Looking Ahead Vinton Cerf, WorldCom

Consider that at the time, the Internet was a set of four primary networks: ARPANet, two packet radio networks (one in the San Francisco Bay area and one in Ft. Bragg, N.C.), the Atlantic Packet Satellite Network (in the United States, Norway, the United Kingdom, Germany and Italy) and a number of LANs (Ethernets, Proteon rings, etc.). There probably were not many more than 400 computers on the Internet, and the public, as well as most of the academic community, was oblivious to its existence. In fact, the Internet's primary communication protocols, TCP/IP, were not rolled out formally on a mandatory basis until Jan. 1, 1983!

Could anyone have predicted that 20 years later, the Internet would reach more than 500 million users, 150 million serving computers and 300-plus million laptops, PDAs, Internet-enabled mobile phones, picture frames and other appliances? I certainly did not. In fact, because the network was then still sponsored wholly by the U.S. government, it was not even clear that public or commercial access would be granted.

It was not until 1988 that the question of commercial access was posed to the U.S. Federal Networking Council, and permission was granted to interconnect the commercial MCI Mail service with the Internet. The World Wide Web didn't exist, although the seeds of its technology had been planted. PCs were only just emerging from their relatively obscure origins.

It seems fair to say that most 20-year projections based on a '80s perspective would have missed some pretty important ideas developed later in that decade. But the sheer impossibility of making such predictions won't stop me from making some now.

The irony of considering the year 2020 is apparent.

When you visit an optometrist, and she tells you that you have "20:20" vision, she means you have perfect eyesight. In other words, you can see at 20 feet what you are supposed to be able to see at that distance. By contrast, it seems guaranteed that my vision of the Internet in the year 2020 is not likely to be very accurate.

To simplify the task, I'll divide my predictions into near-term, midterm and long-range observations with the usual hedging that the chance for error increases as we move further out in time.

The near term, from 2002 to 2006, will be characterized by continued turmoil in the regulatory frameworks associated with broadband, residential access to the Internet.

Regulations (like mileage) will vary, but for the most part, access will tend to be asymmetric. That is, you can receive at higher rates than you can send. Consider that cable modem subscribers can receive at over 1 megabit per second but send at only 128 to 256 kilobits per second. That means a subscriber can receive very good quality digital video but can't send it. So much for high-quality videoconferencing!

Moreover, because each access medium is essentially a monopoly unto itself,the local exchange carrier owns the twisted pair, the cable company owns the hybrid fiber/coaxial cable, the satellite broadcaster owns the spectrum, the Multipoint Microwave Distribution System (MMDS) operator owns the associated spectrum,there is a very real risk of noncompetitive behavior by the service providers.

There is a theory espoused by some which holds that these various high-speed access service modes compete with each other. This is only sometimes true. In many cases, subscribers can get only one or perhaps two of these services, so actual "intermodal" competition is frequently nonexistent. One way to redress this lack of competition is to open up access to each medium so that any ISP can reach any subscriber by means of each access method. That is essentially what happens with the dial network: any subscriber can reach any ISP.

During the near term, debate will center on open access and competition. I cannot predict the outcome, but subscribers have historically insisted on,and ultimately have won,the ability to choose their service providers. During this period, there will be a slower-than-desired but relentless penetration of high-speed access to the Internet, and this will feed a variety of new applications.

I believe fiber will prove to be the best choice for symmetric, high-speed service to residences. It is already the principal choice for high-capacity service to businesses. Without delving into details, fiber may prove cheaper to install and maintain than other media such as twisted pair, coaxial cable. Moreover, one can achieve very high speeds with relatively inexpensive fiber as long as the signals do not need to propagate very far before they are converted and combined into a composite signal carried on laser-driven, single-mode fiber.

Once high-speed services are available, say at gigabit-per-second rates, possibly using Gigabit Ethernet interfaces, it will not seem all that unusual to download or upload high-resolution video files in seconds. A high-definition, digitally compressed one-hour video is about 2 Gbytes, or 16 Gbits of information. At gigabit-per-second speeds, uploading or downloading such a file takes 16 seconds. Video on demand could be transformed: Instead of a continuous, jitter-sensitive service, it may become a simple, high-speed file transfer that has none of the frailties of synchronous, TDM-style transmission.

Over the next four years, we will also witness significant penetration of radio LANs, notably those based on the IEEE 802.11b and later IEEE 802.11a (otherwise known as Wi-Fi) specifications. There are some serious problems with this unlicensed spectrum. I have already experienced multiple service providers colliding with each other within the 2.4GHz band used by Wi-Fi. One is reminded of the early days of commercial radio before the FCC was created. Interference was the norm.

My near-term outlook for third generation (3G) mobile telephone service is more speculative. This is in part because of the enormous sums that have been spent to acquire rights to the appropriate spectrum. Some companies are so heavily in debt because of this expense that it is unclear whether they can afford the capital needed to build out high-capacity mobile service. We should have a clear indication by 2006 about whether the 2 Mbit-per-second burst rate promise of 3G will materialize.

The other major trend to watch is the need to implement the next generation of Internet Protocol, IP version 6 (IPv6). The Internet Engineering Task Force (IETF) standardized the specifications several years ago, but production implementations have only recently begun to emerge, partially because more interoperability needs to be completed.

Security will remain a key concern. It is worth noting that no system can be made absolutely secure against all forms of attack, so intrusion detection and rapid response, as well as resilience through redundancy, will be key elements of future Internet-based systems.

Finally, emerging computer-based speech technologies will create yet another platform on which complex services can be built. I believe a solid base of applications will emerge during the 2002 to 2006 period. After that, speech understanding, synthesis and speaker identification will be standard tools in the Internet repertoire.

From 2006 to 2010, I believe we will encounter dramatic growth in the number of Internet-enabled (or IP-enabled) devices. These often will be simple sensors or appliances with limited functionality that can be controlled through the Net. Obviously, security and authenticity will be critical.

There are already major initiatives in the Netherlands and Japan to Internet-enable automobiles and the devices and instruments they carry on-board. Scores of sensors will provide realtime information about the operation of devices carried in or built into the automobile. Many of these systems will use the Global Positioning System (GPS) or its European counterpart (when that is built) to locate the automobile. Since this positioning information can be made available to every device on board a vehicle, one can expect that devices will effectively know where they are. Consequently, Internet-based interactions will have the potential to use geo-positioning information to enhance the service.

Putting some of this together into a common scenario, one can imagine driving down the street, speaking on a mobile phone with a speech-capable computer that is also on the Internet. As you ask, "Where is the nearest ATM machine?" your Internet-enabled mobile phone conveys your position information. The computer on the other end of the phone can then look up the nearest ATM in a geo-location indexed database and send a map to your car's navigational display.

Far more elaborate scenarios can be imagined based on the central idea that multiple Internet-enabled/addressable devices can be drawn together "on the fly" to provide a service. The array of potential services simply beggars the imagination. Indeed, one of the more intriguing aspects of the Internet is the accessibility of the network's services at the lowest layers (e.g. IP). Literally anyone or anything can emit or receive IP packets, so users can create new products or services at any layer of the protocol hierarchy. It is in part because of this openness that the floodgates of creativity have opened.

During this period, another development is likely to reach very visible fruition. Since 1998, a project at the Jet Propulsion Laboratory (JPL) has been under way. Termed the Interplanetary Network (sometimes called the Interplanetary Internet), this is an effort to extend the utility of the Internet into the solar system.

This is not science fiction. The JPL effort is aimed in part at designing and standardizing an architecture for deep-space communication that simultaneously allows scientists to remotely access and control their instruments by way of the Earth's Internet and also provides the basis for a deep space backbone network (the Interplanetary Network). As long as each new mission's communication systems conform to the Interplanetary protocols, it should be able to support interactions between Earth-based Internet sources and deep space-based assets. Many of these assets in space will make use of standard Internet protocols locally and shift into Interplanetary mode when necessary. Eventually the deep space assets will be used to link platforms around the solar system, not necessarily relying solely on relaying through planet Earth.

Of course there will also be "in situ" networks on the surface of Mars and possibly other planets and/or satellites. Concerns such as conservation of power and self-organization of radio-based networks will drive these deployments, and some of the techniques developed for such space-based networks will have direct application on Earth.

By 2010, there should be at least one or more satellites in orbit around Mars, and a two-planet Interplanetary network should be in operation.

During this time frame, the debate over high-speed access for homes and offices should be resolved, settling on fiber for fixed access and spread spectrum radio for wireless communications. The latter will use a technique called Code Division Multiple Access, or CDMA, to share common spectrum and reduce interference with other users of the same radio spectrum. A variant of this idea is called Ultra-Wideband (UWB), and exploration of this capability was approved this year by the FCC under limited conditions.

It is highly likely that many, if not most, mobile devices will be Internet-enabled, and that suggests that an unpredictable but rich range of products and services will be developed to take advantage of this fact. The Internet should also complete the transition from IPv4 to IPv6. That is an important metric that can be observed quantitatively, providing at least one measure of the evolutionary direction and rate of the Internet.

Shortly beyond 2010, there will be an estimated 2.2 billion users on the Internet and anywhere from 5 billion to 20 billion connected devices. Plainly, managing a system that is roughly three to 15 times the scale of the current telephone system will be one major challenge, if not the major challenge. IPv6 should be the primary core protocol in use at that time, although it is certain that other new protocols will also be in place.

It seems likely that television, radio, telephony, the Web, newspapers, magazines, books, videos, music and a host of other products and services all will be accessible through the evolved Internet.

Intellectual property protection practices will have been resolved through legislation, court battles and the behavior of users. This is not to say that the debate and litigation will be over, as there will always be debate and litigation.

By this time period, it will seem natural to interact with technology by voice and perhaps by gesture. Indeed, it will seem no more strange than expecting doors to open automatically, faucets to start and stop with the wave of a hand, and toilets to flush themselves in 2002. Most of the devices, appliances and gadgets we use will be programmable and controllable through remote servers.

The two-planet Interplanetary Network will have expanded to several other planets, most likely Venus, Jupiter and Saturn. Some spacecraft may be located at solar libration points to act as parts of the Interplanetary Network backbone, and by this time at least one and possibly more than one interstellar, robotic exploration mission will have been launched.

Of course, it will take at least 50 years or more to know the outcome of such a mission, owing to the very long time it will take to carry communication equipment more than four light years away to the nearest star.