Michael Borrus and François Bar
A BRIE Research Paper
Copyright © 1993 by BRIE
March 16, 1993
The authors are grateful for the support of the Alfred P. Sloan Foundation, Directorate General XIII of the Commission of the European Union, and the Los Alamos National Laboratory.
I. Networks at a Crossroads
A. New Technological Possibilities
B. Explosive Growth of Network Use
C. Computing as Driver
D. Blurring of Competitive Lines
E. Policy Confusion
II. Emerging Approaches to the Discontinuity
A. Corporate Networks
1. The Challenge
2. Trends in Corporate Information Networking
3. Network Implications
B. The Internet
C. Traditional Carrier Networks -- Telcos, Cable TV, VANs
1. The Current Competitive Position
2. The Business Services Market
3. Residential Video
4. Telco Choices
III. Policy And Network Evolution
A. Infrastructure as Network Portfolio
B. Policy and the Network Portfolio
C. The Inevitability of a Continuing Role for Government Policy
Once confined to arcane debates among computer experts, the Electronic Superhighway has captured the popular imagination just as the National Information Infrastructure has become so fashionable in policy circles. The promise: a cornucopia of multi-media (integrated video, audio, text, image and data) entertainment, education and industrial application for every firm and household in America.
Comparing tomorrows digital networks with yesterdays highways provides a compelling and familiar vision. Modern transportation technologies revolutionized work and leisure by facilitating the movement of goods and people. Modern information networks are similarly likely to transform todays activities, which increasingly depend upon the transmission and processing of information.
But beyond the hype and the powerful vision, the debate quickly derails. How will we get from todays networks to tomorrows Superhighway? Who really wants it and what will we actually do with it once we have it? Who is going to foot the bill?
Rather than acknowledging the uncertainty, policy and business strategy have been substituting a certain technological image for a highly uncertain market and regulatory evolution: In that dominant technological image, a single network based upon the integration of all of the new technologies would take care of all applications known today (from entertainment video to business communications), with room to spare -- in effect, the all-singing, all-dancing, all-integrated fiber optic Superhighway.
We argue here that the Superhighway image of the future is the wrong image; that it leads to bad strategy and even worse policy. What is emerging in the U.S. is not a single integrated infrastructure, but a portfolio of advanced networks -- phone, CATV, broadcast, wireless, National Research (the NREN), and private. Each of these networks is evolving toward Superhighway-like status (i.e., toward greater intelligence and bandwidth). However, each is driven by very different market and policy dynamics; each is tailored for widely different kinds of applica-tions. The challenge is to ensure that the emerging network portfolio can act as a proper infrastructure -- via development of a logical architecture that enables interoperability and transparent access with varying kinds of performance, privacy and ownership -- even as each piece evolves at its own pace in the market.
The Allure of the Single Integrated Superhighway Infrastructure
The "all-singing, all-dancing" infrastructure vision leads to three central questions which dominate alike policy and the planning of network providers: Who will build the information pipe? (usually Telcos or CATV); what will fill the pipe? (usually entertainment video); who will pay? (usually, not granny). If, as we argue here, the emergence of a National Information Infrastructure will be neither a straightforward evolution of existing networks nor well-defined for an extended period of time -- it will be discontinuous and largely unknown -- then these are simply the wrong questions.
Who should build the pipe?
In reality, a diverse portfolio of networks is being built, which leads to very different questions: What is the proper network portfolio? How can universal access, interconnection, and inter-operability be maintained among the portfolios elements? What levels of performance, reliability and privacy are appropriate? Does ownership matter?
What will fill the pipe?
Demand for fundamentally new technologies is impossible to predict with any degree of accuracy given the inherent market, technical and economic uncertainties. But applications are already emerging that bear little resemblance to either the traditional phone call or the simple image of entertainment video. The catch: All of the emerging applications require extended experimentation and learning by users before they are fully developed, their benefits are understood, and sufficient demand for them is realized to provoke investment. Thus, the right question: How can sufficient experimentation and learning with the new technologies be guaranteed so that demand for advanced applications can be effectively generated?
Who should pay?
Emerging patterns of broadband use suggest that advanced users will increasingly be willing to pay for access to broadband capabilities. But those disparate patterns of emerging "premium" demand may not result in a coherent portfolio infrastructure. Thus, the right question: Can we create incentives (or remove disincentives) for those willing to invest in the evolution of the network portfolio, while minimizing the costs of ensuring universal access for those least able and willing to pay?
Abrupt, Discontinuous Change
New Technological Possibilities and the Explosion in Use
Digitization, data compression, broadband capabilities (high speeds, low latency, and high bandwidth), and ubiquitous computing power in industrial and consumer hands, have all combined to create radically new network possibilities. In domains like private networks and the National Research and Education Network (the U.S. branch of the Internet), where the technologies are most advanced and usage is least constrained by regulation, network use is growing at rates of 10%-20% per month. Such growth threatens to fill existing networks to overflowing and to overwhelm their capacities.
More significant, consider the variety of emerging uses and their distance from the traditional phone call: At one end of the range, interactive applications like concurrent Computer-Aided Design (CAD) among engineers in remote locations require a network capable of switching very large bursts of data very rapidly. The key demands on the underlying network are low latency and high interactivity. Bandwidth is needed to guarantee instant response, not just to handle massive amounts of data. In fact, these applications work much better if the broadband connection is far from saturated (as one network manager put it, the broadband "pipe" should never be filled, to be instantly available when the user needs it).
At the other end of the range are applications such as Xerox PARCs "video window," in which employees from two far-apart locations can keep a window into each others offices open for weeks at a time, aiming to interact as informally as they do around the water-cooler. In this case, the connection setup time or the networks latency are almost irrelevant, but the broadband conduit between the two locations is continuously utilized at its maximum capacity.
Applications with similarly unconventional requirements spread along the continuum between these two extremes. Most involve computer-to-computer communications such as network-supported information retrieval (e.g., Thinking Machines Wide Area Information Servers (WAIS)), or the "knowbots" (knowledge-robots) being sent through the Internet to sift through huge volumes of data and bring back only relevant information. Others, like video-on-demand and interactive TV (e.g., home shopping from smart CATV converter boxes) require fundamental changes in established broadcast patterns.
The Dynamics of Competition
Now consider the varying market and policy dynamics driving the development of advanced networks and the myriad applications examined above. Private networks (e.g., corporate networks like IBMs, or Value-added Network service suppliers like EDS) are largely unregulated by policy. They are driven by corporate competitiveness concerns like globalization, faster cycle times, and the need to link with suppliers and customers outside the bounds of the firm. They are the leaders in developing distributed, cooperative, computer-computer applications for generating competitive advantage in industry.
By contrast, the Internet is publicly subsidized, with little regulation on use, and ties together an enormously sophisticated user-community of scientists and engineers (and increasingly the emerging Nintendo genera-tion of generalists) on a world-side basis. Interactive and real-time applications range from visualization of global climate change, to broadcast of commercial radio and TV, to game-playing and on-line conferencing, to publication of research results in fields as disparate as AIDS research and political science.
Traditional local phone companies are highly regulated by both the FCC and State Public Utilities Commissions, have mandates like Universal Service and Common Carriage, and provide traditional phone, fax, and low-rate transmission of data, plus local access services for large businesses and other carriers (like the long-distance carriers). In the latter applications, they are coming under significant competitive pressure from Alternative or Competitive Access Providers (like Teleport Communi-cations) who are less regulated and, in fact, have been advantaged by regulation to permit them to cream-skim local business opportunities. Driven in similar fashion by a mix of regulation and traditional phone applications, cellular telephone networks and other emerging wireless service providers are entering the local marketplace.
By contrast, traditional long-distance carriers like ATT, MCI and Sprint, are less and less regulated and more and more global in orientation. They provide a panoply of services that range from traditional direct-dialed and operator-assisted long-distance, to highly flexible Software Defined or Virtual Network Services for business applications, to tailored service packages that include network management for large customers, to video-conferencing, integrated digital services, and customized billing.
CATV companies are regulated by local monopoly franchise agreements and increasingly by the FCC for rates, but have no Common Carriage or Universal Service obligations like the phone companies. Their market is fundamentally driven by the delivery of entertainment (and leisure activities) to the home, including broadcast TV and fledgling "interactive" applications like pay-per-view, video-on-demand, and home-shopping. More and more, however, through purchase of Competitive Access Providers and alliances, they are moving toward traditional telecommunications applications like "commuting from home" and local access services.
In a number of domains, the markets of these varied networks are beginning to overlap, as formerly distinct market boundaries are eviscerated by a combination of alliances, technical convergence, and perceived market opportunities that encroach existing turf. It is here that differences in the individual market and policy dynamics create disparate patterns of competitive strength and weakness. For example, CATV companies are far better placed to attack opportunities in entertainment than traditional phone companies, just as ATT is in a much better position than CATV companies to provide switched, high-bandwidth computer-computer applications for industry. Thus, the individual network evolutions are likely to remain distinct -- delivering the network portfolio rather than an integrated infrastructure -- even as competitive bounds blur in the pursuit of emerging market opportunities.
From Confusion to Coherent Policy
As the discussion has hinted, the traditional vocabulary and concepts of U.S. regulatory policy have become outmoded or even irrelevant. In different places, policy is simultaneously fostering and frustrating competition, subsidizing and refusing to subsidize network evolution, facilitating and delaying technological development, requiring and denying public service obligations, permitting and disallowing widespread experi-mentation and use, enabling and blocking the delivery of new applications. There is neither a clear policy direction nor a coherent infrastructure vision.
As we have argued, driven by actual users and their initial applica-tions, there are multiple networks representing different technological trajectories emerging in the face of discontinuous change in the U.S. This is, in effect, a prudent bet-spreading by the market under conditions of extreme uncertainty. There are two fundamental problems with this arrangement: Fragmentation and Unequal Access. The U.S. used to have a single, standardized telecommunications network infrastructure. Deregulation and divestiture have allowed the emergence of a pluralistic and diverse set of networks and service providers, providing innovative and customized applications to specific sets of telecom users. The down side of that approach has been the lack of mechanisms to insure the interconnection and interoperability of the various network components, leading to the increasing fragmentation of what was a coherent infrastructure.
Similarly, while a pluralistic network makes it possible to serve individual users needs, it also tends to make advanced capabilities available only to those rich enough to afford them and aware enough of the potential benefits to demand them. Others risk being left behind. In the past, when "Ma Bell" deployed new technologies in her network, responding to pressure from its most demanding clients, all those connected to the network benefited. Within a pluralistic infrastructure, the challenge is to insure broad diffusion of innovations, and broad access to new services. Failing that, the U.S. would end up with a bifurcated infrastructure, sharply discriminating between the "information rich" and the "information poor." Those denied access to advanced information services will be unable to develop the skills they -- and their employers -- will need to take advantage of the emerging infrastructure.
Policy and the Network Portfolio
The three issues raised at the outset help to situate a new policy agenda that takes account of the problems of fragmentation and access and that acknowledges the reality of the emerging network portfolio.
1) Let market competition drive portfolio development, while policy focuses on three tasks: (i) Defining a "logical architecture" that permits access, interconnection and interoperability among the portfolios networks at several different levels of reliability, security, and performance; (ii) self-consciously mandating large-scale experiments and testbeds in interoperability between the portfolios networks to practically develop the logical architecture; (iii) reviewing the portfolio for gaps in service to major constituencies and encouraging the development of missing elements.
2) Policy should encourage exploratory use of the portfolio among targeted user populations on the Internet model, to ensure the necessary experimentation and learning that can coalesce into demand for new services and applications.
3) Where possible, policy must encourage major business users and network providers to finance large pieces of the emerging portfolio, while finding other additional means as necessary to reduce the costs to poor and initially non-participating constituencies -- including premium pricing or royalty-based pricing mechanisms and novel financing schemes like the sale of bonds.
In our view, Deregulation helped create the current discontinuity, but only an effective combination of competition and proactive policy can deliver a coherent infrastructure for economic development in the next century.
The current changes sweeping information networking in the United States promise a profound infrastructure transformation -- a drastic, discontinuous change rather than simply an incremental evolution of the existing telecom infrastructure.1 It is likely that this transformation of the infrastructure will permit an equally deep transformation in the economic and social processes it supports. At stake may well be the future economic development trajectory of the United States and its wealth and power in the 21st Century.
Dramatic infrastructure change supposes practices and uses that are radically new and therefore unknown. As a consequence, past practices and experiences that have been useful guides in dealing with incremental change provide only limited help in exploring the current transformation. In part, that is why there appears to be such a disconnect between the technological vision that traditional infrastructure providers like the Public Telecom Operators want to provide (e.g., an evolutionary extension of their existing network and services) and what their major customers anticipate needing.
The very nature of the change presents a series of obstacles, not the least of which is that conclusive evidence of dramatic transformation will only be available after the fact. Imagine trying to document the pending industrial transformation that electricity would generate from the vantage of 1880. Instead, we must explore current network provision and use, looking for symptoms of change. We do that by examining the drivers of change -- technology, uses, competition, and policy -- in each of the major networks that together may constitute the future U.S. infrastructure: the local and interexchange carriers, Cable TV, private corporate networks, the Internet, and VANs. The resulting hypotheses will have to be tested against data as it emerges.
If a revolutionary transformation of the network infrastructure is indeed underway, the challenge is to come up with policies and business strategies that acknowledge the discontinuity. Firms and government should proceed from an explicit recognition of the uncertainties rather than from the image of incremental change that drives todays choices. We argue in particular that the diverse networks that must make up the U.S. information infrastructure already exist. There is no need to build one all-purpose network -- the all-singing, all-dancing, all-integrated, all-broadband superhighway. Rather, the central challenge is to encourage a logical infrastructure to come into being through the technical evolution and virtual integration of the existing portfolio of networks.
I. Networks at a Crossroads
This section presents five major symptoms of the transformation we believe is underway. These include 1) the dramatic new technological possibilities associated with advanced networks, 2) explosive growth in the volume and dramatic change in the character of network use, 3) a paradigm shift in computings relationship with networks, 4) evisceration of the formerly clear market boundaries around competition in network provision, and 5) the resulting confusion in domestic U.S. policy toward networks.
A. New Technological Possibilities
Three major technological trends combine to create the potential for wholly new networking possibilities. These possibilities will in turn give rise to new networking practices that will guide the evolution of the network infrastructure along new trajectories.
The first trend is associated with the digitization of the network. Increasing the systems intelligence permits increasing differentiation of network performance, of service (or application) choices, and ever more intimate management and control. Of these, perhaps the most momentous has been the appearance of a "flexibly separable" network management layer, embodied in the software that controls modern networks. Network control no longer required network ownership and could be shared among various participants in the telecom infrastructure: network owners and operators, network users, and third parties. This enabled a new evolution of network practices, building upon network experimentation and re-configuration, which we characterized as a technology trajectory based upon cumulative learning.2 These new practices, deriving not only from the knowledge accumulated by network providers but also from the experience and expertise of network users, constitute a discontinuous change in the dynamics driving the evolution of the network infrastructure.
The second trend results from the emergence of broadband transmission. Increasing bandwidth and speeds now permit transport integration and unprecedented flexibility and performance in network use as infrastructure to economic activities. Until today, networks have represented a principal technological bottleneck confronting information systems simply because they were unable to support widespread geographical extension of the sophisticated practices being developed at the local level. Replacement of the existing physical infrastructure with one drastically more capable is gradually bringing Wide Area Networks capabilities to par with those of Local Area Networks (LANs). Related progress with other transmission technologies, such as wireless trans-mission and data compression, compounds the impact of the transition towards broadband. Combined, these transmission technologies enable a drastic, discontinuous transformation of the underlying physical network infrastructure, which holds the potential to bring the capabilities of networks in synch with those of CPE, thereby removing the current network bottleneck.
The third trend is the dramatic increase in the functionality, performance, and variety of the CPE connected to networks. Years ago, analysts speculated that dramatic increases in the computing power of such equipment might substitute for network need. For example, why use a network to remotely access a mainframes processing power or file storage capacity when you can have equivalent power directly on your desktop? Experience is proving not only that the network is still necessary but that the need for it grows at least in proportion with the terminals power. Moreover, "computer terminals" have become more pervasive in offices, factories, homes and physical infrastructure, embedded within machine tools and office tools, in automobile systems, notebooks and phones, camcorders and CD players, and medical instruments. These products and newly emerging ones like Apples Newton and H-Ps Palmtop, are the leading edge of future high-volume electronics products, all built around embedded computing resources. Such products are increasingly wireless, compact, and need to be networked to fully exploit their potential. As a result, networks must cope with increasing mobility and portability through the incorporation of wireless technology. The overall need for flexible and transparent interconnection will only increase dramatically as such products take hold. This trend towards large numbers of highly sophisticated devices increasingly relying upon a network also constitutes a drastic, discontinuous transformation in the demands being placed upon the network infrastructure in terms of both the transmission volumes and the new pattern of use it will have to support. As a result, it will require discontinuous change in the network itself.
Together, these three trends enable a radical transformation of the underlying information network infrastructure and open radically new possibilities for its management and control. They open the way for many new connections possibilities, which will permit many new use patterns. While they open many avenues -- different trajectories -- it is not yet possible to predict how these trends will precisely combine and play out. What is clear, however, is that they can combine and play out in numerous ways, each of which has different implications for the network infrastructure. In short, there is not one technological trajectory associated with network evolution, rather there are likely to be several. The choice of the "right" trajectory remains wide open at the moment.
B. Explosive Growth of Network Use
While in the past, major network transformations have been driven largely by suppliers (typically through deployment of new, more powerful technologies), the current choices are increasingly user-driven. In this context, "users" may be business or residential customers of the phone and Cable TV companies, academics using the Internet, employees of a corporation using its private network, or even other carriers and network providers. The particular ways in which user pressure builds up, the channels through which it is expressed, and the requirements it places on the network vary across cases. However, in the cases where this pressure has become most acute (e.g., corporate network or the Internet), there is a common pattern to its building up.
This sub-section argues that three usage trends provide concurring evidence of dramatic change. These are the explosion of use (both in volume and variety of applications) of networks that are relatively unconstrained by usage limitations, accelerating demand for network management capabilities, and increasing self-awareness by leading users about their network needs. Taken together, these trends stress existing networks to the breaking point. Only new networks can accommodate this exploding traffic and the related shifts in use patterns. In addition, there may be a fourth user-driver appearing, namely the simultaneous demand for customization and standardization. Users want to be able to tailor the emerging network applications to their needs, but do not want to pay the price of full customization. Thus, they push toward standards to permit open systems commoditization in a multi-vendor environment. Let us examine each trend in turn.
Since digital technology started being deployed through the telecom network, data traffic has been growing much more rapidly than voice traffic. Because it started from a much smaller base, however, the volume of data transmitted over the network remained small compared to voice volume and, until now, could easily be accommodated over the existing network. Further, the predominant types of data applications in use (e.g., file transfer, e-mail) did not by and large conflict with the configuration of the existing voice network.
Two things have happened recently however. First, data networking is growing very quickly in the domains where this growth is the least constrained, in particular the Internet or corporate private networks. The resulting large volumes of data traveling the networks are becoming harder and harder to accommodate within the existing networks. For example, as will be described in greater detail below, data traffic growth on the order of 20% per month is prompting many corporate network planners to overhaul completely their data network over the next few years. Similar data traffic growth was behind the recent upgrade of the Internet backbone from T1 to T3, and has provoked gigabit test-beds for future upgrades.
Second, the type of networking applications spurring this traffic growth are quite distinct from the traditional ones, placing very different demands upon the networks that support them. As we try to understand and forecast these emerging uses, analogies to the current network services can be quite misleading. For example, broadband applications often are presented as extensions of current narrow-band services: we envision making "broadband phone calls" in which the additional band-width could be used to carry information such as video images in addition to voice or to transmit a large file very quickly. While the amount of information transmitted would be much greater than in a traditional phone call, the communication pattern is quite similar. In fact, the actual applications of broadband networking emerging today follow very different patterns, covering quite an extensive range.
At one end of the range, one finds interactive applications based on client-server architecture, which are growing much faster than traditional file transfers. They typically require a network capable of rapidly switching very large bursts of data. The key characteristics of the underlying network are low latency and high interactivity. Bandwidth is needed most importantly to guarantee instant response, not merely to handle the large amount of data. In fact, such applications work much better if the broadband connection is far from saturated (as one network manager put it, the broadband "pipe" should never be filled so it is instantly available when one user needs it). At the other end of the range are applications such as Xerox PARCs "video window," in which employees from two far-apart locations can maintain a window into each others offices open for weeks at a time, aiming to create the kind of informal interaction only possible at a single location today. Here, the connection setup time or the networks latency are almost irrelevant, but the broadband conduit between the two locations is continuously utilized at its maximum capacity. Applications with similarly unconventional requirements are being created in between these two extremes. Most of them involve computer-to-computer communications such as CAD interconnection for concurrent engineering, network supported information retrieval such as Thinking Machines Wide Area Information Servers (WAIS), or the "knowbots" being sent through the Internet to sift through huge volumes of data and bring back only relevant information.
The requirements these emerging applications place on the underlying network differ drastically from those of traditional applications, not simply because they fuel tremendous traffic growth, but also because their traffic patterns have little to do with those of traditional phone calls. These new requirements are becoming impossible to accommodate within traditional networks, and underlie emerging pressures for the deployment of radically different networks. In our view, the extreme stress under which existing networking facilities will soon find themselves constitutes another symptom of the discontinuous change afoot.
In parallel with exploding traffic, users increasingly demand dramatic improvements in network management capabilities. This is reflected in several ways. Network management software has become a high-growth business segment for all types of networks, from LAN specialists like Novells Netview, to T1 specialists to new releases from all of the major computer companies. Simultaneously, major users who manage their own networks are discovering that network management is occupying an ever greater percentage of total network costs -- so much so that they are handing non-essential management off to providers. In response, common carriers are pressing to provide greater functionality to users, whether through enhanced VPNs like those now provided by all the major U.S. interexchange carriers, or through local BOC unbundling toward provision of ONA. Their major competition comes from VANs, who are repositioning themselves explicitly as providers of end-to-end network management capability.
The emphasis on network management confirms a finding of our last study,3 that successful evolution of a telecommunications infrastructure requires network suppliers and network users to combine their distinct skills and knowledge. During what we have described as the telecommunications learning cycle, network users typically learn-by-using how they can best harness available technologies to serve their business objectives, while network providers learn-by-doing the art of deploying networking technologies to serve what they perceive as their clients demands. Neither one (providers nor users) fully possesses the knowledge necessary to promote the evolution of a successful infrastructure. Rather, they both must build upon the others knowledge and cumulatively improve the network and its applications.
Until recently, high-speed networking had been primarily a network suppliers idea. Network equipment providers and network operators have been developing an understanding of how emerging technologies, from fiber optics to ATM switching, would make it possible to build and operate a broadband network. While they had some ideas about the potential applications this would open, they could not point to existing demand for those applications nor, for that matter, be sure that their ideas corresponded to actual user needs.
We now see signs that user awareness about the potential applications of high-speed networking is rising, with the potential to unlock the broadband learning cycle. Indeed, the current traffic explosion on the Internet or within corporate networks is fueled by real applications and real users, not simply network operators pipe dreams. Several factors have combined to create broadband "test beds," where various users are starting to become familiar with these new possibilities. On one hand, the multiplexing of large numbers of narrow-band communications has led large corporate users and network providers alike to set up what can now be used as broadband links and to start experimenting with them. Within local area networks, or within the NREN test beds, further experimentation is underway. On the other hand, as they accumulate more experience with current network capabilities, users become more aware of the constraints imposed on them by the current system and more vocal about demanding change. Pushing the limit of existing networks, they are getting closer to forming clear ideas about the potential benefits of broadband networking.
Overall, we interpret this emerging unlocking of the broadband learning cycle as an important symptom of discontinuous change. It suggests that a critical element -- the possibility for users to experiment and learn -- now exists to initiate a new technology trajectory, based upon the drastically different possibilities of broadband.
C. Computing as Driver
Much of the shift in usage patterns is being fueled by new ways to use computers and by their increasing ubiquity, power and diverse applications. At the leading edge of the business world, a new computing paradigm is emerging, one that de-emphasizes simple linkage of computers in favor of "cooperative" computing.
Two basic elements underlie the full-blown emergence of this paradigm, ubiquitous computing and a coherent infrastructure that permits interconnection for cooperative purposes. The move towards open systems in client-server architectures represents a drastic change in how computers communicate. From file transfer and terminal emulation, demand shifts to interoperability and real-time interaction. This represents a quantum shift in how computers must use the network, in effect, from low-rate/dedicated to switched/broadband. As described earlier, the rise of high-volume, digital, networkable electronics products based on microcomputers will only exacerbate this shift.
In our view, the shift in computer usage will ultimately allow a resolution of the famous productivity paradox -- that information technology (IT) appears everywhere but in the productivity numbers.
As we argued in the last study, one reason IT investments have not translated into higher productivity is that they have primarily served to automate existing tasks. They often automate inefficient ways of doing things. Realizing the potential of IT requires substantial re-organization. The ability to re-organize tasks as they become automated rests largely on the availability of a coherent infrastructure, i.e., a flexible network able to interconnect the various computer-based business activities.
A proper analogy lies with the introduction of electrical power.4 The benefits of electrification appeared when small generators were appended to individual machines, rather than large generators appended to entire factories. This allowed reorganization of the shop floor. Similarly, automation through IT was first done with large mainframes "appended" to the office, without reorganization of the tasks. The following parallel step has been the decentralization of computer power towards the desktop, or the NC machine-tool. These decentralized computers are only now being interconnected, so as to allow and support re-organization. Where this has been effectively accomplished, there are corresponding gains in productivity.
As this occurs, the next step of flexibly integrating network capabilities with diverse business applications is moving beyond the use of telecom as a utility, and toward its use as a real infrastructure for competition. This change in the character of network use constitutes yet another symptom of discontinuous transformation, which section II will explore in greater detail.
D. Blurring of Competitive Lines
This sub-section argues that evidence of discontinuous change can also be found in the evisceration of the clear boundaries that used to separate the various network alternatives. To be sure, the introduction of competition on the U.S. telecommunications scene brought along a variety of new entrants, such as MCI, Sprint and Teleport. By and large, however, they competed within fairly well defined and fairly stable markets: long distance communications, value added services, etc.
Recently, however, network providers that traditionally catered to quite different clienteles, offering very distinct services, are fighting over similar turf. Importantly, this renewed blurring of competitive lines is predicated upon advanced network technologies. This can be seen in the regulatory and competitive battles between Cable TV and telephone common carriers, in the market struggles between traditional VANs, interexchange carriers, and computer makers, and in the move by major users to integrate the Internet with their private networks, thus bringing the Internet into the commercial domain of traditional carriers.
Consider one significant area of intensifying competition, the confron-tation between Cable TV operators and Telcos. The Telcos lobby for permission to provide Cable TV services (often through Cable TV acquisition) and to speed the re-build of their networks to incorporate fiber optics, at least to the "curb." Meanwhile, Tele-Communications, Inc. (TCI), the largest cable TV company in the United States, has established a joint venture with McCaw cellular, the largest U.S. cellular operator, to integrate cellular communications within its cable system, and is acquiring 49% of Teleport, the largest U.S. fiber optic bypass company. Such moves portend discontinuous change because they anticipate wholly new service opportunities that could not be effectively provided via incremental evolution of either Cable TV, cellular or Teleport. Rather, the opportunities can be pursued only by effective integration of network alternatives.
E. Policy Confusion
Finally, the current U.S. policy debate hints at discontinuity because the regulatory choices and dichotomies (monopoly vs. competition, integration vs. diversity) that adequately comprehended past network evolution are inadequate to the current moment. This is reflected in the renewed visibility for networking issues conjoined with deep policy confusion about what directions to pursue. There is a clearly perceived need at the federal level, driven largely by concerns about the international competitive position of the domestic economy, to do something to maintain the traditional U.S. leadership in networking. But there is no clear policy direction to guide the current network transformation and no obvious policy forum to discuss a coherent policy vision.
Thus, the NREN is being directly subsidized by the federal government to provide a high-speed transport backbone for advanced computing resources at the same time that the Telcos are being prevented from moving rapidly toward broadband to the home. The Modified Final Judgment strictures on information services have been relaxed, but the Cable TV restrictions remain, thus barring one of the few foreseeable services that could justify Telco investment in full broadband. But Cable TV investment in telecommunications services is permitted, and without any common carrier restrictions being imposed. One result of the confusion is a proliferation -- a portfolio -- of alternative broadband capabilities, embedded in networks that are tailored to different applica-tions and customers. The concept of universal services is suffering mightily, but the major issues that the emerging network portfolio raises, in particular the need for universal access and interconnection of the portfolio pieces, are not being addressed at all.
Growing awareness that network evolution matters to the nations competitiveness, combined with increased realization that the current telecommunications policy framework may prove inadequate to guide this evolution, provide the opportunity for substantial change in network policy. For example, as "alternative" networks such as Cable TV or the Internet become viewed as constituents of the nations telecommunications infrastructure, the different policy approaches that have governed them suggest some alternatives to traditional telecommunications policy. Such policy changes will be required to accompany the discontinuous transfor-mation of the nations network infrastructure, as we argue in the final section of the paper.
II. Emerging Approaches to the Discontinuity
Although they are seldom characterized as such, the various strategies of corporate users, carriers, Cable TV providers, the NREN and VANs represent alternative ways to prepare for, and survive, the coming dis-continuity. They all contain lessons on alternative ways to approach it and indications about possible futures. This section thus characterizes the challenge as the different major components of the network portfolio infrastructure see it, examines how they are addressing it in their evolution toward advanced network capabilities, and explores the likely evolution of each of the pieces. This is done in three steps, looking first at the leading edge of corporate use, then at the Internet, and finally at the response of traditional carriers (PSTN, Cable TV, and VANs).
A. Corporate Networks
Why should we consider private corporate networks as a piece of the emerging infrastructure portfolio? The reasons closely parallel those for our last study.
Together, the private networks of corporate America have come to represent a sizable portion of the U.S. network infrastructure. Corporate demand for network products and services profoundly shapes the telecommunications infrastructure available to the rest of the economy. Further, large corporations are finding themselves at the forefront of experimen-tation and innovation with high-speed data applications. Their internal users are inventing new ways to apply data networking technologies to existing processes, and creating new processes that take advantage of the technology possibilities. Therefore, the applications developed over private networks provide some indications of what demand for broadband applications might look like. Finally, it is instructive to examine how corporate users justify their investment in advanced networking technologies, since they operate within a set of constraints quite distinct from those of public network providers.
This section examines how corporate users are addressing the current telecommunications discontinuity. It draws upon case studies of private networks, including more particularly those of General Motors, Bank of America, and Hewlett Packard.
1. The Challenge
Over the past decade, corporate data network use has grown very rapidly. For example, total data transmission volume within one companys corporate network has jumped from 0.1 Gb/month in 1980 to 200 Gb/month in 1988, and 600 Gb/month in 1991. Moreover, this growth is accelerating dramatically: during 1991, data traffic growth reached 20% per month. Other companies are facing similar growth in data traffic.
For corporate network planners, this raises two broad categories of issues. First, if such growth is sustained, existing corporate networks will soon become unable to accommodate data traffic. The exponential character of recent growth patterns also means that it will be difficult, if not impossible, to upgrade existing corporate networks. Rather, network managers believe they will have to deploy new data networks.
Second, growth rates vary drastically across applications because companies telecommunications are shifting. In manufacturing companies, for example, network use has been re-focused away from administrative tasks and marketing towards engineering and manufacturing tasks. As a result, the networks they build must remain flexible enough to adapt to new needs as they arise. The network requirements associated with different applications vary widely. For instance, administrative applications may simply require the ability to ship large files, such as payroll data, overnight or to consult an employees record. Such applications can easily be accommodated within current networks. By contrast, engineering and manufacturing applications may involve the interconnection of sophisticated CAD and CAM systems, requiring very rapid transmission of vast amounts of data. Other characteristics of the applications (such as the number of terminals involved, the interactivity, etc.) will also differ. The emerging applications will therefore impose very different technical requirements on corporate networks.
Companies typically estimate that it takes about 3-5 years to deploy a new corporate-wide information backbone, and many have started drawing up plans. Individual approaches vary. Some companies simply ask the telecommunications manager to come up with estimates of future network needs and plans for network upgrade. Others take a much more comprehensive approach. Some of the companies we surveyed have undertaken corporate-wide assessments of their networking needs and their evolution. Such efforts typically involve asking representatives from various parts of the company (ranging from sales and marketing to manufacturing), to predict what their divisions/groups future networking needs would be. Rather than simply extrapolating current growth trends, they were asked to think through the details of how their activities were likely to change. The exercise often is iterative: the goal is not simply to imagine future activities and deduct new network requirements but also to take into account emerging networking technologies and see how they could be used to transform existing processes.
This section consolidates the results of such planning efforts within various companies. Its goal is to show how changes in network use are linked to transformations in business processes and to explore the impli-cations on corporate networking. In the next sub-section, we analyze trends in corporate information networking. Then we explore the implications for corporate network infrastructures, and examine how companies view and evaluate their options.
2. Trends in Corporate Information Networking
The evolution of corporate networking needs is predicated on three fundamental business trends: globalization, shortening cycles and increasing inter-firm relationships. Of course, these trends affect firms differently, according to their economic activity, but all three were prominent among the reasons driving network change among our case studies.
"Globalization" is used by many companies to cover a variety of phenomena. With respect to networking, however, three related meanings of the term appear especially relevant. First, it reflects the fact that U.S. companies are facing simultaneously increasing competition from foreign firms in their home markets, and are aiming to conduct a growing share of their business out of the United States (including increasing international sales, international design and manufacturing facilities, support organizations,...). Second, it reflects their need to draw on dispersed resources. Design, production, marketing or service facilities involved in the making and selling of each product are increasingly scattered around nationally and internationally. This is partly to use the corporate resources better (involve the best design teams in each area), partly for closeness to local conditions (design products where theyll be manufactured or sold), partly to address local content requirements. Finally, it means that companies feel a growing need for interchangeability and replicability among these dispersed units, in order to address two issues. On the one hand, the production processes that they have designed and fine-tuned in one place must be replicable in other regions. On the other hand, interchangeability helps them to stimulate internal competition and create redundancy-based security through internal double-sourcing.
"Shorter cycles" have become a critical basis for competitiveness in all economic sectors. Companies in all economic sectors are making efforts to shorten the various cycles underlying their activities: concept-to-product, order-to-delivery, trouble report-to-repair. They view high-speed data networking throughout their operations as a critical part of the re-organization process that will enable them to achieve this goal, despite increased geographical dispersion.
Finally, companies increasingly find that their success rests upon carefully built linkages with other firms, within what have been described as network-organizations. Such linkages include the partnerships they build with their suppliers and business partners, with their customers, or with the sales and distribution channels they rely upon.
These fundamental trends have a number of implications for network use. First, they create a need for denser, more flexible networking within the company (between and across divisions). This is necessary to support increases in company-wide distributed projects. More cross-functional, cross-division, problem-centered teams/workgroups. These seldom are made up of people who all work in the same locations. Second, they also require better links with the outside through public networks. These become necessary to support increased collaboration with outsiders. Whether for product design and development or marketing and distribution, partners increasingly need the same access to corporate information networking resources as the companys own employees. Typical applications in this area include remote access to technical databases, expert advice, and trouble-shooting. In addition, public networks are becoming important for employees who access information from non-company locations, whether on travel, working at home, or on customer premises.
The combined needs for better internal networks and better public extensions of these private networks corresponds to changing patterns of network use. First among these emerging patterns is the move towards client/server environments: initially confined to Local Area Networks, this new computing "paradigm" will soon extend to private wide area networks, and will require highly capable connections between private and public networks. In conjunction with the emergence of this new architecture (and enabled by it), companies are experiencing the rise of new applications, in particular the growing importance of cross-functional information systems (supporting variable-geometry "virtual workgroups"), of image storage and retrieval systems, and of software distribution mechanisms (broadcast, multicast, or narrowcast). These growing network uses are driving the need for low-delay, high-throughput networks.
The move towards client/server environments has been described as a paradigm shift from a model where wide-area networks provided connectivity between data centers (with a local concentration of computers and terminals) to a model where every desktop is an end-point in the network. Terminals, PCs and workstations are being connected directly to LANs, which in turn are interconnected via high-speed, company-wide networks. Local interconnection of PCs and terminals through LANs (typically 10mbs 802.3 Ethernets) is now widespread. LAN interconnection through FDDI site backbones, then either through bridge/outer/T1s or through the Internet is progressing. As interconnection grows, traffic is skyrocketing.
This relates to two major trends in how computers are interconnected and used. The first is the spectacular decentralization of computing power. Dumb terminals are being replaced by intelligent PCs and workstations, increasingly connected to site LANs. Workstations are now able to handle sophisticated applications once reserved to mainframes (e.g., CAD). This decentralization of computing power drives a second trend towards the centralization of information (or at least of information management). Because so many workstations have now become capable of manipulating data that once could only be stored -- and modified -- within centralized mainframes, there is an acute need to coordinate and consolidate the various databases and applications they handle. For example, it would not be a good idea to have many different versions of the CAD model for a same component being manipulated and modified independently by many workstations.
This implies the need to maintain centralized and widely accessible information resources in servers. Clients then simply "check out" the data, process or display it locally, while letting the network know of their activities. Examples include the centralized database maintained on a single server that may be accessed simultaneously by over 1000 clients worldwide over a wide area network, such as the components catalog, IC design libraries, or trouble-shooting expert systems of computer manufacturers. As this new model becomes the general rule, clients will need high-speed access to the servers, so that it looks to them as if they were accessing local information (LAN-like response times).
3. Network Implications
Far from decreasing the need for network links, the decentralization of computer power in fact requires networks with increasing bandwidth. High bandwidth becomes necessary not so much to transmit large amounts of information (although some of the files are quite large) but most importantly to obtain very short response times. The migration from terminals connected to data switches, to PCs and Workstations connected to a LAN/WAN dramatically increases the potential for each employee to send data through the net. Typically, this ability jumped from 1.2, 2.4 or 9.6 kb/s with terminals or modemed PCs, to 10mb/s with todays Ethernet LANs, to possibly 100mb/s within the next 5 years (with FDDI or DQDB). The number of sites and people they can exchange data with has similarly grown tremendously from a few people in nearby offices, to anybody within the company and beyond, to the increasing number of outsiders connected with the company.
As telecommunications became increasingly deregulated, many U.S. companies have deployed private networks to handle their growing networking needs. These strategies were usually driven by the inability of public networks to handle corporate requirements and by companies strategic need to control their telecommunications. Generally, however, they would prefer to use a public network if one was available that satisfies their requirements for capabilities and control. Indeed, many companies that once deployed private voice networks returned to the public network once services like Software Defined Networks (SDN) and Virtual Private Networks (VPN) became available.
For their data networking needs however, corporate users remain unsatisfied with the public networking options available, and many continue to maintain and use private networks. They are monitoring the most recent offerings of public networks, including SMDS (Switched Multimegabit Data Service) and Frame Relay, but often remain skeptical that these will allow them to use the public networks for their data traffic in the near future. They are exploring a number of paths as they contem-plate the deployment of the corporate network infrastructure that will support these exchanges, including the continued deployment of private routers and leased lines and the use of Value Added Network Service providers.
The migration paths they are choosing for the transition from their existing network to the high-speed replacements they envision are instructive. One computer manufacturer chose for example to phase out completely its private X.25 network, until now used for the majority of corporate wide-area transactions, and replace it with a private Internet based on the TCP/IP protocol. Among the reasons cited for this choice were the need for an infrastructure better able to support increasingly interactive applications, as well as the easy access this allowed to the rest of the Internet community, either through the public NSFNet and its regional networks, or through private carriers such as PSI. Interestingly, such an approach to the provision of its internal network is closely related to the ideas implemented throughout the Internet. In particular, it seeks to encourage network use to foster experimentation and innovation in the belief that employees cant realize an applications worth to them until they have tried it. They are less likely to "push the limits" if they need to justify every cent they spend. Many of the most innovative network uses -- and potentially the most productive -- have resulted from such experimentation.
Other companies we surveyed have been much more concerned with establishing detailed "charge-back" schemes to redistribute their net-working costs among their internal users in an effort to monitor and contain corporate use of telecommunications and control the associated costs. While such approaches force individual users to justify closely their networking use, and thereby keep the companys telecom costs in check, they tend to stifle experimentation, even sometimes reduce telecommunications use. Importantly, they often make it difficult for corporate divisions to embrace the use of new network systems when the benefits would accrue to other divisions or to the company as a whole. The contrast between these two approaches echoes some of the current telecom policy debate: as we seek to move closer to a situation where telecom users only pay for what they use, we risk missing out on some of the more systemic benefits of advanced networking. One of the essential policy tasks will be to strike a balance and design mechanisms that make it possible to mobilize innovative demand by some users for the benefit of all.
Section III of this report will focus on the policy implications of the transformation affecting corporate networking. At this point, however, it is worth emphasizing some of the salient features of this change. First, the evolution of corporate networking suggests discontinuity because it stems from a profound reorganization of corporate activities around the networks supporting them. That reorganization, enabled by the new networks possibilities, in turn places substantially new demands upon the networks. Second, the network applications associated with this change, grounded in hands-on experimentation, are now suggesting real users and real needs for advanced network capabilities. They also demonstrate that network experimentation constitutes a powerful source of innovation, one that needs to be effectively harnessed to promote the transformation of the nations network infrastructure. Third, emerging corporate network uses are showing the limits of purely private networking. Because business reorganization reached beyond the borders of individual corporations to involve growing numbers of suppliers, clients, partners or subcontractors, there will be a need (and a demand) for public networks able to match the capabilities of the private networks individual corporations have built for themselves. Frustrated by the limitations of the public network, corporate users are increasingly willing to play a part in helping the modernization of the public network, provided an adequate policy framework.
B. The Internet
The Internet is a loosely organized system of interconnected computer networks, which primarily serves the research and education community. Its large and growing community of sophisticated users, the diversity of applications and uses it fosters, and the trials now underway within the gigabit testbeds make the Internet one of the boldest real-life experiments in advanced networking today. Three facets of the Internet story are especially relevant.
First, data traffic over the Internet is growing explosively. During 1991, traffic on the Internet has increased by an average 20% per month.5 This translates into a nine-fold increase for that year. If sustained over a two-year period, traffic will increase by about 80 times over two years, 700 times over three years and more than 6000 times over four years.
In fact, this is likely to underestimate traffic growth on the NSFNet backbone since the growth rate itself has increased significantly compared to past years, and shows every sign of increasing again in the future.5 In order to cope with this traffic growth, the NSFNet backbone was recently upgraded from T1 (1.54 Mb/s) to T3 (45 Mb/s) speed. At stake in the current policy discussions of a National Research and Education Network (NREN) is the next phase upgrade to gigabit speeds (roughly 50 times the current T3 speed). This next upgrade is planned for the mid-1990s, if financing is made available.
Second, this growth in Internet traffic has been fueled partly by increased use among existing users but also very significantly by the addition of new users. Initially, the Arpanet -- the precursor of the Internet -- was intended only for the military and military contractors. As it became open to civilian research uses, and placed under the responsibility of the National Science Foundation (NSF), the NSFNets primary users were the academic "elite" of advanced computer scientists and researchers. More recently, two new communities of users have claimed access to the Internet: private corporations are now starting to use the Internet as a Wide Area Network (WAN) to interconnect their Local Area Networks (LANs), and the broader academic community -- including for example academic libraries as well as K-12 schools -- is making increasing use of the Internet.
As a result, the character of the Internets user community changes significantly over time. The character of network use, the kind of applications carried over the network are changing accordingly. Interactive applications now constitute the fastest growing segment among the applications carried by the Internet.6
Significantly, the reason for this widespread enthusiasm is not that the Internet is the best high-speed data network there is. In fact, the networks protocol, TCP/IP, is now quite old and probably much less efficient than newer approaches to high-speed date networking such as Frame Relay or SMDS. But its success stems from the fact that the Internet is often today the only possible outlet for eager users that offers a standardized and stable interface, along with a deliberate focus on openness and interconnection. While these entail problems (see the various worms and viruses...) they make the Internet extremely attractive for very different groups of users, ranging from corporations to academic institutions. By contrast with traditional telecommunications networks, the rules governing the Internet focus on interconnection and interoperability, rather than attempting to closely define the types of applications that are allowed or the rate of return permitted to its constituents.
Further, as a result of its widening use, the Internet is becoming a very fertile experimental ground for high-speed networking applications. Here again, we find it extremely interesting that the most successful "broad-band" applications now supported by the Internet differ significantly from those most prominent in the public debate: for example, they involve cooperative computing (whether by libraries doing cooperative cataloguing or corporations linking CAD workstations) rather than video transmission.
Third, the policy environment surrounding the evolution of the Internet is dramatically different from the traditional telecom policy environment. The policy goals being discussed, the policy mechanisms that have permitted the growth of the Internet and that are now envisioned to guide its future, are actually quite foreign to the current telecommunications debate. One of the important factors in the success of the Internet comes from the "governance mechanisms" (rather than regulation) that guide its use and evolution, in particular, the direct focus on inter-connection and interoperability among the various constituent networks. Further, while not always explicit, cross-subsidies of all kinds -- in particular of the kind that probably would no longer be tolerated within the public telecom networks -- pervade the Internet. These include the initial military subsi-dies that supported the Internets basic technology development and the public subsidies channeled by the NSF into the exploration of high-speed applications. Current discussions about the conditions under which for-profit use of the Internet by private companies should be allowed -- the so-called "privatization of the Internet" -- in effect explore new possibilities for cross-subsidizing Internet use between various categories of users.
The meaning of these mechanisms goes beyond simple transfers of money. Indeed, what is at issue here is to find ways for various com-munities of users to share a common network infrastructure in ways that are mutually beneficial. The success of the Internet largely rests upon its ability to combine different visions (ranging from DARPA military research objectives at its inception to the goals of academic users and the wider needs of the K-12 community). Its development could therefore draw upon a variety of rationales: the Internet has been a tool for government agencies, a basic research prototype for scientists designing fast packet switches or optical transmission devices, a knowledge infra-structure that supports education and libraries, or a tool to promote more universal access to information services in the K-12 community.7
The Internet experience holds fascinating lessons, as we elaborate a policy framework for the current transition. At stake is the deployment of a network infrastructure able to support joint experimentation and learning among various users and able to serve as a conduit for the diffusion of network-related innovations across various user communities. The significance of the Internet as experiment lies not simply in the technologies it helps develop, but -- more importantly -- in the new usage dynamics it helps uncover, the new network management mechanisms it tests, and in the new policy strategies it forces us to explore.
C. Traditional Carrier Networks -- Telcos, Cable TV, VANs
From their wildly disparate starting points, the broad variety of telecommunications carriers in the United States are managing the transition to advanced networks in very different ways. The major players include ATT and its principal interexchange (IEC) competitors like MCI and U.S. Sprint, traditional suppliers of local access and exchange services like the Bell companies (BOCs) and GTE, value-added network suppliers (VANs) like BTNA, GEIS and EDS, CATV suppliers like Cox and Viacom, and alternative fiber-based access providers (AAP) like Teleport.8 As table 2 suggests, each carrier type has different strengths and weaknesses in confronting the transition. While each is moving toward advanced capabilities, they are doing so at different speeds, usually with a different mix of available technologies.
By virtue of past investment, strategy and regulation, each network type is tailored for different market segments -- e.g., long-distance voice (IEC), local voice (BOCs), business data (VANs), residential entertain-ment (CATV), and local business access (AAP). Each therefore also follows an evolutionary logic rooted in the core business and associated capabilities of the past. Yet, more and more, traditional competitive boundaries blur as common markets emerge for the provision of advanced networking capabilities. Each carriers network choices are strongly path dependent on past decisions that open different probabilities for successfully exploiting the common set of future opportunities.
Table 2: U.S. Carrier Network Competitive Position
|Advanced, increasingly broadband platform but dependent on local networks for access|
|BOC||Narrowband||Ubiquitous local access network but lagging
Mix of old/new tech. technologically and still heavily regulated
Tailored for business services
|Cream-skimming position and tailored intelligent services, but carrier access dependent and less ubiquitous|
|Ubiquitous local broadband capability but not switched
Regionally fragmented; tailored to passive entertainment
|AAP||Digital fiber||Advanced, local cream-skimming, but geographically
restricted to few major urban locations
This section describes the dynamics of change confronting the major carriers from the perspective of the most traditional and least prepared of the players, the local telephone companies (hereafter abbreviated as: "RBOCs" or "Telcos").9 That choice of organization is made in part because the Telcos face direct competition from all of the other carriers. Their dilemma conveniently permits us to frame the issues facing everyone. But there is a more significant reason. By virtue of their ubiquity, universality and bottleneck position, the local Telcos will remain a critical piece of any coherent national communications infrastructure. Because they are saddled with the most costly, slowest to evolve piece of the national infrastructure, they represent the limiting constraint on the national evolution to advanced network capabilities. Therefore, the pace and character of overall infrastructure evolution in the United States will be largely determined by the speed and choices of the local Telcos in trans-forming their networks: The decisions they make may well determine how coherent, advanced and universal the infrastructure finally becomes.
1. The Current Competitive Position
The Bell companies face a remarkably tough transition. Their major strength, the ubiquity of their narrowband network, confronts several major vulnerabilities. These include regulatory control over the speed and extent of adoption of new technologies, proliferating competition from the interexchange carriers, alternative access providers and VANs in areas where new business services could pay for network upgrade, and an alternative infrastructure (Cable TV) dominating the one service (enter-tainment video) that might justify broadband to the home. Table 3 lays out the competitive terrain.
Table 3: Telco Competitors
|Intelligent Services||IECs, VANs, Privates|
|High-Speed Data||Privates, IECs|
|Broadband Access||AAPs, CATV|
|Broadband Switching||IECs, VANs, AAPs|
Key: POTS = Plain Old Telephone Service; IECs = Interexchange Carriers; AAPs =Alternative Access Providers, VANs = Value-Added Network providers; privates = private networks
Torn between the responsibilities imposed by regulation and the need to respond to competition in almost every area of advanced service provision, the local U.S. Telcos have exhibited great diversity and little coherence. Their relative confusion and the rapid advance of the other networks in the infrastructure portfolio call into question whether the Telcos will ever be capable of playing the integrative back-bone role for which their networks could be uniquely qualified.
The current competitive position of the regional Bell Companies underscores this dilemma. Relative to their major competitors, they are laggards. With variation among them, they have been slow to adopt Signaling System 7, the first step toward offering enhanced intelligent services, and slow to adopt broadband transmission, the key to offering advanced broadband services to business users and interexchange carriers. Table 4 compares the RBOCs to the interexchange carriers along these dimensions, while Table 5 compares RBOC fiber deployment, showing the great variation between the companies as a result of strategy and available constraints (including variations in local regulation).
Table 4: Network Status of Key Players (as of January 1, 1992)
% Digital switches
% digital IOF
% Fiber IOF transport
Key: IOF = Inter-Office Feeder; SS7 = Signaling System #7;
SCP = Service Control Point
Source: Estimates based on industry conversations
Table 5: RBOC Fiber Deployment
Fiber-miles per thousand access lines
Aggregate Fiber Invest-ment per Access
Aggre-gate Fiber Invest-ment
|Southwest- ern Bell||38.3||41.43||270,300||352,300||453,000||11,818||489.6|
Source: Column 1 and 2 calculated from columns 3-7, taken from Tables 2, 6, 9, and 12 of "Fiber Deployment Update," End of Year 1991, Industry Analysis Division, FCC, March 1992.
Part of the problem is an extremely conservative investment philosophy, which justifies investment almost exclusively on the basis of up-front cost savings rather than anticipated new revenue streams. But even so, their basic network is in need of upgrade, particularly of the operating systems (management and control), which are keys to offering stable, low-cost, diverse services. Indeed, industry estimates suggest that there is as much as a 20% difference between RBOC per-line operating costs and those of the lowest cost alternatives -- a result in equal parts of less efficient operations and much higher overhead.10 These troubles are also reflected in comparatively slow response and new service introduc-tion times. For example, MCIs maximum target time to roll out new services is roughly equivalent to the Telcos current minimum. 11
Higher costs and slower technical change are not the only reasons the RBOCs are competitively vulnerable. Their income structure combines with regulation to make them particularly susceptible to competitive entry. Table 6 estimates the percentage of RBOC revenues coming from their major categories of service provision. About half of the RBOC revenue stream comes from charges imposed by regulation (to replace the settle-ments procedure that used to provide cross subsidies to local service) and is paid by customers for the privilege of accessing the local Telco network. Another third comes from local exchange services, of which the most profitable segments by far are directory, intra-LATA toll and a few special business services (e.g., 800 service).
Table 6: RBOC Revenues by Service Category (%)
|Local Exchange Services||30%-35%|
Source: BRIE estimates based on FCC dataThe bulk of access charges and profitable services (and approximately half of total revenues) are accounted for by the small percentage of customers who are either major businesses or interexchange carriers.
The concentration of profitable services and major customers invites the entry of competitors. By tailoring their networks to provide local access to major customers (i.e., the strategy of AAPs) or to target profitable business services (i.e., IECs, VANs), new entrants can cream-skim with impunity since they need not carry the costly burden of providing the rest of local services to the majority of customers. Regulatory decisions have increasingly encouraged this kind of competition -- most recently in September, 1992, with the FCCs decision further pushing competition in local exchange services by requiring that the Telcos provide similar access for local competitors as they provide for IECs. And emerging advanced revenue opportunities, like Advanced Intelligent Network (AIN) services and broadband transmission and switching, appear certain to be subject to the same logic. Perhaps a full third of broadband access revenues, a quarter of intra-LATA toll revenues, and 10-15% of total Telco exchange revenues are at risk over the next decade from such cream-skimming competition.12
Since a majority of Telco costs are more or less fixed, associated with maintaining the local exchange network, competition in the profitable service segments will likely have the perverse impact of forcing local subscribers to bear rising prices for basic services: competition will erode margins on profitable services and reduce the base of the largest customers. That will force fixed costs to be spread across fewer customers and to be recovered by raising prices (subject, of course, to local regulatory approval) on unprofitable services where there is little or no competition. Ultimately, rising prices may invite new entry even in those services.
In fact, the RBOCs require major network investment in the near term just to defend their existing market positions, let alone to provide new services. Relative to customer demands and the best competition, significant investment is needed to rehab and stabilize the core network, to adopt new technologies that lower the cost and increase the reliability of the network, and to develop the management and control systems (support systems) that would permit the Telcos to be more flexible in services, billing and customer responsiveness. Such investments could generate significant cost savings and help to position the Telcos to provide new services, but existing capital and expense budgets appear less than adequate to accomplish the necessary investment in a competitive time frame. Indeed, industry estimates of the investment shortfalls needed for defensive purposes run to several hundred million dollars per Telco.13
Meanwhile, the full costs of changing out the narrowband network to competitively provide advanced services are enormous, the more so if they can not be carried in the ratebase. Here too, a minimum of several additional hundreds of millions of dollars of annual investment per Telco are required to attack ISDN, AIN and broadband opportunities. But the Telcos own discounted cash flow analyses suggest that there is little prospect of a positive payback this century from investment to provide these new services.14 Emerging competition is forcing the Telcos toward advanced networks as the only viable response, but the sheer expense and regulation may well prevent them from getting there in a timely fashion.
2. The Business Services Market
BOC problems are compounded by the rapid moves of other carriers into advanced services, particularly in the local business services market. There are two principal categories of advanced local services, advanced intelligent services (as defined by Bellcores AIN concept) and broadband access and switched services.15 Here we survey the prospects for each in turn.
AIN is a set of network specifications developed by Bellcore to de-couple development and provision of new service capability from switch software.16 Typically, AIN adds software and computers (called Service Control Points or SCPs) to the public phone network, connecting them via Signaling System 7. A predefined set of triggers -- e.g., hook flash, busy signal, specific number -- indicate AIN handling by SCPs capable of providing an array of intelligent services.17 New services can then be created by adding new SCPs and writing SCP software without having to rewrite switch software. In essence, much as the computer world is moving toward client-server architectures, AIN creates a more distributed architecture of the current phone network, with intelligence decentralized from switches where it largely now resides.
Achieving AIN requires upgrading the current network (e.g., full SS7 implementation is necessary), changing its architecture, and significantly re-tailoring its management and control systems. That implies at best a gradual roll-out to dense LATAs as SS7 is fully implemented by the mid-late 1990s. Bellcore then estimates the potential incremental revenue gain for an RBOC from implementation of AIN at $1-3 billion around the turn of the century. Other industry projections are less optimistic, anticipating only a tenth of the Bellcore projections. Whatever the revenue gains, AIN will not be provided universally before well into the next century.
More important than incremental revenues, however, AIN may be necessary just to permit the Telcos to achieve the flexibility -- both applications and configuration flexibility -- necessary to compete in advanced business services with IECs, VANs, and even CATV.18 For example, AIN would provide an ideally flexible and capable backbone, and the management and control support, for personal communications network providers (PCNs), who are likely otherwise to build their own networks or ally with alternative local distribution options like CATV (more about which, below).
Although the competitive pressures surrounding AIN are real, they pale in comparison to those of the business broadband market. From the Telco perspective, business broadband includes two principal opportunities, providing broadband access to other carriers and large customers (who would do the switching themselves), and providing switched broadband services. The prerequisite for access is optical fiber in the feeder and inter-office portions of the network.19 Switched services would eventually be based on fast packet technology in the Asynchronous Transfer Mode (ATM), although Frame Relay and SMDS are already-available precursors. Switched services would thus add ATM switches and SONET electronics to the fiber plant. Switched broadband could permit Telcos to offer the kinds of services, like interactive CAD/CAM, that are being developed over private networks.
While most BOCs are deploying fiber rings in dense urban business areas served by high-capacity central offices, the vast majority of T1 lines originating in those offices normally terminate in switches that have no local fiber distribution. Moreover, as a percentage of their total networks, the Bell Companies have put virtually none of the necessary infrastructure for broadband switched services like SMDS or Frame Relay into place. In essence, the local Telcos are capable of providing limited broadband transport in major city business centers, but lack end-end digital broad-band connectivity. While they can provide limited broadband access to large customers and IECs -- industry estimates of exactly how limited range from 10%-40% of major business centers -- they are incapable of providing any switched broadband services.
Meanwhile metropolitan fiber nets like Teleport (now in 20 major U.S. cities) have already wired local business concentrations, providing alternative access for businesses and carriers (whether interexchange or VANs) that is generally cheaper, more reliable and more responsive than the Bell companies -- in some cases, by considerable margins. Almost all major urban centers in the United States now have at least one AAP. California is representative, as table 7 suggests. AAPs appear to have garnered up to a third of the access market in many cities, mostly as a result of serving the access needs of the IECs.20 And now, with the recent FCC decision on local exchange services, the AAP networks may be interconnected through the Telcos to provide direct competition in the most lucrative switched services to local businesses. To be sure, APP success is based on a pure cream-skimming strategy; but that only makes the competitive threat more daunting, aimed directly at the profitable services that could otherwise pay for upgrade of the overall Telco network.
Table 7: California Alternative Access Providers
|Los Angeles||MFS, TCG|
|San Francisco||MFS, TCG|
|San Jose||MFS, TCG|
|San Diego||Linkatel, TCG|
Key: MFS = Metropolitan Fiber Systems; TCG = Teleport Communications Group
Source: Trade journal reports
To combat the competition, the Telcos have been speeding up roll out of fiber and the associated electronics and management/control supports, at least to dense urban centers, with BellSouth, Nynex, and BellAtlantic the most aggressive. (Recall table 5.) Universal extension of broadband access, or even extension beyond major urban areas, would seem to be too costly to contemplate in the near term.
In switched services, the interexchange carriers like ATT, and traditional VANs (like EDS) have been impinging on the local Bell company business services market even as they fight between themselves for more national or global opportunities. The basic strategy is to differentiate themselves with enhanced management capabilities and industry-specific applications. On the one hand, this means a race to deploy the new variety of "fast-packet" alternatives like SMDS or Frame Relay to capture "bursty" high-capacity opportunities like cooperative computing, to provide bandwidth on demand to financial or medical services, and to carry bulk transfers between distributed LANs. Indeed, ATT, Sprint and at least six major VANs already offer frame relay. On the other hand, there is pressure to focus and specialize, to differentiate by tailoring services to vertical industry segments in transportation or distribution.
In effect, there appears to be a segmentation emerging in the market between those providers like ATT and BTNA (British Telecom North America) positioning themselves to become network integrators offering user-customized services and end-end global network management and the rest, who seek refuge in specialized business applications niches.21 The IECs have several significant assets to draw upon. These include sophisticated billing and accounting resources that can be used strategically to define and serve major customer usage patterns, virtual private network (VPN) platforms, and, of course, increasingly global connectivity. These network resources permit the IECs to assemble long-term, customer-tailored packages of managed network capabilities (like those ATT has under Tariff 12), which they hope will lock-in customers over time. Such packages provide a reasonably stable revenue base from which to launch new services and network upgrades -- i.e., exactly the stable opportunity that is denied to the local Telcos, who lack the capabilities and the regulatory permission. (We return to this point in the Conclusion.)
With a few exceptions (e.g., EDS, BTNA), the traditional VANs find their competitive opportunities somewhat more constrained by their historical role and evolution. Most provided data processing services to vertical markets (e.g., EDS started by processing Medicare/Medicaid transactions). As a consequence of the bewildering array of proprietary standards in computing, most also developed network management expertise in converting and interconnecting disparate protocols. As they add broadband capabilities and intelligence to their networks, they should be able to tailor services both to the vertical markets they have served for some time and to specific network management opportunities that fall within their traditional domains of competence. The latter might include distributed database services and management of light, sporadic, multi-protocol data traffic. An eventual evolution toward true ATM networks would, however, push VANs providers back toward integrated voice services, bringing them out of their niches and more fully into competition with the IECs and Telcos. Some regional VANs, like TC Systems, already offer centrex and ISDN services over their existing networks. Like the other network providers, the possible evolutionary paths to exploit those opportunities are open-ended.
Over time, it is conceivable that the major IEC and VAN business carriers can effectively re-integrate the private network fragments which embrace multi-vendor equipment and services in a distributed environment, exploiting significant economies of scale and scope involved in assembling and applying network management expertise and in tailoring applications for users. In so doing, they will likely siphon off significant service revenues from local Telcos, in effect by-passing the local network with direct connections to those major local business customers involved in global business operations. Needless to say, the Telcos face several daunting obstacles in responding effectively.
First and perhaps most fundamental, BOCs must transform their service delivery strategies from those that address a rather uniform mass-market to those geared toward increasingly sophisticated and specialized needs of an increasingly segmented market. BOCs will have to forsake the more traditional arms-length relationship between public utility provider and user to pursue collaborative solutions (with customers and 3rd party vendors).22
As suggested above, none of this seems likely to happen fast, not least because of the expense of providing switched broadband conjoined with regulatory universal service obligations. It would be surprising if the Telcos were to capture more than 10%-15% of the emerging switched broadband services market before the next century unfolds in depth.
3. Residential Video
The local Bell companies also face an entrenched competitor in Cable TV in the one residential service that might justify broadband to the home, residential provision of video/TV entertainment. With service to over 60% of homes, and with over 90% of U.S. residences passed, Cable TV provides a full alternative broadband distribution option -- albeit one that at the moment is primarily a non-switched, co-ax, tree and branch, one-way network. The costs of re-fitting the entire Cable TV infrastructure to carry switched communications services are enormous ($30-$290 per subscriber) but probably an order of magnitude less than the converse, i.e., extending the local PSTN via fiber optics and video electronics all the way to the home ($1,500-$3,000 per subscriber).23 The latter investment hardly appears a feasible option at the moment. Industry estimates suggest that wiring just high density residential concentrations offers a huge negative payback for the foreseeable future, decades into the next century.
Meanwhile, the Cable TV industry is not standing still. The evolution of the Cable TV networks is being fundamentally driven by a single application, provision of entertainment TV/video to the home. All the network upgrades in progress have a single major rationale, expansion of channel capacity to 150 channels by putting fiber in the trunk portion of the network (an incremental investment comprising only 19% of total plant capital costs).24 With that many channels, video-on-demand can effectively be delivered without switching. Add switching in the head-end, and multiple packages of perhaps 1000 channel "events" are available, 150 at a time. Digitize a few of the channels, and two-way communication is possible, broadband to the home and narrowband back. With those kinds of future options, Cable TV providers can ward off the long-range threat of local Telco entry to their market. In that sense, most of their latest moves into other communications services should be understood as "BOC-repellent."
Indeed, both regulatory barriers and competitive considerations currently prevent realization of the least-cost alternative for switched broadband to the home -- interconnecting the Telco feeder network with the Cable TV local distribution network of broadband co-ax to the home.25 Still, the Cable TV firms claim to have an ideal architecture to link with PCN networks and thereby achieve alternative local distribution of telecom services, or with metropolitan fiber nets to provide competition to local Telcos in the business services market -- the core of the Telco revenue stream (counting access services). Forays in these directions are just beginning. For example, McCaw/TCI in Oregon, Cox in California, and Comcast in Philly are all using local CATV as backbone to link cell cites, while TCI, Cox and Time Warner, among others, have linked with AAPs.
Again, the precise evolutionary paths to an advanced local communi-cations infrastructure appear to be wide open and need to be tested through experimentation and learning. Nonetheless, the very existence of this only comprehensive, nation-wide, non-Telco, local distribution network provides real opportunity for Cable TV firms to move from provision of entertainment to provision of infrastructure. Already, a panoply of non-entertainment services are selectively available over the Cable TV distribution network, ranging from educational services to telecommuting -- although the former only exist because local munici-palities have demanded them in local franchise agreements. The central question is whether Cable TV providers are in a position to permit the kind of experimentation and learning necessary to effectively develop the range of services that they could provide as an alternative advanced infrastructure. If not, their choices will be driven by alliances with other network providers, and are likely to result not in a fully competent infrastructure, but an entertainment-based network with some cream-skimming business opportunities sitting on top.
4. Telco Choices
Overall, as indicated above, the responsive choices of the Bell companies are complicated because, given the competitive constraints and limited resources, they can not choose to pursue all advanced capabilities effectively and simultaneously. Indeed, at least in the short-term, they seem to face a significant choice between pursuing advanced intelligence and pursuing broadband transmission opportunities. While both sets of services would benefit from basic network upgrades, AIN could be pursued independently of broadband and vice versa -- expenditures to achieve one do not lead incrementally toward the other. Moreover, the associated management and control systems are likely to be significantly different, further (and expensively) complicating the Telcos situation. The dichotomy can be seen in the disparate strategies of the different RBOCs. Some, like BellSouth, seem more committed to the pursuit of broadband, while others, like BellAtlantic, seem to be aiming more for intelligent services.
But neglecting one opportunity for the other may also be a strategy for failure. For example, spending to implement Bellcores Advanced Intelligent Network concept could permanently sacrifice provision of broadband transmission services to business users because it would give the first-mover advantage to well-situated, cream-skimming competition. Nor does pursuit of either option seem to lead to the universal extension of the one capability throughout the network. For example, incremental moves toward broadband for businesses do not appear to be an effective strategy for reaching broadband to the home because there is almost no geographic overlap between business central offices and residential ones.
To be sure, under the enduring press of competition, the local Telcos are not standing still. As indicated above, incremental investment is occurring to upgrade their service platforms, as are exploratory invest-ments toward advanced services. For example, over 100 fiber-to-the-curb or fiber-to-the-home broadband trials are being pursued, although most of these are aimed at proving-out fiber technology for plain old telephone service. Only if the costs can be justified on that basis are local regulators likely to permit the Telcos to include the enormous investment for residential broadband in the rate-base. As much as anything else, that regulatory logic conditions the competitive strategy of the Bell companies.
As the dynamics play out, it appears that the local Telcos face a dichotomous future,
post-discontinuity. They could end up being the network of last resort for all those small
businesses and residences that cannot afford, or are not served by, the more specialized
pieces of the infrastructure portfolio. Or, they could become an integrated transport
backbone, providing the universal access and interconnectivity that the network portfolio
needs to become a true infrastructure. Dramatic changes in both strategy and policy are
necessary to achieve the latter outcome.
We have argued throughout that discontinuous change rather than a simple incremental evolution is characterizing the transformation of the network infrastructure in the United States. Driven by actual users and their initial applications, there are multiple networks representing different technological trajectories emerging in the face of discontinuous change -- a prudent bet-spreading by the market under conditions of extreme uncertainty. Each network choice carries with it the potential of strong path dependencies; i.e., networks optimized for advanced business services or for network management of cooperative computing may not be appropriate for residential broadband. Moreover, for each alternative path, there is no obvious moment when demand will "take-off" to justify the associated investment risks. Rather, the evidence strongly suggests that takeoff is only likely after extensive experimentation and learning.
In fact, given the uncertainties, it is impossible to predict demand, appropriate levels of investment, or the ultimate requirements these will place on the network infrastructure.26 While this is manageable when network evolution is incremental, it is much less so when change is rapid and discontinuous. Under these conditions, misallocation of effort and investment must occur, although the inappropriate choices will only become apparent in the revealing gaze of hindsight. In part, it is this difficulty that has convinced U.S. regulators to abandon policy in favor of the market. But, as we argue below, that decision is not likely to be any more effective in dealing with discontinuous change. Markets are lovely vehicles for creating equilibria under known conditions, but are not particularly effective mechanisms for dealing with drastic change under conditions of extreme uncertainty.
A. Infrastructure as Network Portfolio
The inability to predict, however, does not imply the kind of inability to plan that requires relinquishing all choice to the market. Rather, policy must acknowledge the discontinuity and recognize that the boundary con-straints on policy have changed. Multiple networks exploring a range of technological alternatives should be encouraged, given the uncertainties. Long-term interconnection and access between the networks is necessary if the total system is to comprise a true infrastructure. Incremental moves must not sacrifice long-term potential. Potential evolutionary paths must be explored through experimentation and learning until alternatives and tradeoffs are clear. Open architectures and maximum flexibility would appear to be a necessity in an uncertain future.
The choices for policy become clearer when arrayed against those that dominate the current policy debates captivation with the single integrated broadband infrastructure. In that dominant technological image, a single network based upon the integration of all of the new technologies would take care of all applications known today (from entertainment video to business communications), with room to spare -- in effect, the all-singing, all-dancing, all-integrated broadband network. But this is only one of the many possible technical development trajectories. Prudence would dictate that other trajectories ought at least to be explored before this one is exclusively blessed by policy.
Still, the all-singing, all-dancing vision is compelling, not least because the technical evolution to this common child appears straightforward. Simply add fiber optics and fast switching to upgrade the public phone network, and voila. But thinking of the current network challenge as a well-defined upgrade leads to a series of questions about how to get there, which dominate alike policy and the planning of network providers: Who should build the network (usually, Telcos or Cable TV), what will fill the pipe (usually, entertainment video), and, most profoundly, who should pay for it (usually, not granny)? If, as we argue, the evolution to broadband is neither an upgrade nor well-defined, rather discontinuous and largely unknown, then these are simply the wrong questions to tackle the current transition.
1 Who will build the pipe?
The Broadband ISDN vision implies one main network. Because there are several contenders to build it, the question then is framed as "Whose network is more amenable to being upgraded?" Because of the substantial investment required to build such an integrated network, building two competing integrated networks appears out of the question. The only two logical contenders, then, are the Telcos and the Cable TV companies, one with a narrow band but switched network and the other with broadband but no switching
In reality, something altogether different from this image is actually happening in the United States Multiple networks are simultaneously evolving toward true broadband capability. In effect, a broadband net-work portfolio is being built, with little likelihood that, in the medium term, any of the pieces will replicate the "all-singing, all-dancing" image. The networks combine various aspects of the new technologies and target different users and needs. The interexchange carriers are for the most part already fully broadband to principal business customers, as are metro-politan fiber nets (i.e., Teleport) and major parts of private corporate networks. The NREN is being similarly upgraded for intensive computer applications. Cable TV reaches the home, but only for entertainment, and is being re-engineered to provide limited communications services to a narrow class of at-home workers. Even the local phone companies, charged with common carriage and serving residences, are evolving toward broadband for business as a kind of competition-repellent to maintain their core customer base.
If the "all-singing and dancing" network is misleading, then the right question is, what is the proper network portfolio, and how can universal access and interconnection be maintained among the portfolios elements?
2 What will fill the pipe?
Demand for fundamentally new technologies is impossible to predict with any degree of accuracy given the inherent market, technical and economic uncertainties. As a consequence of belief in a straightforward network evolution, however, firms typically plan by extrapolation from the present. Two perfect examples of how this is misleading are predic-tions that gross over-capacity must result from current plans and that the only service likely to justify Telco investment in broadband to the home is residential video-on-demand.
Consider Michael Nolls highly publicized article.27 He argues that if all the people in the United States spent all their waking hours watching images/information over the net while talking on digital phones, this still would not generate enough traffic to justify the construction of an all-fiber network. The problem, of course, is that this predicts traffic based on current schemes where people talk to people, or people talk to computers. In fact, based on emerging applications for broadband that already exist, the biggest chunk of traffic is likely to be machines talking to machines, with the human element in the communications chain receiving essential streams of information via visualization of data in band-width intensive graphical interfaces. Moreover, as we have seen, broadband capabilities may be necessary whether or not the pipes are fully filled, if only to empower the substantial subset of business applications that require the speed and bandwidth on-demand, even if only temporarily.
Indeed, the emerging applications suggest an alternative question: Not what will fill the pipe, but how can sufficient experimentation and learning with the new technologies be guaranteed so that demand to fill the pipe can be effectively generated?
3 Who should pay?
Because the great majority of users (residences) do not perceive a need for broadband, and because the new applications it makes possible remain largely undefined, they do not want to pay for the upgrade. However, underlying the notion of an upgrade of the existing network to an integra-ted broadband infrastructure is the basic idea that it will in fact be financed through the rate base. This wrong question thus pits consumers against producers, and state regulations against federal competitiveness concerns.
The emerging patterns of use of broadband suggest two alternative contentions. The first is that, given sufficient experimentation and learning conjoined with the difficulty of effective network management, major users are willing to pay for broadband capabilities -- and to pay to embed them in publicly accessible networks -- even if they carry a premium over cost. The second is that major networks will change-over to broadband to serve that business opportunity and to repel competition to their core business revenues, but there is little evidence that the resulting investment patterns will result in an integrated broadband infrastructure reaching all citizens. If advanced users are in fact willing to pay for what they need, but what they need does not automatically result in widespread infrastructure, then the essential question becomes: Can we create incentives (or remove disincentives) for major business users to pay for the evolution of the network portfolio, while minimizing the costs of ensuring universal access for those least able and willing to pay?
B. Policy and the Network Portfolio
There are reasonable policies available to deal with the questions raised above -- although they are not the policies that are currently fashionable. Our principal recommendation is for policy to shift focus from concern with regulating monopoly or promoting competition to concern with ensuring that the appropriate technologies reach the appropriate users in a timely fashion and in ways that involve novel financing to replace the bygone cross-subsidies that guaranteed universality. Three corollaries, which build on the recommendations of our last study and which correspond to the three questions raised above, help to elaborate. These are:
1) Let market competition drive portfolio development, while policy focuses on three tasks: Reviewing the portfolio for gaps in service to major constituencies; pursuing development of missing elements; and self-consciously mandating experiments and testbeds in forced access and interconnection between the portfolios networks to assure the possibility of infrastructure coherence in the future.
Given path dependencies, there is no guarantee that any single category of network provider will offer an optimal infrastructure solution to bridge the discontinuity. Rather, a portfolio of networks appears necessary given the uncertainties. Only market competition is likely to provide a broad portfolio by stimulating development of the variety of networks that comprise the alternative technological approaches.
But competition cannot be the exclusive choice of policy-makers either. All of the cases suggest that competition will hasten change but also simultaneously cause both over- and under-investment in the socially optimal infrastructure portfolio. Moreover, because markets usually do not offer an optimum solution for standards formation, there is no guarantee that markets will provide access or interconnection among the elements that comprise the network portfolio. For example, VANs play an important role in filling the gaps between the larger pieces of the network portfolio, but the business dynamics are unlikely to make them the appropriate solution for access and interconnection because they will only provide that where the big business opportunities lie. In essence, competition can not entirely substitute for continued regulation.
Wherever there are major constituencies not being served by a nations given network portfolio, there is a need to fill the gap via public policy in the interests of universality. Such actions could take a variety of forms, from direct public funding of development, as with the Internet in the United States, to the imposition of common carrier requirements upon certain network providers, to old-fashioned cross-subsidy. Similarly, ensuring access and interconnection between the elements of the network portfolio remains an essential obligation of government. Policy should explicitly pursue testbeds for interconnection between the portfolios networks, just as it now licenses field-trials to test new technologies/ services within individual networks like the PSTN.
2) To ensure the necessary experimentation and learning, policy should permit essentially free exploratory use among targeted user populations of special national significance, while conducting fundamentally different field trials for others.
The corporate cases and our earlier study suggest the essential role that experimentation and learning play in network development. Some elements of the network portfolio, like large corporate private networks and businesses well served by VANs, will engage in that experimentation on their own. Most other user populations served by other network elements may not. This includes major residences and probably also small- and medium-sized businesses. Their experimentation and learning will have to be publicly fostered.
For many such populations, the Internet case suggests that subsidized experimentation and use is a public good that can be critical to learning, and thus to accumulating the appropriate network portfolio. This would be a particularly useful way of ensuring that important user populations of special national significance (whether for economic, security, social welfare, or other reasons) enjoyed early exposure and access to the new networks.
Network providers and other user populations will have to develop initial experimentation and learning the old-fashioned way -- through field-trials. But current trials, with their emphasis on technology and limited populations, are likely to be inadequate to the task of bridging the discontinuity.28 Field trials will thus have to be re-thought -- both extended in scope and participation (users, network providers and third-party service providers on the Minitel model) and probably financed differently (perhaps through a royalty on eventual service provision).
3) Where possible, policy must encourage major business users to finance large pieces of the emerging portfolio, while finding other additional means as necessary to reduce the costs to poor and initially non-participating constituencies.
The corporate cases suggest strongly that lead users will pay for what they need and, consequently, can finance a substantial part of the network portfolio. The cases also suggest that, where permitted, lead users would also pay to embed what they need in public infrastructure. This would probably remain true even if a premium was initially demanded of major users, so long as long-term cost reductions were guaranteed. Disincentives that prevent such outcomes -- like regulatory restrictions on new service provision -- should be removed. Moreover, a variety of incentives could help to encourage major users to finance infrastructure development, from tax breaks to a temporary royalty stream on services they thus helped to initiate.
However, extending the advanced network portfolio universally to currently disinterested communities is likely to require other novel financing schemes. These could come via the sale of obligations like bonds or even equity -- i.e., making individual citizens part owners of the advanced networks that reach them -- or through the subsidy of general tax revenues, of a special tax or import tariff, or even through reinstitution of cross-subsidies where appropriate. The worst possible way to accomplish this would be via straightforward inclusion in the rate base, in effect, making the least capable pay for what they do not want. A better way might be to link modern infrastructure to another social goal like a cleaner environment. A tax on pollution or energy inefficiency that went to subsidize the universal extension of modern network infrastructure could probably be defended politically.
C. The Inevitability of a Continuing Role for Government Policy
While the story we told last time could be described as incremental evolution along a given trajectory, we are now facing a discrete jump onto a new trajectory. Underlying each trajectory and the evolution it can sustain is a physical system. We are now putting in place drastically different physical networks with much greater capabilities. What we described in our last study was how companies could harness the newfound capabilities of the management and control layer. Firms strategies reflected incremental evolution in the use of a given physical infrastructure. With radical change in the underlying characteristics of the system should come new capabilities and new user-awareness. The sum total of shifts should, much more so than the last time we examined these issues, lead to networks that support fundamentally different industrial and competitive capabilities, and eventually to a new develop-ment trajectory for the economy that resides on the network infrastructure.
We doubt whether an incremental evolution to these new capabilities and their eventual economic development trajectory is even possible. Specifically, it appears that the technical and investment requirements for increasing network intelligence are not the same as for increasing band-width. Similarly, the technical and investment requirements for increasing entertainment applications are not the same as for increasing computer-to-computer interaction. The result is that evolutionary trajectories down one technical or application path may very well preclude effective exploration of alternative paths. That is why multiple networks -- the portfolio -- must be built and ultimately why the all-singing and dancing network is the wrong image for the moment. As we emphasized in the last study, the appropriate trajectories can only be gauged through experience in production and use, and by learning-by-doing. The overall investment in the infrastructure and necessary education (including experimentation and use) is likely to be quite staggering, but investment is already going forward. The issue, simply put, is to accelerate the development of relevant technologies, encourage access, interoperability and experimen-tation, and achieve a VIRTUAL integration of the evolving physical networks into a LOGICAL infrastructure.
These incredibly high stakes lead us to speculate about the inevitability of a continuing role for government policy in developing telecommuni-cations in the United States As we emphasized earlier, deregulation and market competition can not provide all of the necessary elements of the portfolio or even universal access and interconnection among them. As a result, there will certainly be relative winners and losers as network evolution occurs in the United States The U.S. political system remains reasonably responsive, especially at the margins, to losers, who in turn will have little choice but to invite political intervention. The outcome of that all-too-typical U.S. politics -- as recently occurred with re-regulation of CATV rates in the Cable TV Act of 1992 -- will be a high degree of unfocused and a-strategic government policy-making, perhaps even a thorough-going re-regulation. The only feasible alternative is for government to play a proactive role by addressing the issues outlined above before they provoke the losers to clamor for redress. Deregulation thus helped to create the current discontinuity, but only an effective com-bination of competition and proactive policy can manage the transition while delivering a coherent infrastructure for economic development in the next century.
|AAP||Alternative Access Provider|
|ATM||Asynchronous Transfer Mode|
|BOC||Bell Operating Company|
|BTNA||British Telecom North America|
|CAD||Computer Aided Design|
|CAM||Computer Aided Manufacturing|
|CPE||Computer Premises Equipment|
|EDS||Electronic Data Systems|
|GEIS||General Electric Information Services|
|ISDN||Integrated Services Digital Network|
|LANs||Local Area Networks|
|NREN||National Research and Education Network|
|NSFNet||National Science Foundation Net (backbone to the Internet)|
|ONA||Open Network Architecture|
|PSI||Performance Systems Inc.|
|PCN||Personal Communications Networks|
|PSTN||Public Switched Telephone Network|
|RBOCs||Regional Bell Operating Companies LATAs|
|SCPs||Service Control Points|
|SMDS||Switched Multimegabit Data Service|
|TCP/IP||Transmission Control Protocol / Internet Protocol|
|VANs||Value Added Networks|
|VPNs||Virtual Private Networks|
|WAN||Wide Area Networks|
1 Throughout this document, our use of the term telecommunications infrastructure covers the ensemble of information networks in the United States. More specifically, it includes common carrier networks (local and long distance), Cable TV networks, the Internet, VANs, and corporate private networks.
2 Bar F. and Borrus M., with Coriat, B., Information Networks and Competitive Advantage: Issues for Government Policy and Corporate Strategy Development, (Brussels: CEC/DGXIII, BRIE, OECD, 1990).
4 Paul David, Computer and Dynamo: The Modern Productivity Paradox in a Not-Too-Distant Mirror, CEPR Publication No 172, Stanford University, 1989.
5 This growth figure is an estimate given by Jordan Becker of ANS. Available data shows somewhat smaller but nonetheless dramatic growth on the NSFNet backbone (data available from Merit, via anonymous ftp to <nic.merit.edu>/statistics/nsfnet/). See Hart et al. "The Building of the Internet: Implications for the Future of Broadband Networks," Telecommunications Policy, November 1992, and BRIE Working Paper #60 (Berkeley, Calif.: BRIE).
6 See Hart et al., "The Building of the Internet: Implications for the Future of Broadband Networks," Telecommunications Policy, November 1992 and BRIE Working Paper # 60 (Berkeley, Calif.: BRIE).
7 Lewis Branscomb, "Information Infrastructure for the 1990s: A Public Policy Perspective" in Building Information Infrastructure: Issues in the Development of the National Research and Education Network, Brian Kahin ed., McGraw-Hill, 1992.
8 BTNA = British Telecom North America, GEIS = General Electric Information Services, EDS = Electronic Data Systems
9 RBOC = Regional Bell Operating Companies, the regulated local exchange carrier subsidiaries of the Regional Holding Companies (BellAtlantic, NYNEX, Ameritech, BellSouth, Southwestern Bell, USWest, and Pacific Telesis) created in the divestiture of ATT; Telco = local telephone company.
10 Based on consulting reports and industry estimates using Form M FCC data.
11 Based on industry conversations.
12.Very rough estimates based on industry market assessments.
13.Based on industry conversations.
14 Based on industry conversations.
15 We do not deal here with narrowband ISDN (Integrated Services Digital Network). Because it can be delivered with relatively modest changes in network architecture and is already being delivered today, it falls outside of our focus on advanced, high-bandwidth networks. However, our category of broadband switched services would include broadband ISDN.
16 The AIN description which follows is based on Bellcore and Pacific Bell documentation.
17.There are a large variety of anticipated services that include Network Call Completion (i.e., calling party can leave message for later delivery to a currently busy exchange), Do Not Disturb (i.e., outbound calls operate normally, inbound calls are rejected or routed elsewhere including to voicemail), Follow Me (subscriber-tailored schedule for call routing at different times to different locations), and a host of similar ones. New services are being defined with each new trigger developed in successive releases of AIN software.
18 Recall that applications flexibility is the ability to deliver different applications over a fixed network configuration; configurations flexibility is the ability to reconfigure the network to provide wholly new applications that would be otherwise impossible.
19 This description is based on Bellcore and Pacific Bell documentation.
20.Based on industry conversations.
21.On this point and for an elaboration of what follows, see the background paper by Jeffrey Sprafkin on the evolution of VANs, "VANs Role in the Advanced Network Infrastructure," BRIE Working Paper #63 (Berkeley, Calif.: BRIE).
22 Ibid., at p. 35, citing Permut, Telephony, June 15, 1992, p. 94, 96.
23 See Bruce Egan, The Case for Residential Broadband Telecommunications Networks, (Brussels: CEC-RACE, 1992).
24 Based on CATV trade journal estimates.
25 See Egan (ibid).
26.See Elton M., "Integrated Broadband Networks: Assessing the Demand for New Services," (Brussels: CEC-RACE, 1992).
27 Noll, Michael A., "Voice vs. Data - An Estimate of Future Broadband Traffic," IEEE Communications Magazine, June 1991, V29 N6, p. 22.