you were asking about were the human problems.
We certainly did think about security because we knew that would be an issue, so we encouraged the development of the crypto field -- the things that have led to RSA and the public key system and so on.
We didn't incorporate a lot of that at that time because, in fact, if you look at the history of computing and telecommunications, nobody has been willing to pay for security. So every time any manufacturer says, "I'll offer it at an extra price," nobody buys it. And so it has been very difficult to get anybody to put the work into developing it. I think we are at the point where people will pay something for it. Not a lot still, but enough that it will solve some of those problems.
SVR: Who will bring the bigger pipes, the wider bandwidth to the home? Will it be the telephone companies or the cable companies? Who is going to win that war? Roberts: I think that fight is going to be a very interesting fight, and there are actually three participants. There are the wireless companies, there are the cable companies, and there are the telephone companies.
The cable people have a real problem. They are going to have a big win at the moment with simple cable modems; there will be a pretty good penetration over the next few years of that because cable modems are a big step up for the user. You can get to 10 megabytes at the home pretty easily. It is very hard to get the other direction, the back direction. But the forward direction for data flow into the house is good. And of course you can always use the telephone for the back direction if the cable is a one-way cable altogether. So that will be part of it.
That doesn't really allow them to win the battle because until they get a good back channel -- that is, wide band -- they haven't really provided the home with the full service for video conferencing, for voice, for serving. I think a lot of homes will be doing some serving too; they'll have their own thing that they are selling, their own thing that they are doing. So I think that it is very important to have a fuller duplex service, and the cables are going have a hard time doing that.
The telephone companies have a much easier time doing it and, in fact, have a much greater infrastructure and much more money. When they get their act together and the right technology -- which will take about two more years to really have the right technology with which to get to the house -- it will be a combination of fiber to the neighborhood, maybe not to the curb but to the near neighborhood, then wire to the house -- that will be a pretty effective way to get to the house.
Fiber will need to go to the house eventually, but it will take quite a while to get to that. The whole process of installing fiber in the neighborhoods is going to be a major undertaking for the telephone companies. It will take them 5 to 10 years overall. But it will be quite cheap to get to each house and they will be able to deliver, initially, probably 10 megabytes and later on considerably more full duplex to the house -- which is what people really can use.
I believe that in the long run the telephone companies will do much better in that battle. In the short run, the cable companies will probably do reasonably well. And that means that they will have some incentive to go and install more, better two-way cables. And they may, in fact, do very well.
The wireless people will come in and win part of it as well because, particularly at first, low usage is very attractive. But at any significant volume, people will want to get away from that because there are not enough airwaves available. The FCC does appear to be open to giving enough spectrum, though, to make that a viable option.
There is the whole other issue of what the bandwidth and the structure of the whole Internet are, and how that will grow. Today we are in a position where, if nothing is done, we are going to have a meltdown in two years or so.
SVR: Why will we have a meltdown? Roberts: Because the TCP, which is the protocol we are using for flow control, is an Internet protocol that is sitting there banging back and forth trying to control the rate at which each person sends so that they don't exceed the network. As the volume on the network grows, the speeds of everything have to go up, so we are raising the speed of the entire network and the trunks and the access lines that everybody is working from.
As that goes up and I start operating from each company and other places at 100-megabyte speeds or greater, then that oscillation, which takes the same amount of time because of the speed of light -- for example, a server in New York talking to me in California always takes 20 milliseconds to get to me no matter what the network does, because of the speed of light -- that oscillation is 40 milliseconds at minimum, plus the operating systems delays and everything else.
The result is that you are losing more and more data in the oscillation peaks. And every time you lose it you have to retransmit it, and it gets worse and worse. Efficiency goes down, overall, as the speed of the network goes up. TCP just has to change.
SVR: For the average user, who doesn't understand packet flow and why packets are lost, how does that person go about diagnosing what is wrong? Roberts: At the moment there is not much of anything they can do about it because it isn't their computer; it is the network. Unfortunately, that is a bad scenario because if you lose a little bit more everybody retransmits more and the volume goes up rather than remaining stable. So it gets worse and worse as things get worse.
That is a problem which we have got to change, in the way I've talked about, by getting flow control in the network, real flow control at the electronic level.
The problem with real-time audio or voice or video or anything like that is, once you lose it you can't retransmit it. It is too late. Now if you are playing audio slowly, you could retransmit it and it would just take longer if you are just getting an audio file. But if you are playing it in real time, you have a real problem because it is gone once the network loses it. But the network won't lose anything under the 1998 scenario that I am painting.
SVR: That sounds very optimistic. Roberts: I think it is going to happen pretty smoothly. It will take, in fact, a lot of system changes on everybody's part because what is going to have to happen is that the underlying layers will have to be changed out. And then the routers are going to have to be changed out. And then the equipment is going to have make one evolutionary roll. But that is not too difficult. I mean, that is a two- or three-year process. And it is growing so fast that that is probably quite reasonable.
Now, at that point in time, if I'm offering a Web site which has, say, a Web page that people are interested in, and I want to offer a quality service, what I would do is say, "Okay, this Web page has a megabyte of data and I want a two-second response time for my people because that is where people start getting uninterested is two seconds." I want to be able to get it to them in two seconds so I want the data rate associated with that.
And if it isn't there, if you ask the network "What data rate am I getting to this user?" -- and you will now know what data rate you are getting -- and I am only able to get a megabyte per second, or 28 kilobytes or some amount that that user can support, then you will choose a Web page or a reduction of the photographs or something else that fits, that gets there in two seconds. You will be able to make intelligent choices so that you get the quality of service that you are asking for. For example, with the audio, if you had two compression rates, you could offer to get to him with the high compression if that is all his line would take.
SVR: People consider you the father of the Internet, the elder statesman. Do you like that title? Roberts: Well, yes. I conceived of the network and went out and designed a large fraction of it, although I got everybody in the R&D community to work with me to help build it. And I went through and made it happen with Congress and with government, and I went through the process of building it. So I really did make it happen many years earlier than it would have happened any other way.
It was a huge fight because the telephone industry was convinced that this was crazy and stupid and it would never work -- that packet switching was the worst thing ever to come along and was just going to blow up in our face.
SVR: So it took a lot of self-determination to see it through? Roberts: Yes. It was like any new thing that somebody is trying to make happen. We had to fight all our critics who said this was ridiculous. In fact, within my own community, within the computer community, nobody wanted to do it; and by saying nobody, a lot of the universities and research labs didn't want anything to do with it because they wanted the computer all to themselves. They didn't want anybody else to get on it. SVR: Why were you so convinced this was a good idea? Roberts: Because I knew we couldn't grow a community, our information and so on, by having every computer separate. If every computer had stayed separate and all we could have done was transfer tapes around between each other or something, we would never have been able to grow the information base or the literature base or the knowledge base or the use of the software.
When somebody does something and has a big database for every piece of software, you want to be able to get at it from all over the world.
I won an Erickson Award in '81. The Erickson Award is the equivalent of a Nobel Prize, because Nobel doesn't give anything in this field. The Swedish National Academy gives out the award in communications every three years with money from Erickson instead of Nobel.
In any case, that award was based on the work I'd done in 1969. And I asked why it took so long for them to get around to it (if it was supposed to be based on recent accomplishments). The reason was that from their point of view they had just heard of it and understood it in the last few years and it just became a real thing in Sweden.
SVR: What are your interests outside of work? Roberts: My primary interest is life extension and supplements to extend age and improve performance. SVR: Any ideas that you hadn't known prior to your exploration of this topic? Roberts: Yes. And I have published and given speeches on the subject quite a bit. The thing I have concluded is that evolution really has tried to get rid of us after childbearing age; it doesn't really need us around, given the old caveman scenario, at that point. And therefore we have built-in mechanisms that kill us off, that don't need to be there. SVR: Like what? Roberts: Our dopamine declines from age 45 to age 100, and then we are dead. It kills you automatically. And it is caused by a build-up of an enzyme called MILB in your brain which exists for no reason whatsoever except to kill you. It is easy to fix with a drug called Deprenyl. You just take the drug and that goes away and you don't get killed that way anymore. SVR: I've never heard of that. Roberts: Well, there are lots of those things. If you look at the human body you will find about 15 things that I've identified that are programmed to kill you. SVR: Aside from the supplements, what are some other things that contribute to longevity? Roberts: The thing that has really affected me is that there are some things which affect your mental capability. People's mental capability seriously declines with age. I found that I wasn't doing as well as I was in my 20s and 30s. I wasn't able to do programming and technical work -- which requires a lot of memory and creativity -- anywhere near as well. I could do management and decisionmaking, but I didn't have the same ability to do programming and creative things that take a lot of intensive concentration. I started taking a number of different supplements to restore certain things and fix the blood supply to the brain.
It has totally changed my capability to perform on that level. I am now the leader in the scientific community in terms of this low-control activity, and I've been able to program and do designs that I couldn't have possibly done five years ago.
. . . . . . . . . . . . . . . . . . . . . . . .
Every item on this site is copyright © Web Networks, Inc.
Disclaimer: Web Networks, Inc. is not responsible for anything found on sites other than those hosted by Web Networks, Inc.