Inventing the Internet again

George Gilder
Forbes ASAP
June 1, 1997
Print ArticleIn the early 1960s, working on America's second-strike capability, Paul Baran conceived the Internet. Now he wants to save the Net itself.

FOR THE FIRST TIME IN HIS LIFE as an engineer, Paul Baran was "scared stiff." That can happen to people who stumble too close to the abyss of 20th-century history and look over the edge. Born in 1926 in a house in a corner of Poland that had been claimed by three different nations during his parents' tenure, brought to America by his family at the age of 2, Baran was a child of European tempests.

But now, in the heady Southern California of the 1950s, the young Hughes Aircraft engineer found himself working in an American crucible. He was a design engineer for the Minuteman missile control system. Unlike the liquid-fueled Titans of the previous era, which required hours of preparation before they could fly, Minuteman could be instantly rocketed into the sky. To the Pentagon this seemed safer. The solid-fueled rockets would not be vulnerable for hours on the ground awaiting fueling. But Baran and his colleagues knew that this would be the most deadly and dangerous military system ever built. One accident and a cloud of missiles was on its way.

Appreciating the risks in the proposed design, Hughes summoned Warren McCullough from MIT as a consultant on human behavior. An expert on command and controland a psychiatrist and brain surgeon to bootMcCullough explained the emerging facts of life. Throughout history, he told the Hughes engineers, the real command of the battle migrated to the men closest to the enemy. The man in the crow's nest, not the officers on the ship's bridge, was in de facto control. What he saw and reported determined the captain's orders. Regardless of nominal chains of command, the real governance of history moved to individual people on the front lines, often frightened or panicked at the time. But in the nuclear age, no such single person, necessarily fallible, could ever be trusted.

Analyzing the technical problems of creating a command- and-control system for Minuteman, Paul Baran found himself abruptly in the crow's nest, stricken by historic terror"scared stiff," as he recalls. It was clear to him that the problem was systemic; it could not be solved by tweaking the command-and-control schemes then being proposed at Hughes.

To explore the problem more broadly, Baran in 1959 left Hughes for RAND, the not-for-profit (the name stands for "R&D") set up after World War II to harbor the systems analysis skills developed during the war. At RAND the formidable strategist Albert Wohlstetter was demonstrating that in a matter of minutes Soviet short-range missiles could take out all U.S. foreign strategic air command bases encircling the Soviet Union. Then the Soviets could say stick 'em updemanding surrender on the basis of the vulnerability of remaining U.S. missiles to superior Soviet forces. In many vivid papers and speeches, Wohlstetter relentlessly presented his argument that U.S. forces faced a "missile gap." The famed Alsop brothers, leading columnists of the day (Stewart was the father of the computer writer), echoed the Wohlstetter claims. John Kennedy listened and made the gap a theme of his 1960 presidential campaign.

Wohlstetter and his colleagues urged that the Pentagon redeploy its strategic forces to the United States and endow them with a second-strike capabilitythat is, to withstand a first strike and retaliate in kind. Greatly reducing the temptation to go first, this posture would escape the dangerous hair-trigger tenterhooks of the early cold war.

A viable second-strike capability, however, assumed that the command, control, and communications systems would remain intact. It was here that Baran fretted. He saw that one nuclear explosion at high altitude would affect the ionosphere for many hours and thus wipe out all long-range, high-frequency radio communications. In addition, one strike at the centralized switching nodes of AT&T would destroy the rest of the control network. The missile system would endure, but it would be deaf and blind.

Plunging deeper into history than Kennedy had, Baran resolved to design a communications system that could survive a nuclear attack and save the second-strike deterrent. He took inspiration from another idea of MIT's McCullougha parallel computer system with adaptive redundancy. Like the human brain, such a system could reconfigure itself to work even after portions were destroyed. But using the noise-prone analog circuits of the time, it was impossible to build the necessary switches. Baran concluded that all the traffic would have to be digital. Moreover, the digital traffic would have to be broken into short message blocks now called "packets," each containing its own routing information, like a DNA molecule, and able to replicate itself correctly whenever a transmission error occurred. With many additions and permutations, his original design is today termed the Internet, and it is shaping the emerging history of the 21st century.


Baran, though, is not satisfied with his creation. Contemplating its vulnerability to terrorism and other attack, he feels pangs of fear that echo his alarm of 40 years before. As more and more of the critical systems of advanced industrial society migrate to the Net, they become susceptible to new forms of sabotage, espionage, hacking, and other mischief. Air traffic controls, train switches, banking transfers, commercial transactions, police investigations, personal information, defense plans, power line controllers, and myriad other crucial functions all can fall victim to cybernecine attack. If the Internet is to fulfill its promise as a new central nervous system for the global economy, its security and reliability problems will have to be addressed.

Seventy-one years old, still with his Ph.D. economist wife Evelyn (their son David is director of information technology at Twentieth Century Fox Home Entertainment), Baran remains in the crow's nest, buffeted by inklings and insights of historic threats and opportunities. In a sense, Baran's current projects merely fulfill the far-reaching logic of his original concept, elaborated at RAND between 1960 and 1962 and published under the title On Distributed Communications in 11 compendious volumes in 1964: a survivable "network of unmanned digital switches implementing a self-learning policy at each node, without need for a central and possibly vulnerable control point, so that overall traffic is effectively routed in a changing environment."

To fulfill this scheme, Baran specified all the critical functions of the Internet: packets with headers for addresses and fields for error detection and packet ordering. He described in detail the autonomous adaptive nodes found in Arpanet IMPs (interface message processors) designed by Bolt, Beranek & Newman (BBN).

Baran also included features only recently and selectively introduced, such as encryption, prioritization, quality of service, and roaming ("provisions to allow each user to 'carry his telephone number' with him"). He described a web of peer nodes each connected to three or more other nodes, and he offered the first of the distributed routing algorithms that have multiplied over time.

Unique to his vision was its grasp of the economics of a network that could handle "the expected exponential growth in the transmission of digital data." Declaring that "it would be possible to build extremely reliable communications networks out of low-cost unreliable links, even links so unreliable as to be unusable in present-type networks," he estimated that the price of the system would be some $60 million per year. That was some 20 to 30 times less than what was being paid by the Department of Defense for their leased communications systems without any of these features. It was two orders of magnitude cheaper than new analog national systems being proposed at the time by each of the three military services.

Thus Baran not only conceived the essential technical features of the Internet, he also prophesied the cliff of costs over which digital technology would take the networking industry. By imagining the compounding effects of Moore's law three years before Moore's own famous prophecy, Baran stressed the key economic drivers that impelled the prevalence of the Web as the universal Net.

The system of communications that Baran attacked in the early 1960s at RAND was the imperial establishment of AT&T. As Baran explains, "While AT&T did have digital transmission under examination, it was in the context of fitting directly into the plant by replacing existing units on a one-for-one basis. A digital repeater unit would replace an analog loading coil. A digital multiplexer would replace an analog channel bankalways a one-for-one conceptual replacement, never a drastic change of basic architecture. I think that AT&T's views on digital networks were most honestly summarized by AT&T's Joern Ostermann after an exasperating session with me: 'First, it can't possibly work, and if it did, damned if we are going to allow the creation of a competitor to ourselves.'"

In 1972 the company sealed its fate by turning down an opportunity to buy the entire Arpanet. As Larry Roberts explained in Where Wizards Stay Up Late, "They finally concluded that the packet technology was incompatible with the AT&T network." So it was and so it still is. The existing phone system remains the chief obstacle to the final triumph of the Net. But the logic of digital communications is inexorable. It will displace all the existing establishments of television and telephony.


These days Baran's vision, however, goes far beyond wireline communications. Baran takes the Internet model and extends it boldly to wireless communications. On June 23,1995, on the occasion of the Marconi Centennial, marking the 100th anniversary of the invention of the radio, Baran gave a momentous keynote speech in Bologna, Italy. In it he demanded a radical reconception of wireless networks.

"The first 100 years of radio," he declared, were marked by a perpetual "scarcity of spectrum....One of the very first questions asked of young Marconi about his nascent technology was whether it would ever be possible to operate more than one transmitter at a time. Marconi's key British patent #7,777 taught the use of resonant tuning to permit multiple transmitter....[Yet] even today, with over 30,000 times more spectrum at our disposal than in Marconi's day, entrepreneurs wishing to implement new services encounter the same perpetual shortage of frequencies."

Focusing on the most desired bands between 300 and 3,000 megahertz (UHF), Baran asserted that when you "tune a spectrum analyzer across a band of UHF frequencies," you discover that "much of the radio band is empty much of the time. This unused spectrum might be available for transmission if we could take measurements and know exactly when and where to send the signal."

As an example, he cited "the many millions of cordless telephones, burglar alarms, wireless house controllers, and other appliances now operating within a minuscule portion of the spectrum and with limited interference to one another. These early units are very low power dumb devices compared to equipment being developed that can change its frequencies and minimize radiated power to better avoid interference to itself and to others.

"In part," he declared, "the frequency shortage is caused by thinking solely in terms of dumb transmitters and dumb receivers. With today's smart electronics, even occupied frequencies could potentially be used."

The chief reason for the apparent shortage of spectrum, he concluded, is regulation of it. Echoing his earlier critique of wireline communications, he declared that "the present regulatory mentality tends to think in terms of a centralized control structure, altogether too reminiscent of the old Soviet economy. As we know today, that particular form of centralized system...ultimately broke down. Emphasis with that structure was on limiting distribution rather than on maximizing the creation of goods and services. Some say that this old highly centralized model of economic control remains alive and well todaynot in Moscow but within our own radio regulatory agencies."

The heart of the problem is the concept of spectrum as public propertyas scarce real estate or a precious natural resource. Spectrum is nothing of the kind. It has been created by a series of brilliant technical innovations, beginning with Marconi and continuing in a steady stream of high technology oscillators and digital signal processors: from mag-netrons and kystrons to varactor multipliers and surface acoustical wave devices, from gallium arsenide and indium phosphide heterojunctions to voltage-controlled oscillators and Gunn or IMPATT diodes. Spectrum is chiefly a product of inventors and entrepreneurs. Americans will rue the day when foreign governments and international organizations begin auctioning and taxing, marshaling and mandating the use of these mostly American technologies.

The real estate model applies chiefly to broadcasters and others using analog modulation schemes in which all interference shows up in the signal. A television signal requires some 50 decibels of signal to noise power, or l00,000-to-1. By contrast, error-corrected digital signals can offer virtually perfect communications at a signal-to-noise ratio well below 10 decibels, or 10,000 times less. Moreover, new digital systems can divide and subdivide the spectrum space into cells and differentiate calls by spread-spectrum codes or even isolate particular connections in space by space-division-multiple-access-devices that function as "virtual wires" allocating all of the spectrum to each call.

Baran pointed out that "any transmission capacity not used is wasted forever, like water over the dam. And there has been water pouring here for many, many years, even during an endless spectrum drought." Although Baran urged as an ideal the transfer of the 480 megahertz of spectrum currently occupied by analog broadcasters to fiber optics and cable coax, he said, "We don't have to wait [for this ideal solution].... The existing spectrum can be more efficiently used by resorting to smart receivers and transmitters."


To conceive of Baran's model of wireless, begin by thinking of the human eye and comparing it to a radio. Like a radio, the eye is essentially a device for converting photons into electrons, pulses of electromagnetic energy into electrical currents. Geared for visible light rather than radio frequency signals, the eye is a receiving antenna. As radio technology moves up through the microwaves toward the infrared realmwith infrared wireless links from Canon now reaching 155 megabits per secondmany of the differences are dissolving.

Yet, in the crucial index of performance, the radio is drastically inferior to the eye. While most radios can receive signals across a span of frequencies ranging from the kilohertz to the megahertz, from thousands to a few million cycles a second, the eye can grasp signals with a total bandwidth of more than 350 trillion hertz (terahertz). That is the span of visible light, from 400 terahertz to 750 terahertz, red to purple.

How is it that your eyes command 350 terahertz of bandwidth and your FM radio around 20 megahertz, 17 million times less? It is not chiefly the special powers of the retina and other optical faculties. Radio antennas can collect an even larger span of frequencies. The difference is mostly behind the receiver. Backing up the eyes is the processing power of some 10 billion neurons and trillions of synapses. Backing up the radio antenna is a lot of fixed-analog hardware. Eyes are smart and aerobatic while the radio is dumb and blind.

In Baran's vision, the future of wireless is the replacement of current dumb radios by smart digital radios that resemble eyes. Coupling radio technology with computer technology, the antenna can acquire a brain. Smart radios can eventually process gigahertz of spectrum (billions of cycles a second). They can sort out the frequency channels much as eyes sort out arrays of color, and pin down codes and sources of radiation much as the eyes descry different sources, shapes, and patterns of light. For example, a smart radio could process phone calls, videos, teleconferences, geopositioning codes, speed-trap lasers, and emergency SOS's.

The result will be a transformation of the nature of the spectrum. The current real estate model will give way to a new view. Rights to spectrum will roughly resemble drivers' licenses for use on the highways. Today you use your 350-terahertz eyes to survey the highway in front of you and avoid other traffic. As long as you do not collide with other users, pollute the air, or go too fast (use excessive power), you can drive anywhere you want. As radios are computerized, they will be able to "see" the radio frequency spectrum as your eyes see the roads. Smart radios will be licensed to drive in open spaces in the air as long as they don't collide with other radios, overpower them, or pollute the airwaves.

As Baran argues, the fulfillment of this dream is at hand. It is the broadband digital radio or software radio. Essentially, the radios used in cellular or PCS (personal communications services) phones will be able to differentiate among frequencies; they will be able to tell which direction a signal is coming from and isolate it in space; they will be able to identify the language of codes and protocols and waveforms that it is using and download software translators. No longer caught in a dedicated set of channels, time slots, protocols, data types, and access standards, radios will be smart and agile rather than dumb and fixed frequency.


This will not happen tomorrow. But like any technological vista, it illuminates the future. It opens the way to a new wireless paradigm, fully in place shortly after the turn of the century, that will mandate an entirely new model of wireless regulation and a new method for judging the evolution of companies and their prospects. In general, the companies on the path to broadband digital radiosthe smart radiowill prevail over companies that hook their futures to hardwired machines linked to narrow spans of frequencies. Moore's law, the doubling of computer power every 18 months or so, is enabling the creation of broadband cellular radios in which most of the processing occurs in digital form.

Some of the first smart radios were built for the military. In Operation Desert Storm, the cacophony of allied combat radiossome 15 of them using a variety of frequencies, modulation techniques, encryption codes, and waveform standards, such as AM or FM or PCM (pulse code modulation)created a virtual Babel in the sand. Units needed a separate radio system for every radio (or radar) standard. As a result, the Pentagon launched the Speakeasy projectone smart radio that could process all the different standards in software. Made by Hazeltine and TRW, the first prototypes were demonstrated successfully in 1994. Because standards change over time and hardware improves at the pace of Moore's law, a software programmable radio also saves money. Rather than upgrading the system in hardware every time the technology changes, software radios can be upgraded merely by downloading a new software module.

Speakeasy engineers have spread the word through the cellular industry. Stephen Blust, now at BellSouth Wireless, is leading an international effort to create smart radio standardsthe MMITS project. Today, with the advance of an array of new digital technologies, including CDMA, TDMA, GSM, DECT 1900, SMR, PHS, and a spate of others, every urban area is becoming a Desert Storm of incompatible radios. Not only are these systems unable to communicate with one another, but they also require separate spectrum and base station equipment. All this redundant processing has raised the costs and reduced the universality of wireless and prevented cell phones from displacing wireline telephony.

The solution to complexity, however, is Moore's law: Put it on a chip. Reducing this Babel of complexity to silicon microchips, with hundreds of millions of transistors on centimeter slivers of sand that ultimately cost less than $2 to manufacture, smart radios can radically simplify the cellular landscape. Freed of most wires, poles, backhoes, trucks, workers, engineers, and rights of way, cellular should be far cheaper than wireline.

For example, the conventional analog base station that receives your cellular calls and connects them to the telephone network requires a million-dollar facility of 1,000 square feet. This structure may contain a central-office-style switch to link calls to the public switched telephone network, huge backup power supplies and batteries to handle utility breakdowns, and racks of radios covering every communications channel and modulation scheme used in the cell. This can add up to 416 radios, together with all the maintenance and expertise that multiple standards entail.

In the near future, one wideband radio will suffice. Digital signal processors ultimately costing a few dollars apiece and draining milliwatts of power will sort out all the channels, codes, modulation schemes, multipath signals, and filtering needs. Gone will be the large buildings, the racks of radios, the arrays of antennas, the specialized hardware processors. Gone will be the virtual honeycombs towering in the air in time and space with exclusive spectrum assignments and time slots, and possibly gone will even be the battalions of lawyers in the communications bar.

All this apparatus can be replaced by a programmable silicon base station in a briefcase, installed on any lamppost, elevator shaft, office closet, shopping mall ceiling, rooftop, or even a house. The result, estimated Don Cox of Stanford, the father of American PCS at Bellcore, could be a reduction of the capital costs of a wireless customer from an average of some $5,555 in 1994 to perhaps $14 after the turn of the century. That is a paradigm cliff of costs.

As smart radios are delivered in the first years of the new century, they will allow escape from the zoo of conflicting protocols. Base stations will be programmable in software, able to handle any popular protocols, including the new technologies that will be emerging. The world of wireless will escape the bondage of air standards, where if you live in a GSM (global services mobile) area, you are forced to use GSM, and if you live in a CDMA (code division multiple access) area, your communications-poor cousins visiting from Europe will have to give up their GSM phone and demand to borrow yours (will they ever give it back?). Under the new regime, different standards mean different software loaded into RAM (random access memory) in real time. Any cell can accommodate a variety of access standards, channel assignments, and modulation schemes, and the best ones will win.


To get there from here, however, will require heroic achievements in the technology of radios. Every radio must combine four key components: an antenna, a tuner, a mixer, and a modem. Easiest is the antenna. Even though antennas too are converging with computer technology and becoming smart, for many purposes a shirt hanger will do the trick. It is the other components that deliver the message to the human ear.

Tuners usually employ the science of resonant circuits to select a specific carrier frequency or frequency band. The cellular band, for example, comprises 25 megahertz at around 850 megahertz. The PCS band comprises some 30 megahertz at around 1,950 megahertz. A mixer converts these relatively high microwave frequencies into an intermediate frequency (IF) or to a baseband frequency, which can be converted to a digital bitstream.

Familiar in the PC world, a modem is a modulator-demodulator. In transmitting, it applies an informative wiggle (AM or FM, say) to the carrier frequency. In receiving, it strips away the carrier, leaving the information.

In the old world of dumb radios, transceivers join all these components into one analog hardware system. In the new world of smart radios, only the antenna and the front-end mixer are analog and hardwired. Channels, frequency bands, modulation schemes, and protocols all can be defined in software in real time. The radio becomes a programmable microwave eyea device that can see whatever colors of RF you want to send it.

The key to digital radio is the analog-to-digital converter. It takes a radio or intermediate frequency and samples it at least at a rate double the frequency to translate it into a series of numbers. Imagine a strobe light illuminating a dancer. The light will have to strobe at least twice as fast as the dancer moves or you will not be able to detect the dance. Indeed, in a phenomenon called aliasing, you may see a different, slower dance, as you see a tire rotating slowly in the wrong direction on a film. In a similar way, an ADC strobes (samples) the dance of inflected frequencies on the carrier wave. The resolution of the ADC is measured in bits, setting how high the number can be that defines the waveform and, in samples per second, determining how high a frequency the ADC can capture without aliasing.

Ultimately, early in the next century, the advance of analog-to-digital converters will dispense even with the mixer. Then the all-software radio will be here. Analog-to-digital converters (ADCs) will be able to translate microwave frequencies directly from the antenna into a digital bitstream. Alcatel has already accomplished this feat in the GSM cellular band at its labs in Marcoussis, France. But so far this almost totally digital radio is a stunt rather than a product. That will change.

Most of today's ADCs cannot function reliably in real time at microwave frequencies (above 300 megahertz). Therefore, mixers are vital. Whether digital or analog, a mixer is essentially a multiplier. As invented by E. H. Armstrong, the father of FM, mixers are superheterodyne. They use local oscillators (LOB) to multiply the carrier frequency with a lower frequency. The key result is a frequency that represents the difference between the LO frequency and the carrier. This frequency is an intermediate frequency that holds all the information borne by the carrier but at a level that can be processed by existing ADCs.

By far the most effective mixer is the paramixer invented by Steinbrecher Corporation of Burlington, Massachusetts, now owned by Tellabs and renamed Tellabs Wireless. This device can range gigahertz of frequencies with a spur-free dynamic range (a range of volumes without spurious crackles or harmonics) that could capture the sound of a pin dropping at a heavy metal rock concert. For a fully digital superbroadband radio, a cascade of these still-costly devices is still the best bet. The pioneer of this technology since it was conceived a decade ago by MIT professor Donald Steinbrecher, Tellabs's Burlington operation introduced the Steinbrecher MiniCell in May for wireless local loop and interior cellular applications.

Tellabs has had trouble selling its wideband radios for cellular applications, for which they may be overdesigned. With the increasing spread of CDMA, which ordinarily uses only one to three channels, the initial gains from a broadband radio are small. But for a wireless local loop, with many thousands of customers in the Third World using all available channels, a broadband base station could offer large effficiencies. Replacing a large number of costly custom radios with one programmable device, the MiniCell may find its niche.

As ADC technology continues to advance, however, it will relieve pressure on the mixer, opening the way to still cheaper and lower power solutions. With the expiration of Steinbrecher's patent on the paramixer, the business is opening up. Watkins-Johnson has created a tiny mixer device in gallium arsenide the size of your smallest fingernail. So has Mini-Circuits of Brooklyn, New York. "It has 50% less performance than Steinbrecher's, but it costs only 10% as much. Many customers say, 'It's a deal,'" observes former Steinbrecher CEO and president R. Douglas Shute, now contemplating a startup.

AD converters are now edging toward microwave frequencies. Both Analog Devices and Comlinear, a National Semiconductor company, have introduced 40-megasample-per-second products at a resolution of 12 bits. This allows more of the mixing to move into digital multipliers. The first of the digital downconvertor chips came from Harris Corporation of Melbourne, Florida. Harris now has parlayed its expertise in RF and mixers into the creation of a sophisticated programmable machine that demonstrates the management of multiple modulation schemes in one cellular radio. Introduced on the floor of the Fifth Annual Wireless Symposium Exhibition in late February in Santa Clara, California, the Harris smart radio showcases its programmable HSP50214 digital downconvertor chip and is run from a PC. With an array of displays, the machine is designed to allow configuration and testing of smart transceivers from a Windows PC.

With high-powered digital signal processors and leading-edge ADCs, Analog Devices is a paragon of the digital radio paradigm. At the CTIA (Cellular Telecommunications Industry Association) meeting in San Francisco during the first week of March, Analog introduced a wideband smart radio tuned to the cellular band but applicable through the PCS band as well. A reference design to be used by infrastructure manufacturers, it displays an array of new chips from Analog comprising a specialized ADC called the 6600, tunable filters called the 6620 and the 6640 that function as a digital tuner, a SHARC DSP chip that performs the modem and channel-coding role (any advanced DSP will do), and a "sinfully cheap" Watkins-Johnson mixer chip the size of your fingernail. Incorporating an automatic gain control and a received signal strength indicator, the ADC is customized for smart radio applications.

The antenna is from Radio Shack (most any will do). From a Windows PC using Visual Basic, Analog engineers can move from one cellular channel to another and from GSM to CDMA to DECT 1900 to IS-136 to the Japanese Personal Handyphone system (PHS). As manufacturers around the globe converge on a single intermediate frequency of 70 megahertz, the reference radio could adapt to any cellular band, from 850 megahertz on up. All you would have to do is change or retune the mixer. Accordin~ to Tom Comine Gratzek, Analog Devices's director of base station marketing at the Analog communications center in Greensboro, North Carolina, customers say, "Shazaam!"


Interest is acute at all major telecom equipment manufacturers, from Ericsson to Motorola, and champions include every telecom company that thinks it may have guessed wrong in the GSM, TDMA, CDMA wars. BellSouth, for example, is slipping into a GSM ghetto, but it dreams of deploying smart radios that can play any popular standard and allow it to filch (i.e., service) CDMA customers. Also a TDMA orphan, AT&T could buy cheap, all-purpose base stations that allow it to sell any favored brand of service. Ericsson is using the technology to create indoor GSM base stations that can fit in a closet, and if worst comes to worst (as it will), Ericsson will also offer CDMA, perhaps initially as an overlay for data.

By drastically enhancing efficiency in the use of spectrum, broadband digital radios will lend new force to the industry's move up the frequency ladder toward bandwidth abundance. They enable the seamless convergence of the cellular band not only with the PCS band but also with an array of other applications such as the low-powered ISM (industrial, scientific, and medical) bands at 900 megahertz used by Baran's Metricom startup, the 24-gigahertz band of Associated Communications, the 28-gigahertz band of Local Multipoint Distribution Service (LMDS) used by Cellular Vision for wireless cable, and the 38-gigahertz band of WinStar. This up-spectrum bias assures the continued success of companies pressing the frontiers of microwave integrated circuits, low-noise amplifiers, power amplifiers, and other devices that function in the gigahertz.

Going over the cliff of costs, the industry can introduce radically new products. We have just undergone the epoch of the personal computer, climaxing in 1996 with PCs outselling TVs in units for the first time. We are now entering a new era when a new form of PC will be dominant. It may not do Windows, but it will do doors. Tetherlessly transcending most of the limitations of the current PC era, the most common PC will be a digital cellular phone.

It will be a dataphone, as faithful readers of these pages will know. It will be as portable as your watch and as personal as your wallet. It will recognize speech and convert it to text. It will plug into a slot in your car and help you navigate streets. It will consult electronic yellow pages and give directions to the nearest gas station, restaurant, police headquarters, or hotel. It will collect your news and your mail and, if you wish, it will read them to you. It will conduct transactions and load credit into a credit chip on a smart card, which can be used like cash. It can pay your taxes, or help you avoid them, or soothe you with soft music as you do your calculus homework. It will take digital pictures and project them onto a wall or screen, or dispatch them to any other dataphone or computer. It will have an Internet address and a Java run-time engine that allows it to execute any apples or program-written in that increasingly universal language. Or it will dock in a more powerful machine to perform more demanding functions. It will link to any compatible display, monitor, keyboard, storage device, or other peripheral through infrared pulses or radio frequencies.

And, oh yes, it will unlock your front door or car door, open your garage door, or even play Jim Morrison songs, if you are old enough to care for those swinging Doors of the 1960s (amazingly enough, my teenage daughters do).

Sorry, though, Nokia, your model 9000, which comes closest today to this new machine, will not cut it, at least in the United States, because it is based on Europe's increasingly obsolescent GSM standard. Also offering the right form factor but the wrong access standard is the IBM-BellSouth Simon, which is based on the U.S. analog cellular system (AMPS) or CDPD (cellular digital packet data). The most common PC will not be a GSM or CDPD device, because it will soon need to provide bandwidth on demand while draining the lowest possible power, whenever it is not plugged in. Thus the first PC of the new paradigm will probably have to be CDMA, built from the bottom up to provide bandwidth on demand, according to TCY/IP Internet standards, at a handful of milliwatts of communications power.

Among the companies soon to supply such machines, resembling the popular U.S. Robotics Pilot, are Sony, Qualcomm, Lucky-Goldstar, and Samsung. In cooperation with Alcatel, the European giant, which has just announced a CDMA program, Qualcomm base stations will soon contain a GSM link that can allow such CDMA dataphones to tie seamlessly to GSM systems in Europe. This will permit European carriers to use CDMA to expand capacity without jeopardizing their GSM customers.

Inspiring the Baran vision of wireless is the spectronic paradigm, in which most of the industry, from personal computers to cellular phones, moves on into the microwaves and is discussed more in terms of megahertz and gigahertz than in the usual metrics of mips and bits. The spectronic paradigm tends to favor the manufacturers of gallium arsenide, indium phosphide, and silicon germanium devices. Even as Philips and other firms push silicon bipolar chips toward microwave frequencies, the industry will move to higher domains of spectrum where gallium arsenide and indium phosphide tend to prevail. For the power amplifiers needed in every cell phone, gallium arsenide is superior to all the silicon variants. Pushed by the advance of the spectronics paradigm, the current ride of Vitesse, Anadigics, TriQuint, and other gallium arsenide innovators is likely to continue.

The major long-term winner is silicon germanium. Pioneered by IBM fellow Bernard Meyerson and tested and sampled by Analog Devices, silicon germanium combines much of the manufacturability of silicon with the high-frequency operation of gallium arsenide. IBM has recently contracted with Hughes's communications division to develop silicon germanium microwave devices.

As the technology advances, the broadband radios will be ideal to offer video teleconferencing, World Wide Web, and other image-rich wireless content, including CDMA bandwidth on demand. Data, not voice, will be the critical application. As people brandish their dataphones around the globe, linking to convenient displays through IR connectors, users can break out into a tetherless telecosm where they can work or play, study or pray, anywhere they go.

A major supplier of wireless in Third World countries may be NextWave, the aggressive CDMA vendor for PCS, now preparing an IPO. As a "carrier's carrier" providing only infrastructure and network services and leaving the sales and marketing to the locals, NextWave will join its complementary sister company in space, Globalstar, at the heart of a CDMA fabric of culture-independent worldwide communications. Watch Motorola's obsolescent Iridium, with its exclusive spectrum requirements and its effort to bypass all local infrastructure, sink like a stone.

The new paradigm of wireless joins Baran's two key inspirationsInternet and smart radioto burst the chains of geography. People who want leading-edge computers and communications can get them wherever they may live. Using Globalstar, Teledesic, and other low-earth-orbit (LEO) satellite systems that will be available as the smart radios roll out, students in the Third World can study or work in the First World. Teachers and entrepreneurs in the First World can serve and employ people around the globe. Imagined gaps between the information rich and poor will collapse in an infoscape equally accessible to all.

Baran has not spent his life in speculation or prophecy. Living at the heart of Silicon Valley in a walled and radiantly flowered community a few minutes down Middlefield Road from Netscape, Baran sits at the epicenter of a series of entrepreneurial creations. His home-office PCs and Power Macs are linked to the Internet through the Palo Alto Cable Co-op by cable modems from Com21, which he founded and now chairs. To run multimedia programming down twisted-pair wires, the regional Bell operating companies now propose to use discrete multitone technology (DMT), the basic technology conceived by Baran for Telebit and now the leading digital subscriber loop (DSL) method, taken up and perfected by Amati Communications, just down the road in San Jose. StrataCom, recently purchased by Cisco for $4 billion, began as a leveraged buyout spinoff from Baran's Packet Technologies.

Metricom, a Baran company with investments from Bill Gates, among others, offers wireless Internet services through Baran's neighborhood and at campuses across the country. Baran's company, Equatorial Communications, introduced spread spectrum commercially as a way of delivering information from satellites below the noise floor required by the FCC. Spread spectrum is now, in the form of the CDMA of Qualcomm and Globalstar, the world's fastest-growing communications technology. And it is the basis for the flourishing, unlicensed wireless systems, such as Metricom, operating at less than one watt of transmit power in the ISM (industrial, scientific, medical) bands.

Collectively, the visionary concepts of this once-myopic and still-modest engineer offer the foundation of an effort to reinvent the Internet in an increasingly wireless form and reshape the communications policies of the nation and the world.

George Gilder is a contributing editor of Forbes ASAP. He also publishes the monthly Gilder Technology Report. For a newsletter subscription, call 1-800-888-9896. For more information about GTR and its September conference, email

Like any demiurgic force, the Internet has spawned many creation myths. Vividly depicted in Where Wizards Stay Up Late: The Origins of the Internet, by Katie Hafner and Matthew Lyon, is the MIT version, centered on the late J. C. R. Licklider and Lawrence G. Roberts, now chief technical officer at Connectware in Silicon Valley. At MIT, ARPA, and Bolt, Beranek & Newman, the popular "Lick" was undoubtedly a major influence in early computer circles, but his Internet role seems elusive (I guess you have to have been there to grasp the significance of his bromides on "Man-Computer Symbiosis"). But Larry Roberts was the ARPA executive who, in 1967, launched and specified the project that evolved into the Net. Bob Metcalfe's historic MIT thesis, Packet Communication (1973), begins with a quote from Roberts. Although Roberts says he never read any of Baran's work, it was more voluminous and detailed than Roberts's, preceded it by at least four years, and was widely propagated by RAND in the intimate society of defense oriented computing.

In any case, in October 1996 NEC gave an award of 5 million yen apiece to the three key inventors of the Internet: The honorees included Vinton Cerf for designing TCP/IP (with Robert Kahn), and Tim Berners-Lee for conceiving the key protocols of the World Wide Web. Both are household names in the industry. The third was Paul Baran, for his prior invention of packet switching and routed digital computer networks survivable under attack and orders of magnitude cheaper than their analog precursors. In other words, as early as 1962 Paul Baran, the least known of the honorees, conceived the essentials of the Internet.