Skip to main content

The Next Big Thing

Laying claim to being a father of the Internet can be a dicey proposition-just ask Al Gore. But it is one title that Dr. Larry Roberts can claim with some conviction. In 1966, Roberts was ARPA's chief scientist when he began work on the Advanced ResearchProjects Agency Network (ARPANET), part of the Defense Department. The project specifically involved helping turn Leonard Kleinrock's packet-switching theory into a working network-which played a major part in the evolution of the Internet.Since those days, Dr. Roberts has continued to advance data-networking technology, whether with Telenet in the '70s, NetExpress in the '80s or ATM Systems in the '90s. Today, he is CTO and chairman of Caspian Networks, an Internet infrastructure company.He lives in Silicon Valley in northern California and is, in a strange way, master of the valley. He recently took time out to discuss-with
Broadcasting and Cable Associate Editor Ken Kerschbaumer-the Internet and where streaming media fits in.

As someone who has been involved with the Internet from the very beginning, what is your take on the whole direction of the Internet movement in the past year or so?

I think that e-commerce, of course, has been a lot of the focus, which is fairly misleading. E-commerce is actually only a small part of the Internet business. I think what has happened in the last year, in the last two years really, is something that's really remarkable. But what people haven't noticed as much as e-commerce is that the industry has taken off in internal business activities. For example, companies getting material from each other-parts ordering and parts information-all of the business traffic has moved to the Internet, pretty much.

As for the traffic on the Internet, it's been doubling each year or maybe a little faster until last year, when it went up by four times. This year, so far, it has grown by eight times, and it will continue to grow for the next four years or so. But around 2005, I think Internet traffic will have to start tapering off, because we will have reached the amount of money that is spent on telecommunications today. I don't think people will spend more of the gross domestic product on telecommunications than they do today.

In any case, they're rapidly shifting over from the telephone and other things, and spending money other ways. And the carriers who are providing Internet access are really struggling to find technology to install, because they can't grow fast enough.

Where do you see the Internet fitting-as a medium-compared to television and radio?

I think you'll immediately start seeing streaming audio and video take off in business activities, because more and more businesses are being connected at very high bandwidth. There's a lot of high-bandwidth stuff going on in business, and almost all the business-related connections are at least a T-1 connection.

I also believe radio, through the Internet for the home, is going to take off very shortly. The bandwidth for the average user at 56 kilobits is sufficient for radio. The question in my mind is, when are we going to start seeing Internet radios-devices that are made for that? I suspect that, within a year or two, we'll start seeing actual devices that are just purely Internet radios.

Then what are the implications for traditional radio stations and their businesses?

One of the implications is that I will be able to select whatever I want to hear, say the news coverage in Florida of the election, and I'll just select that. I won't have to listen to the rest of the stuff. But one of the questions that introduces is: How do programmers get paid? Ads don't work as well, and I think subscriptions are annoying to people, because they like to jump around and do whatever they want to do. So my guess is that there is going to be a very easy mechanism for me to pay a flat bill each month to my ISP, and whoever I'm talking to gets a piece of that. It won't cost very much because there are millions of listeners.

The result is Internet radio is going to be very attractive for people who want to listen to exactly what they want to-the kind of music they like, or the news when they want, or the traffic when they want.

The real issues are twofold. One is, how do subscribers pay, which I think is easily solved. And the other is, how do your subscribers select content easily. When you're in your car or when you're doing other things, it's not easy to select a Web site. In your car, I suspect it's got to be speech interpretation. And I think it could be the same in the home. The radio will listen to what you say, because you don't want to put your hands and eyes on something.

What will be the impact of wireless wideband?

Wideband wireless is going to be in place next year. The technology that I've seen will virtually eliminate fading, because it uses two antennas rather than one. So when one fades, the other doesn't. The effect of that is actually dramatic, because it means that you can go almost anywhere and not have any substantial fading.

What about advertising? Will people accept commercials as part of the experience?

Well, I think that, once subscribers start paying for subscriptions, then they don't have to experience the commercials, and I think that is what they'll choose to do where they can. But you will have people who want to hear about various products, and the Internet is actually a pretty good way to do that. The challenge is how do programmers insert advertising, because people will be able to fast-forward past the commercials-there's absolutely no reason why they have to listen to the stream.

If you have TiVo, you've already seen how that works. You just space over them. I think that the commercial people have to take a different tack. They'll have to find out what consumers want, and make better guesses as to who consumers are and what they want. But I don't know if people will accept it.

You've been credited with being one of the fathers of the Internet. How different is where we are today from your vision?

Well, I didn't expect we'd necessarily get into the areas that we've gone into. The primary emphasis when I started it was to get information from any computer to anybody, anywhere, and we do have that. But while that expands the knowledge base, it doesn't change commerce, and it doesn't change entertainment. And now what has happened is we've gone through e-commerce, which had never worked before, and we didn't really expect it would work. It has worked pretty well, even though too many people started companies.

When you first started getting involved with the Internet, and you started to realize some of its potential, when did it become apparent to you that this would become a medium that could be as widely accepted as television and radio?

In the 1970s, I don't think I ever thought that it would get this big, because we didn't have personal computers. It was all mainframes, and the potential was far more limited. But, by the early 1980s, I think it was pretty clear that it was going to grow this big. In fact, in 1981, I made a prediction that voice would convert in 2001 to the Internet, which, in fact, I think it will.

So, I was pretty accurate in predicting that communications would transfer over to the Internet. As for television, because it had the highest bandwidth, I was projecting it to hit the Internet probably in 2010. Now I think it will be 2005. Things are changing much faster.

There has been a lot of heartache with start-up companies that came along and seemed to see the Internet as a way to compete with traditional media companies and programmers. And there were others who said companies couldn't compete unless they made the content interactive to take full advantage of what the Internet has to offer. How important is it to have content that's interactive, as opposed to just streaming the equivalent of a video sequence?

I think what consumers want out of streaming audio and video is the same experience as we have today, except selectable. In other words, I would like to be able to go to a Web site that has the news I want at the moment, the information I want or the program I want to see. I don't want to be controlled by the programmer's timing.

I think the other part of the picture is that there will be a new activity in interactivity where people want to interact with other people. They'll do that in things like aviator games, where they really can start interacting with other people.

How important is standardization in streaming media?

Well, I don't think that's a big problem, because you can always download the software. In almost any device we build in the future, you'll be able to do that.

Are there any other areas where you do see some standardization problems?

One of the problems I see is that the Internet network isn't supporting streaming media very well today. And it won't, unless the person originating the stream provides more information to the Internet about what kind of media they're sending and the requirements for quality of service. I think that companies like mine are going to build equipment that can support quality service over the Internet so that we can support video and voice far better.

But, again, the server sending the content has to say something more-in terms of the protocols to the Internet-like what speed it needs or what delay variance it wants, or it won't be handled properly. The protocol that is used today-UDP-is about the worst at giving information of any I've ever seen. It just doesn't give any information about the speed and the quality of the service required. Of course, when people built the Internet, they didn't think there was going to be any quality of service.

How are companies attacking this quality-of-service issue?

There are many people attacking it at the edge, but we're probably the only ones really attacking it at the core-to really fix it overall. It will get fixed in the next couple of years, because I don't think it's a big mystery as to how to undertake fixing it. But the real issue with standards is at least getting something at the beginning of a transmission, some command that tells the Internet what you want.

So the command will sort of tell the Internet that it needs this much speed to deliver the content?

Right, this speed and this delay variance. Those are the two primary things.

I guess a similar quality-of-service problem comes up when cable operators offer cable modems with access to the Internet at speeds up to 100 times faster than 56k, with speeds relying on traffic. And on the DSL side, the problem seems to be just meeting demand for rollout. How do you see those services rolling out over the next couple of years, and how will operators improve their services?

There are three services really, because wideband wireless will start competing, very heavily. My guess is that cable will continue leading for about two or three more years, and then both wireless and DSL will surpass it in actual bits moved.

The problem facing cable with cable modems is they have to rearrange their bandwidth allocation, because, as people start using more and more bandwidth, they're going to overload the 30 megabits that are available. They're also going to have to offer fewer channels and more Internet bandwidth, because they'd really like to use several hundred megabits for the Internet so they can handle more people. But they're going to have to change their equipment a lot, and that's one of the problems that cable is going to have.

Another problem they face is most people who use a cable modem see great service when they start, and then it gets worse and worse. But DSL won't have that problem. In fact, as the business improves, they'll put fiber closer and closer to your house so it will be easier and easier. That's why I think DSL will overtake cable modems in time very easily, in terms of bits moved. And people will find it to be a more consistent service that isn't always changing as the cable operator forgets to invest.

Fixed wireless has a huge upside potential, because it basically can give you 10 megabits or wide bandwidths, in every cell, but the cells can be as small as you want.

How does fixed wireless work?

It uses a new technology that has less fading because it uses multiple antennas at the transmitter and multiple receive antennas-two and two, preferably. And then you just have a packet-delivery system, which basically sends and receives packets of information.

The technology has been available for a very long time. But, the solution to the radio problems didn't come up until recently. So, now we can really do 10 megabits or more in a cell, and at that point, you go to a radio store and pick up your device, take it home and start using it. That dramatically simplifies the installation process. And so, I think that's why it can outstrip the others for a period of time, until they can get DSL to your house.

So, when do you see that service rolling out?

Everybody's saying next year.

In the online-entertainment area, there have been a number of non-starters and companies that have gone out of business. Do you think these companies were just ahead of the technology curve? Did the people who were investing in them not understand the technical side?

I don't think the technology is there yet. For radio delivery, it's just barely there. And for TV, it's too early for people to get good quality. What's happening is people are always going to be experimenting at the fringe, and if something happens, they'll find some users who can afford it or have wide enough bandwidth access.

I think you'll always see that kind of activity at the start of an area. But we're going to see such a fast growth in the Internet bandwidth, at about eight times per year. The average delivery to a person is under a kilobit today in the U.S., but if that's 8 kb next year and 64 kb the next year, then, in two years, we have tremendous capability for any kind of high-quality radio. And then, by 2005, we have very good-quality TV. The kind of thing we're doing at Caspian Networks will support the quality of service, and the streaming and its efficiency will be better. Secondly, the devices the services will then go on will progress a bit, and there will be devices that are just Internet radios or Internet TVs. And that will change the market, dramatically.