Still Netting after all these years

Still Netting after all these years
By Nick Wingfield

Staff Writer, CNET NEWS.COM

One would expect the man known as the "father of the Internet" to resemble an aged Fred McMurray--retired, nostalgic, and wrapped in a cardigan. But nearly 30 years after joining the team of engineers that erected ARPANET, the Department of Defense-sponsored predecessor to the Internet, Vinton G. Cerf is still very much on the job.

Now responsible for overseeing MCI's Internet backbone, the largest carrier of Net traffic in the world, the 50-something Cerf is as excited about the present and future of the global network as he is about the past. And what a past it is.

Cerf joined the original ARPANET team in 1968 as a programmer, helping to cobble together the first nodes of the network a year later. At the time, ARPANET consisted of a handful of room-sized host computers on university campuses connected for the purpose of sharing computing resources. Within a few years, email became a hugely popular application within the small society of ARPANET users, helping to fuel the growth of the network.

In 1974, after picking up a Ph.D. in computer science from UCLA, Cerf and several colleagues produced what would become a vital, albeit clumsily named technology for the Internet, the transmission control protocol/Internet protocol (TCP/IP). As it caught on over the next decade, TCP/IP provided a common dialect that allowed networks of all stripes--with names like CSNET, BITNET, and NSFNET--to be linked into a common internetwork, or, as we know it, Internet. Today, that same protocol underlies every Internet application, from email to the Web to audio streaming.

It's not hard to see how Cerf's celebrity has risen above that of his ARPANET colleagues. Charming, cultured, and a bit of a dandy, he defies the myth of the stereotypical nerd engineer. From the time he was in high school, Cerf was determined to make an impression by dressing in coat and tie and carrying a briefcase. "I wasn't so interested in differentiating myself from my parents, but I wanted to differentiate myself from the rest of my friends just to sort of stick out," he said.

In conversation, he speaks with equal delight about wine (Montrachet white burgundies are good, older Bordeaux are risky) and the cochlear implant that almost completely restored his wife's hearing recently after a lifetime of deafness. If the Internet can possibly be represented by a single face, it is Cerf's friendly bearded and balding visage. (People magazine placed him on its 1994 25 most intriguing people list.)

There's also an air of diplomacy to Cerf, a quality that has aided him throughout years of contentious debates with engineers over the course of Internet standards. At a 1992 meeting of the Internet Engineering Task Force, as a heated argument over the future of Internet protocols turned angry, Cerf diffused the controversy by stripping down to a T-shirt bearing the slogan "IP on Everything." Even as the Internet becomes a high stakes arena for businesses, Cerf believes cooperation is critical for keeping the Internet in order.

These days, Cerf seems flattered but slightly uncomfortable with the idea that he is the father of anything but two grown sons. MCI, though, clearly relishes his legend status, regularly trotting him out to speak and schmooze his dazzled fans. In 1994, the telecommunications company jumped at the opportunity to rehire him (Cerf developed MCI Mail for the company in the early 80s), as if it were a record company presented with the chance to hire Elvis as a vice president.

Cerf remains emotionally connected to his ARPANET days, exchanging email with former colleagues over the network they created. But he displays little nostalgia for an era when Net access was limited to a few dozen engineers eager to share their computers. Cerf is a supporter of Ethernet inventor Bob Metcalfe's law: The value of a network increases in direct proportion to the number of people connected to it.

NEWS.COM talked with Cerf at our San Francisco headquarters about everything from Internet standards and MCI's pending merger with British Telecom to the now famous T-shirt incident.

NEWS.COM: If you could rebuild the Net from scratch, what would you do differently?
Cerf: One thing I'd definitely do is pick a bigger address space than 32 bits. I have a rationale for having ended up with that number, but looking at it in 1997, it doesn't hold too much water. I can only be forgiven that this was a decision made 20 years ago when I didn't have adequate foresight to realize what an explosion [the Internet] would be.

I would probably go in and integrate security into the system more fully than I did. I don't mean to claim all this credit; many other people did this. Collectively, we didn't do some things. But remember the time period in place where this work was being done: It was the 1970s, the Vietnam war was going on, and most of the work was being done by graduate students in universities. If we had gone in with any kind of military-sounding crypto, even given that we were allowed to do that (which we would not have been because it was all classified anyway), I think we wouldn't have gotten very far.

NEXT: MCI, consumers, and Net telephony

MCI, consumers, and Net telephony

MCI hasn't been as aggressive in going after the consumer Internet market as AT&T has been. Why?
We have tended to focus on the business community for two reasons. The business community is better outfitted statistically to take advantage of Internet products and services than the consumer market. This isn't to dismiss the consumer market by any means--it's clearly big and growing. The current stats are that something like 35 to 40 percent of American households have PCs and maybe a third to more than that have modems. That's a significant population right there.

We are preparing to handle more of the consumer community as the equipment and software penetrates the market. Plainly, other companies have shown that there does exist a market and there clearly exists demand. Whether the demand is sustainable under reasonable pricing formulas is still an interesting question to answer. We've already been through the tragedy of that situation. Others have gone through that, AOL in particular. So we are conscious of how big and volatile this market can be, and we want to serve them properly and not go in and advertise things that aren't sustainable.

I think some level of service is sustainable at $19.95, but I don't think it's unlimited level of service, or alternatively, it may be unlimited access to nonpremium services. I'm hoping two factors will enter into this to make things better for our consumers. One of them is simply engineering to scale.

The second thing I hope will happen is that people will understand--as they have with the telephone business--that paying for what you use is often an attractive arrangement. If it's really flat rate for everyone and the ratio of the heaviest user to the lightest use is very big, the lightest user is going to say at some point, "Wait a minute, I pay $19.95 a month and use it 3 hours a month. He pays $19.95 a month and uses it 500 hours a month. I think I must be paying for his use. This doesn't sound right." At some point, the market will rationalize, and when that happens you'll see a much more reasonable pricing structure that's self-sustaining. Let's face it, we don't want the Net to go away, but it has to sustain itself as a business or it will go away.

[MCI CEO] Bert Roberts says he doesn't think Internet telephones will be much of a challenge to regular telephones. Do you agree?
That's right. First of all, the total capacity of the Internet today, as large as we like to think it is, is quite small compared to the total telephone network on a global scale. So my back-of-the-envelope estimate, which is not very scientific, is that if we took every bit of Internet capacity that the world has now and turned it over to Internet telephony, we might handle three percent of the world's load. And that's probably generous.

On the other hand, as time goes on we will be able to support more and more of those kinds of applications--not just telephony, but other real-time kinds of services. Multiparty gaming, for example, is becoming a very popular kind of activity that our partner, British Telecom, is introducing. So we should be able to handle more of it, but there are circumstances when being able to do real-time service through the Internet is just what you need. In a hotel room, I have one telephone line. And in the home, I have one telephone line. I'm on the Net and I want to buy something, but I want to talk to customer service. Do I want to hang up the phone and kill my Internet connection to talk to the customer service guy who then wants to say, "Now, if you look at the page on the screen..."

What do you think of the regional Bell operating companies complaining to the FCC that Internet traffic is clogging up their networks?
We have detected an anomaly in some of their complaints. One of the suppliers said that they had to spend an extra $14 million or something to upgrade their service in order to deal with the longer connect times for Internet usage. At the same time, they were also advertising that people should get on the Net and [offering them] free time, and buy a second phone line, by the way! We said, "Gee, this doesn't seem to match up too well. Maybe your marketing department and your engineering people should talk to each other."

What are some of the things you're doing to the MCI infrastructure to allow it to handle things like streamed audio and video?
We have been increasing the absolute capacity of the network by increasing the speed of the SONET (synchronous optical networks) interfaces that we use. Our current backbone runs at 622 mbps. We are going to lay in an additional ring of OC12 [622-mbps fiber-optic lines] capability during the remainder of 1997 just to augment the total capacity. We are pressing all of the router vendors--but most noticeably Cisco because we use a lot of their equipment--on the packet forwarding rates that we can expect from their newer engines in order to use that high-capacity fiber that we have available.

We're also working together with Intel, Cisco, and BT on quality-of-service measures that we can introduce into the network to distinguish real-time traffic from other traffic. This is still in the laboratory stage. I haven't deployed anything in the production network, but we're seeing some very good results so far in the lab. I would emphasize that if it works in the lab, that still leaves you some distance away from production capability.

Together, we already have something called Concert Internet Plus, which is a big global business Internet backbone that is still under construction. Of course, in the United States, we have the internetMCI backbone that carries Internet traffic domestically with about 200 international connections around the world. When we combine all of this, there's an opportunity to build a common global backbone. I can't tell you how excited I am about that! For one thing, it makes it possible to offer uniform services to our businesses and customers that one would not otherwise be able to offer.

The second thing it does is that the size of the corporation will be such that our ability to put capital to work is significantly increased. I anticipate that our Internet efforts will have their fair share of capital allocations in order to continue to grow and meet all our customer's demands.

NEXT: What we want from the Internet

What we want from the Internet

Can you address this myth that the Internet was created as a Department of Defense military project meant to survive nuclear war?
The first project that ARPA funded in wide area computer networking was the ARPANET. It was often mistakenly given this attribute of nuclear resilience. The fact is it was designed for resource sharing; it was really to solve a problem that ARPA had when it was supporting computer science research. Every single computer science department in the country said, "We have to have the best machine in the world every year or we can't do world-class research." ARPA couldn't afford that every year. So the question was is there any way to connect them together so that everybody could share? That's sort of where the ARPANET came from.

Obviously, operating out of the Defense Department, there also had to be a rationale for spending money on this. The answer was computers should be part of the command-and-control system in the military. ARPANET focused on the resource sharing aspects of that, and a set of protocols were developed. Then came the new applications of packet switching to packet radio and packet satellite, which also had a clear military potential: ships at sea, mobile units on the ground, in addition to fixed units in the terrestrial part of the world that the ARPANET could address.

When Bob Kahn, who pursued this at ARPA, realized that he had different kinds of nets that had to be interconnected, now we had a different problem. It was clearly an Internet problem. By the time I got involved in all this in '73 on the Internet side, it was clear that there was a real mandate to make this technology work for the military. When the Gulf War was executed, Internet technology played a pretty key role in some of the communications, and it worked.

I think of you as part engineer, part diplomat. Do you agree that there's a lot of diplomacy involved in working within standards groups?
Well, I wish that I could do a better job of it than I have in the past. Whenever you try to deal with a large number of people--especially engineers with very strong, almost religious opinions--you need a certain amount of diplomatic skill to make things come together. In today's world, it's not just engineering passion that drives the debate, but economic interest as well. Let's face it: Greed is a very powerful force. It can be used in very positive ways because it potentially drives a very positive economic engine. It calls for yet a higher degree of diplomatic skill than anyone, including me, really possesses.

Now that businesses have a serious economic stake in the Internet, do you think standards are being adhered to well by companies like Netscape and Microsoft?
All of these companies are building many of these products on standards that are developed by the IETF [Internet Engineering Task Force, a volunteer standards body], the World Wide Web Consortium, and others. In some cases, they're industry-level conventions that just become common without any formal approval. What typically happens is that in the course of trying to achieve interoperability, one seeks standardization. But in the course of trying to achieve product differentiation, one seeks to become incompatible, to retain customers because only your software can do something. And there is a true tension there.

I think if we've learned anything as a community in the last ten years, we've learned that customers want choice, customers want interoperability, and customers don't like being locked in. As much as I can understand the Netscapes and the Microsofts of the world wishing to do things that are special to their products to advance the state of the art, I think they will ultimately find that their customers will insist that their products interwork. I think that's been a bedrock for the Internet community, and I don't think that's going to change.