\documentclass[11pt]{article} \usepackage{times} \input amssym.def \input amssym.tex \usepackage{epsfig,psfig} \setlength{\textwidth}{6.2in} \setlength{\textheight}{9in} \setlength{\oddsidemargin}{.2in} \setlength{\topmargin}{-0.25in} \setlength{\headheight}{0in} \def\tm{\raise5pt\hbox{{\rm\tiny TM~}}} \newcommand{\hsp}{\hspace*{\parindent}} \makeatletter \def\section{\@startsection {section}{1}{\z@}{-3.5ex plus -1ex minus -.2ex}{2.3ex plus .2ex}{\normalsize\bf}} \makeatother \makeatletter \def\subsection{\@startsection {subsection}{1}{\z@}{-3.5ex plus -1ex minus -.2ex}{2.3ex plus .2ex}{\normalsize\bf}} % put a period after section or subsection number in header \def\@sect#1#2#3#4#5#6[#7]#8{\ifnum #2>\c@secnumdepth \def\@svsec{}\else \refstepcounter{#1}\edef\@svsec{\csname the#1\endcsname.\hskip .75em }\fi \@tempskipa #5\relax \ifdim \@tempskipa>\z@ \begingroup #6\relax \@hangfrom{\hskip #3\relax\@svsec}{\interlinepenalty \@M #8\par}% \endgroup \csname #1mark\endcsname{#7}\addcontentsline {toc}{#1}{\ifnum #2>\c@secnumdepth \else \protect\numberline{\csname the#1\endcsname}\fi #7}\else \def\@svsechd{#6\hskip #3\@svsec #8\csname #1mark\endcsname {#7}\addcontentsline {toc}{#1}{\ifnum #2>\c@secnumdepth \else \protect\numberline{\csname the#1\endcsname}\fi #7}}\fi \@xsect{#5}} % put a period after theorem and theorem-like numbers \def\@begintheorem#1#2{\it \trivlist \item[\hskip \labelsep{\bf #1\ #2.}]} \makeatother \thispagestyle{empty} \begin{document} \begin{center} {\Large\bf {The history of communications and its implications for the Internet}} \bigskip \\ Andrew Odlyzko \\ AT\&T Labs - Research \smallskip \\ amo@research.att.com \\ http://www.research.att.com/$\sim$amo \smallskip \\ Preliminary version, June 16, 2000. \\ \vspace*{1\baselineskip} {\bf Abstract} \\ \end{center} % \setlength{\baselineskip}{1.5\baselineskip} The Internet is the latest in a long succession of communication technologies. The goal of this work is to draw lessons from the evolution of all these services. Little attention is paid to technology as such, since that has changed radically many times. Instead, the stress is on the steady growth in volume of communication, the evolution in the type of traffic sent, the qualitative change this growth produces in how people treat communication, and the evolution of pricing. The focus is on the user, and in particular on how quality and price differentiation have been used by service providers to influence consumer behavior, and how consumers have reacted. There are repeating patterns in the histories of communication technologies, including ordinary mail, the telegraph, the telephone, and the Internet. In particular, the typical story for each service is that quality rises, prices decrease, and usage increases to produce increased total revenues. At the same time, prices become simpler. The historical analogies of this paper suggest that the Internet will evolve in a similar way, towards simplicity. The schemes that aim to provide differentiated service levels and sophisticated pricing schemes are unlikely to be widely adopted. Price and quality differentiation are valuable tools that can provide higher revenues and increase utilization efficiency of a network, and thus in general increase social welfare. % There are sound arguments in favor of price and quality differentiation. Such measures, most noticeable in airline pricing, are spreading to many services and products, especially high-tech ones. However, it appears that as communication services become less expensive and are used more frequently, those arguments lose out to customers' desire for simplicity. % The balance moves away from the need to operate % a network at high utilization levels and to extract the highest % possible revenues. Instead, catering to user preferences becomes % more important. In practice, user preferences express themselves through willingness to pay more for simple pricing plans. In addition, there is a strong ``threshhold'' effect to usage-sensitive billing. Even tiny charges based on utilization decrease usage substantially. In a rapidly growing market, it is in the service providers' interest to encourage usage, and that argues for simple, preferably flat rate, pricing. Historical evidence suggests that when service costs decrease, such arguments prevail over the need to operate a network at high utilization levels and to extract the highest possible revenues. Communication services have long exhibited many characteristics of modern high-tech industries, namely high fixed costs, network effects, and difficulty in allocating costs among many products and services. Therefore the historical lessons of this paper are likely to be applicable to many pricing situations in the future. When prices are high, and purchases infrequent, airline-style ``yield management'' is likely to dominate. On the other hand, when prices are low, and a service is used many times a day, simple pricing and uniformly high quality are likely to be more common. Historical analogies as well as current expenditure data also suggest that in the ``digital convergence'' of broadcasting and point-to-point communications, it is the latter that will dominate in shaping the evolution of the Internet. The current preoccupation with professionally produced ``content'' is probably more a distraction than help in planning for the future. Content has never been king, it is not king now, and most likely will not be king in the future. The development of the Internet is likely to be determined by the same growth of the myriad unpredictable commercial and social interactions that have fueled other communication services. \clearpage \tableofcontents %\input d53.toc % for final run, comment out \tableofcontents, edit file.toc, add a line % \input file.toc, % and make the "input" line work \clearpage \begin{center} {\Large\bf {The history of communications and its implications for the Internet}} \bigskip \\ Andrew Odlyzko \\ AT\&T Labs - Research \smallskip \\ amo@research.att.com \\ http://www.research.att.com/$\sim$amo \smallskip \\ \end{center} %\vspace*{2\baselineskip} \setlength{\baselineskip}{1.5\baselineskip} \section{Introduction} \hsp The goal of this work is to draw lessons from the history of communications. The view is broad, covering many technologies, and many decades or even centuries of data, and thus is of necessity not very detailed. However, it does produce some clear patterns. This is remarkable in that the history considered here covers many services, including regular mail, the telegraph, and the telephone. This history also covers many types of economic environments, including government monopoly, private untrammelled monopoly, government regulated monopoly, and competitive markets. The only constant factors have been relentless technical progress and growth in volume of communications. The historical record carries lessons for pricing, quality-differentiated services, the valuations of the dot-coms, as well as for the question of whether indeed ``content is king'' on the Internet. Many of these lessons carry over to general ecommerce. The Internet has historically treated all packets equally, and pricing has been predominantly through flat monthly rates depending only on the size of access links, not on usage. Today, there is a strong momentum towards changing both of these principles. Influential figures in the data networking community make statements such as the following: \begin{quote} Although flat-rate continues to be the predominant form in which Internet access is sold, that form of pricing is unviable. Flat-rate pricing encourages waste and requires 20 percent of users who account for 80 percent of the traffic to be subsidized by other users and other forms of revenue. Furthermore, flat-rate pricing is incompatible with quality-differentiated services. \hspace*{+3in}Pravin Varaiya, abstract of \\ \hspace*{+3in}INFOCOM'99 keynote lecture \end{quote} There is extensive work on development of multiple service levels, typically with some implied guarantee of performance, usually referred to as Quality of Service (QoS). If those are deployed, they will necessarily require pricing schemes more complicated than the standard flat rate ones. The driving force behind these developments is economic, based on the view that it is not feasible to provide satisfactory levels of quality for all transmissions at a reasonable cost. There is also concern (expressed in the quote above) about fairness of light users subsidizing heavy ones under the current scheme. While the momentum is towards introduction of QoS and more complicated pricing systems, there have always been voices arguing in favor of preserving the simplicity of both operations and pricing of the current Internet. Their counterarguments can be substantiated through evaluations of the costs of multiple service levels to end users, and on considerations of what utility users derive from data networks [Odlyzko7, Odlyzko8]. Shortage of bandwidth is less of a problem than is often thought, and the general complexity of the system is a large barrier. An important observation behind many of the arguments in those papers is that what users really care about are fast transactions. As a result, whenever they can afford it, they run their links at low utilizations, since the main benefit they derive comes from the ability to burst at high speeds for short periods. In such an environment, flat rate pricing is most appropriate, and QoS does not help much. The historical approach of this paper also leads to a skeptical view of most of the ongoing work on QoS. There have been many calls for usage-sensitive pricing similar to that of Pravin Varaiya quoted above. For example, an investigation of phone service in New York City in 1905 concluded \begin{quote} that, so far as large cities are concerned, unlimited service is unjust to small users, favors large users unduly, impedes expansion of the telephone business, tends to inefficient service, and that, as a financial proposition, is unsound. \hspace*{+3in}p.~246 of [Stehman] \end{quote} As will be shown in Section 14, the case for usage sensitive pricing in local phone calling in 1905 was much stronger than it is on the Internet today. However unsound it may have seemed, unlimited local calling for a flat monthly fee for residential users has persisted in most of the United States for all of the 20th century. It seems worth exploring why it did, what effects that has had in comparison with other countries that adopted usage-sensitive pricing, why it is being re-introduced in many of those countries, and what conclusions we can draw from this about the likely future of the Internet. If we compare the telecommunications industries in different countries, we find few signs of harm from the ``unsound'' practice of flat rates for local calls. Table 1.1 shows that U.S. citizens use their phones considerably more than inhabitants of other affluent industrialized countries at only a slightly higher cost. Thus at least from this high level view, % it appears that both consumers and service providers benefit from flat rate pricing. \begin{table}[htb] \begin{center} Table 1.1. International comparison of telephone industry revenues and usage in 1997 \\ ~ \\ \begin{tabular}{lr@{}lr@{}l} country & \multicolumn{2}{c}{revenues as} & \multicolumn{2}{c}{minutes of phone} \\ & \multicolumn{2}{c}{fraction of GDP} & \multicolumn{2}{c}{calls per person} \\ &&& \multicolumn{2}{c}{per day} \\ ~ \\ Finland & ~~~~~~~2&.52\% & ~~~~~~~16&.6 \\ France & ~~~~~~~1&.93 & ~~~~~~~10&.6 \\ Japan & ~~~~~~~2&.06 & ~~~~~~~10&.6 \\ Sweden & ~~~~~~~2&.05 & ~~~~~~~20&.7 \\ Switzerland & ~~~~~~~2&.66 & ~~~~~~~13&.0 \\ U.K. & ~~~~~~~2&.51 & ~~~~~~~12&.7 \\ U.S. & ~~~~~~~2&.86 & ~~~~~~~36&.9 \\ \end{tabular} \end{center} \end{table} One can question whether history has much to teach us about the Internet. After all, incidents such as the transatlantic telegram of 1866 that almost bankrupted the U.S. State Department (see Section 13) show massive incompetence in cryptography and data compression as well as in basic business practices and diplomacy. Thus one can object that today we are much better prepared to handle a sophisticated technology. On the other hand, many of the practices that service differentiation would imply have been applied in the past. For example, Section 14 relates how over a century ago an Englishman was reprimanded for using his phone to summon the fire department to a blaze in a neighbor's house. History never repeats exactly, but there are frequent recurring themes that illuminate the present and hopefully can help make better choices for the future. While QoS and sophisticated pricing schemes are widely regarded as necessary for the future of the Internet, the trend in other communication services has almost uniformly been in the opposite direction. Many examples are presented in later sections, and I will quote just a few here. In long distance voice telephony, the most popular plans are the simple ones that are independent of time of day or distance. Even in the Internet arena, the trend so far has been largely towards simple flat rates. A decade ago, the mass market online services, such as CompuServe, Prodigy, and AOL charged not just for minutes of connect time, but even for individual email messages. Email charges were eliminated first. By the middle 1990s, these services were also forced by subscriber complaints and competition from other ISPs to switch to unlimited access for a flat monthly fee. Although simplicity in pricing has been the general trend, it is important to recognize that there is persuasive logic to differentiated services. Even in the absence of quality differentiations, there are good reasons for pricing according to usage, as is done with electricity. Some of the arguments in this direction will be reviewed later. A major problem with such measures, though, is that usage sensitive charging, even with very low rates, lowers usage, often dramatically. Sometimes that is desirable. Sometimes, though, it goes against public policy or against interests of service providers. Western Europe is beginning to re-introduce forms of flat rate local phone calling precisely in order to stimulate Internet usage, which lags behind that in North America. Numerous examples are presented later to show that the best way to stimulate usage is to have flat rates. The Internet is expected to continue growing and to penetrate ever more deeply into our lives. Hence the competition among ISPs will be won by those who can entice their subscribers to make the fullest use of the services that are offered. Flat rates or approximations to them are among the most powerful tools in this competition. The general trend has been for each communication service to offer higher quality and lower prices. This has gone hand in hand with higher usage, and higher total revenues. (The telegraph is a partial exception to this trend, and it is instructive to study its history for that reason.) At the same time, pricing has tended to become simpler. A particularly telling instance of the trend towards simplification in communication is the tendency for prices to become distance-insensitive as a service evolves. The basic justification for the development of quality-differentiated service levels on the Internet is the perception that transmission and switching resources are expensive, and therefore high quality cannot be provided uniformly for all traffic. Yet that does not seem to have been a barrier for other communication services. The distance dependence in pricing of mail, telegraph, telephone, and even data network services has historically decreased, or even disappeared completely. A Martian who saw just the evolution of price schedules might be tempted to argue that if the extra cost of network facilities to send messages to distant locations is not worth charging for, then the extra cost of overprovisioning the network to provide uniformly high quality to all transmissions is likely to be negligible as well. As will be explained in Section 6, distance dependence in pricing is often a sign not of substantially greater costs for long distance service, but a form of price discrimination. Long distance transmission is regarded as more valuable. Thus the decrease in distance dependence of pricing provides an even stronger argument than that of the hypothetical Martian. Perhaps an even more interesting point of view goes even deeper into an analysis of possible forms of pricing. A schedule of charges based simply on distance is already a major concession to the desire for simplicity, and away from optimality in either revenue extraction or in relating prices to costs. One can take the historical evidence of this paper that communication pricing has been evolving towards simplicity and argue that by simple analogy, the Internet is also likely to move in that direction. A sounder procedure is to investigate more carefully what factors contributed to historical developments, and what influence they might have in the future. There are, after all, reasonable arguments against simplicity. A large part of this paper is devoted to an investigation of the reasons for various pricing strategies. Sophisticated strategies for quality differentiation and price discrimination are common. They are perhaps most noticeable in airline yield management, but are spreading to other areas. For example, Coca Cola is experimenting with vending machines that will automatically raise prices when temperatures are high. We can expect such practices to be widely adopted for two main reasons. One is that the economic evolution of our society is increasing the role of fixed costs in the provision of goods and services. Therefore pricing on the basis of marginal costs is becoming untenable, and it becomes necessary to price on the basis of customers' willingness to pay. That calls for approaches such as those of airlines and Coca Cola. The other reason we can expect such practices to spread is that modern information technology is making them possible. In the past, Coca Cola might have wanted to price its drinks depending on customers' thirst, but could neither predict the degree of that thirst, nor could it adjust prices in a timely fashion. Now it can do both. While there are sound arguments for quality differentiation and price discrimination (treated briefly in Section 5), there are also economic arguments against them (described in Section 7), based on economic models that favor bundling. Perhaps more important, such practices conflict with the strong preference that people have for simple pricing, especially for flat rate schemes. (Section 8 is devoted to this subject.) There are many reasons for this preference. The insurance aspect is important (``How do I know how high a bill my teenagers will run up?''). So is another aspect, neatly summarized by Nick Szabo's phrase [Szabo], ``mental transaction costs;'' ``Yes, I can save by optimizing my usage, but do I want to, if the savings come to fractions of a penny, and require my attention dozens of times a day?'' The main point of this paper is to observe how the conflict between the need to optimize and people's reluctance to optimize has been resolved historically. The general conclusion is that complicated pricing and quality differentiation strategies have been applied and continue to be applied for expensive or infrequent purchases. On the other hand, the search for optimal efficiency has usually been abandoned in favor of simplicity for small and frequent transactions. The long history covered in this paper illustrates how the decreasing prices of communication services, the fruit of improving technology and of economies of scale, have gone hand in hand with the trend towards greater simplicity. It seems reasonable to expect that a similar division between the quest for optimal efficiency on one hand, and simplicity on the other, will hold in the future. It will be easier for service providers to implement complicated pricing schemes. There will also be more tools (and intermediaries) helping consumers to navigate the maze of choices. However, both factors have been operating for a long time and that appears only to have strengthened people's desire for simplicity. The Internet is leading to an unprecedented proliferation of choices in goods and services, straining human capacity to deal with all the complexities. (As an example, voice response systems are used widely, yet have not diminished the need for human assistance. Instead of just simplifying old tasks, they have led to more complicated choices, which require human assistance on top of the automated systems.) Therefore the balance between optimization and simplicity is likely to remain where it has been, in spite of the developments in technology. That is a major reason that airline tickets % which are expensive and bought relatively infrequently are much more likely to be sold through complicated schemes than voice calls. Even in the U.S., which is affluent, and has a large air transport industry, there is only about one round trip per person per year by air, at a cost of around \$400. Thus airplane tickets are expensive and are bought relatively infrequently. Further, air transport is a mature industry, and costs are declining slowly. On the other hand, on average each U.S. inhabitant is engaged in about a dozen phone calls per day, each typically lasting about three minutes, and costing either nothing or a fraction of a dollar. How much thinking about the price and value of each call would people want to do? The same arguments apply even more strongly to Internet usage. % People with broadband access appear to visit a dozen or more % sites per day, and this number is only likely to increase. Will we want to be bothered about the cost of each Web page download, or any of the myriad other small transactions we will be engaging in? Moore's Law is providing ever powerful silicon chips, but the nerve cells in our brains are not getting any faster. Human attention may be the ultimate scarce resource. The historical trend of increasing volumes and decreasing costs in communications is illustrated by comparing the purchase of a book today to the receipt of a letter in England in the the early 19th century. In 1834, about 5 letters were mailed per person per year in England. In most parts of the country, the recipient had to learn that a letter was waiting, go to the post office and pay a fee that averaged what about two hours of work cost. (See Section 12 for more details.) In contrast, in the U.S. of 1994, there were about two pieces of mail delivered per person per day. The cost of each piece was under two minutes' wages for an average worker. That same year, 1994, the U.S. book industry sold 2.13 billion books, or about 8 per person. The average price was \$11.19 per volume, a little less than the average hourly wage. The full cost of the acquisition also included going to the bookstore to find the book, pay for it, and so on. Today, half a dozen years later, we are in the in the Amazon.com era. Those additional non-monetary book purchase costs can be decreased, as we can order books with a few keystrokes and mouse clicks, and have them delivered to our house a day or two later. Thus in terms of relative costs, books today are acquired more often and cost less than letters 170 years ago. % No wonder letters in those earlier times were written carefully. As costs decrease, and volumes increase, it is only natural that the care we devote to each message or transaction goes down. Thus it is to be expected that we should be less willing to spend time and energy optimizing our transactions. Section 3 reviews the remarkable and continuing increase in the volume of communications, and the effects this has had on the care we put into individual messages. % its decreasing unit costs. % the mail volume statistics takes the population of UK to % be basically just that of England, Wales, and Scotland, % which in 1831 had 16.3 M people. If also include Ireland % (which apparently was part of the same postal system), would get % additional 7.8 M people, which would depress stats on ave. no. % letters per person. However, apparently the volume of mail % in Ireland was low. Before considering growth and price statistics, in Section 2 we look at the relative values of different types of communication. Currently there are widespread expectations that ``content,'' usually denoting professionally produced multimedia traffic, will dominate the Internet. However, even if the volume of multimedia traffic does exceed that of other transmissions, designing the network primarily for such traffic is likely to be a mistake. Historical evidence, including current statistics, shows that people are willing to spend far more on point-to-point communication than on entertainment. Thus it is unlikely that professionally produced multimedia traffic will determine the evolution of the Internet. To continue attracting the resources that we are devoting to communications, the Internet will have to cater to user desires for flexibility, universal reach, informal associations, and other spontaneous interactions that no ``content'' can satisfy. If the above prediction about secondary role of content comes to pass, it will fit a historical pattern. For example, in the early 19th century, U.S. postal policy was preoccupied with broadcast-type communication, namely through newspapers. Distribution of newspapers was subsidized to an extraordinary degree by regular letters, which brought in most of the money. Eventually public pressure forced policy makers to lower the prices of letters, which led to an upsurge of traffic. The next few sections review some of the main arguments for and against price and service quality differentiation. Section 11 then discusses pricing trends in several areas related to communications. Finally, Section 12 starts consideration of communications by examining the history of the postal system. This is continued in the following sections on the telegraph, the telephone, and various related services. Historical analogies are interesting, but may not always apply to a new technologies. The Internet is unique, and some of the important ways it differs from other services are examined in sections 20-23. Perhaps most important, prices have been increasing in recent years for most of the Internet. This is a seldom-noted phenomenon that is almost unique among modern high-tech industries in which demand has been increasing and technological progress has been stunningly fast. It is likely that this increase in transmission prices has been responsible for much of the demand for QoS and complicated pricing. Once prices start plummeting overall, as they have begun to do on some routes, it seems likely that providing more bandwidth will be seen to be more attractive than forcing complicated schemes on users. There are two other ways in which the Internet is often claimed to differ from other communication services. One is in the possibility for a ``tragedy of the commons,'' in which flat rate pricing leads users to saturate any links that are offered. The other is in the independence of traffic from distance. However, as sections 21 and 22 show, neither of these is likely to be a serious problem for the Internet. The general conclusion that I draw from the evidence presented in this paper is that the Internet is not unique. Most likely it will continue to evolve the way other communication services did, namely towards higher quality for all transmissions, lower prices, and simple prices. We are likely to end up with a mix of traditional flat rate offers, together with some form of the ``expected usage pricing'' scheme proposed in [Odlyzko8]. That proposal is discussed in greater detail in Section 5. In this model, a customer (residential or business) would be offered unlimited usage for a year, say, with the price determined just by the capacity of the link and that user's record. Service providers would act as insurance companies, using prior usage, demographics, and other information to fashion customized bundled offers that would meet the demands for simplicity and flexibility. % xxx The Internet is likely to evolve in the direction % of the electric utility. we do not have several grades % of electricity, and don't have to decide % xxx separate sockets for a shaver, a radio, or a light % xxx special cases: uninterruptible power supplies, higher voltage, etc. This paper pictures pricing of communication services as the outcome of a competition between the need to optimize on one hand, and to cater to user preferences for simplicity on the other. This view is unconventional. Most standard works, such as [MitchellV], concentrate on optimizing some conventional quantitative utility function, and treat customer desires as irrelevant, or else as irrational annoyances that interfere with clever and efficient schemes. In particular, such works treat preferences for flat rates dismissively, if at all. For example, we find in a paper on pricing of local phone service the following statement: \begin{quote} ... Clearly a movement to a positive per call charge would increase aggregate economic efficiency. Yet nearly all proposals for a move to [usage-sensitive pricing] have met stiff consumer resistance. The reluctance seems to persist even when customers face the prospect of a [usage-sensitive pricing] plan that would, on average, result in a lower monthly bill. \hspace*{+3in}[Panzar] \end{quote} That paper then goes on to propose a usage-sensitive pricing plan that would hopefully help wean customers from their apparently irrational reluctance to embrace such schemes. This paper approaches the problem differently. It presents extensive historical and current data about customer preferences, explores the reasons for such preferences, and considers the effects of catering to them. It argues that however irrational such preferences may seem, they should be paid attention to. It shows that, generally, such a course of action is likely to be best for rapidly growing communication services like the Internet. Technology is boosting the capacity of all networks. Therefore customers have to be encouraged to increase usage, to fill this growing capacity. Carriers that fail to persuade their customers to move on to higher bandwidth services will be left behind in the competitive race. Flat rate pricing is optimal in encouraging use. For cases where flat rate pricing plans are not feasible, this paper also suggests usage sensitive plans that try to appear as close to flat rate to the user as possible. Standard works on telecommunications economics usually say nothing about the effect of flat rates on usage. The only work I have found that treats this subject in any detail is the rather obscure report [Kraepelien]. This is obviously not the definitive history of communications. No work could possibly fill that role. Even a complete coverage of communication pricing, the subject of most of this paper, is impossible in a work of this scope. (The book [Smith] devotes over 400 pages to a study of postage rates alone, and yet covers only a few countries and only up to World War I.) In particular, practically no attention is devoted to technology, regulation, industrial organization, and many other crucial aspects of communication services. Their effects are noted only through the decreasing prices and increasing quality and quantity that users experience. Instead, I pick out a few threads from this rich subject that combine to present a coherent picture of relevance for the Internet. The main themes are growth of communication, the qualitative change this growth produces in how people treat communication, and the evolution of pricing. The stress is on the user, and in particular on how quality and price differentiation have been used by service providers to influence consumer behavior. The selection of historical data in this paper is heavily weighted towards the early periods of each technology that is considered. One reason is that those early periods tend to be more colorful. For example, today telegraph company lawyers do not hire people to cut down the poles of a competitor (as happened in England in 1861, see p.~63 of [Kieve]]). Also, railroads that compete for the same bottleneck mountain pass do not hire private armies that engage in shooting wars (as happened in the infamous 1878-1879 contest between the Denver and Rio Grande Railroad and the Atchison, Topeka, \& Santa Fe). A more substantial, but related, reason for concentrating on early stages of communication services is that they help clarify the evolution of those services. The contrasts between the early and mature stages of a technology bring into sharper contrasts the issues of how people treated those services, and what the role of pricing was. The general conclusion is that the forces that have worked to produce simplicity in pricing in the past are likely to be at least as influential on the Internet. Optimal efficiency will be sacrificed in the interests of satisfying customers. Further, professionally produced ``content'' will not be the main factor determining the evolution of the Internet. Communications is likely to grow in revenues and importance, but will do so in often unpredictable ways, as society adopts the new tools to suit its needs. {\em Note:} A section at the end, between the Acknowledgements and the References, lists sources for the data presented in the tables. It is worth emphasizing now that all the monetary figures are in current dollars, and thus are not adjusted for inflation. Price indices for the 19th century are rather imprecise, and in any case inflation-adjusted costs do not present an accurate picture of the burden of prices back when living standards were much lower. Therefore some prices are stated in terms of hours of work required to pay that price. \section{Communications: grit and glamor, or why content is not king} \hsp The Internet is widely predicted to produce ``digital convergence,'' in which computing, communications, and broadcasting all merge into a single stream of discrete bits carried on the same ubiquitous network. The popular images of convergence are heavily tinged with the flavor of Hollywood. ``Content is king'' is the universal buzzword, where content is usually taken to mean professionally prepared material such as books, movies, sports events, or music. The % market and technology race is supposedly to determine which organization or alliance will dominate in providing content to the users, ideally in advanced multimedia formats. A recent article concludes that ``[the Internet] has become a mass medium used mostly by relatively passive consumers, and as such major content providers will dominate it'' [MargolisR]. The book [Winston] also presents the Internet as the next step in the evolution of mass media. Many industry leaders appear to base their strategies on this thesis. For example, at Global Crossing, its latest CEO, Leo Hindery, is attempting % ``to turn this global Internet-based network into a mature % content distributor.'' \begin{quote} to turn this global Internet-based network into a mature content distributor. ... ``I don't want to be anyone's dumb pipes,'' says Hindery. ``If all you do is racks and servers, that's dumb. What we're doing is melding the network and the content.'' \hspace*{+3in}[Krause] \end{quote} This preoccupation with content is not peculiar to North America. Norio Ohga, once CEO and recently chairman of Sony, says that ``[w]ithout content, the network is nothing'' [Schlender]. Juan Villalonga, the chairman of the dominant Spanish communications carrier Telef\'onica, bases his strategy on the belief that ``[t]he key ... is content. Without it, ... phone companies risk becoming simple commodity pipelines'' [Baker]. Unfortunately for these companies, content is not the key. Content certainly has all the glamor. % , and tons of publicity. What content does not have is money. This might seem absurd. After all, the media trumpet the hundred million dollar opening weekends of blockbuster movies, and leading actors such as Julia Roberts or Jim Carrey earn \$20 million (plus a share of the gross) per film. That is true, and it is definitely possible to become rich and famous in Hollywood. Yet the revenues and profits from movies pale next to those for providing the much denigrated ``pipes.'' The annual movie theater ticket sales in the U.S. are well under \$10 billion. The telephone industry collects that much money every two weeks! Those ``commodity pipelines'' attract much more spending than the glamorous ``content.'' A reasonable objection to the comparison above is that what matters in directing investments is profits, not revenues. That is true. However, one cannot have profits without revenues. Further, profits of the telephone industry have dwarfed those of Hollywood. Even if we look at profitability in terms of return on investment, it is not clear that movies have been notably more profitable than communications, especially if one adjusts for risks. Those who invested in Disney in the early 1980s (but not recently) have done very well, but Sony took a bath in its takeover of Columbia Pictures. Some creative talent has done very well. An outstanding example is Steven Spielberg, who became a billionaire, while minimizing his risks through careful structuring of the deals. On the other hand, for most actors and writers, the financial rewards are much slimmer and spottier, as Hollywood is very much a ``winner-take-all'' market. In communications the risks may also be rising, as the Iridium debacle demonstrates, but so are potential returns. At this moment Wall Street gets attracted primarily to the prospects for rapid growth, figuring that profits will show up some time in the future. While the actual returns that Wall Street seems to be expecting may be ludicrously overoptimistic, the general principle appears valid. The histories of the telegraph and the telephone show the same pattern. Usually many companies jumped in with unrealistic hopes, most failed, but the industry as a whole prospered. Therefore the rest of this section will concentrate on revenues and growth rates of various sectors of the high-tech economy, and what they say about the current and potential role of content. Table 2.1 presents statistics that show the relative sizes of several sectors of the U.S. economy. The data was drawn primarily from [USDOC2], and the year 1997 was the most recent for which all the relevant time series were available. The precise description for how the figures were obtained is given in the section just before the References. There is considerable overlap in different categories in Table 2.1. For example, the \$187.5 billion of advertising industry revenues pays for almost all television broadcasting, and that provides much of the funding for the movie industry. Further, consumer expenditures on phone services are already contained completely in the general telephone industry figure. Some of the categories, such as sporting goods and airlines, are included just for illustration. % , as their relevance to the arguments of this paper is slight. %\begin{table}[htb] \begin{center} Table 2.1. Selected sectors of U.S. economy. $$ \begin{array}{lr@{}lr@{}lr@{}l} \multicolumn{1}{c}{\mbox{industry}} & \multicolumn{2}{c}{\mbox{1994 revenues}} & \multicolumn{2}{c}{\mbox{1997 revenues}} & \multicolumn{2}{c}{\mbox{annual growth}} \\ ~ & \multicolumn{2}{c}{\mbox{(billions)}} & \multicolumn{2}{c}{\mbox{(billions)}} & \multicolumn{2}{c}{\mbox{rate}} \\ \\ \mbox{telephone} & $~~\quad$\$199&.3 & $~~\quad$\$256&.1 & $~~~~\quad$8&.7\% \\ \mbox{\hspace*{+.25in}long distance} & 81&.0 & 98&.5 & 6&.7 \\ \mbox{\hspace*{+.25in}wireless} & 16&.8 & 33&.5&25&.9 \\ \\ \mbox{U.S. Postal Service} & 49&.6 & 58&.3 & 5&.5 \\ ~ \\ \mbox{advertising} & 151&.7 & 187&.5 & 7&.3 \\ \\ \mbox{motion pictures} & 53&.5 & 63&.0 & 5&.6 \\ \mbox{\hspace*{+.25in}movie theaters} & 6&.2 & 7&.6 & 7&.0 \\ \mbox{\hspace*{+.25in}video tape rentals} & 7&.0 & 7&.2 & 0&.9 \\ \\ \mbox{broadcast industries} \\ \mbox{\hspace*{+.25in}television broadcasting} & 31&.1 & 36&.9 & 5&.9 \\ \mbox{\hspace*{+.25in}radio broadcasting} & 10&.5 & 13&.5 & 8&.7 \\ \mbox{\hspace*{+.25in}newspapers} & 47&.2 & 55&.3 & 5&.4 \\ \mbox{\hspace*{+.25in}magazines} & 17&.4 & 19&.9 & 4&.6 \\ \\ \\ \mbox{consumer spending on ``content''} & 113&.9 & 133&.5 & 5&.4 \\ \mbox{\hspace*{+.25in}subscription video} & 29&.2 & 41&.5 & 12&.4 \\ \mbox{\hspace*{+.25in}home video (rental and purchase)} & 17&.8 & 20&.4 & 4&.6 \\ \mbox{\hspace*{+.25in}home video games} & 3&.1 & 4&.4 & 12&.4 \\ \mbox{\hspace*{+.25in}newspapers} & 12&.8 & 13&.6 & 2&.0 \\ \mbox{\hspace*{+.25in}consumer magazines} & 9&.5 & 10&.1 & 2&.1 \\ \mbox{\hspace*{+.25in}consumer books} & 20&.2 & 20&.9 & 1&.1 \\ \mbox{\hspace*{+.25in}recorded music} & 14&.7 & 14&.9 & 0&.5 \\ \\ \mbox{consumer spending on phone service} & 70&.5 & 85&.4 & 6&.6 \\ \\ \mbox{sporting goods sales} & 53&.5 & 64&.1 & 6&.2 \\ \\ \mbox{airlines} & 88&.3 & 109&.5 & 7&.4 \\ \\ \mbox{national defense} & 281&.6 & 270&.5 & -1&.1 \\ \\ \\ \end{array} $$ \end{center} %\end{table} What is striking is how highly valued communications is. If we combine the revenues of the phone industry with those of the postal service, we obtain a figure larger than military spending, and almost three times higher than the revenues of the airline industry. Just the spending on phone services is higher than all advertising outlays. So say good-bye to all those plans for financing the Internet through advertising! Yes, advertising can help fund some services, but it will not provide the generous revenue streams that are needed to support a communications infrastructure as large as the phone system. To obtain the funding that many dot-coms seem to be planning on, it will be necessary to get contributions from more than advertising. Ecommerce can help, but even that probably will not be enough, and it will be necessary to persuade people to pay for a large chunk of their communications. The question is, what are people willing to pay for? Table 2.1 shows that some advertising-supported business models might indeed be feasible. For example, sales of recorded music come to about \$15 billion per year. If one eliminates the overhead costs of the physical distribution system for CDs, one could probably provide the artists with as much money as they make now, and the music labels with as much revenue for their central selection and promotional activities (and their profit) for under half of the \$15 billion. That would be about half of the advertising revenues that the radio industry collects for broadcasting music. Getting that much extra funding from advertising might be possible for an Internet music service that allowed listeners much greater selectivity and thereby led to more listening (as Napster appears to be doing on college campuses). However, such a move appears feasible only because recorded music is a relatively small market. We could not hope to obtain enough advertising money to pay for anything as large as the phone system. Although Table 2.1 makes a powerful case by itself, it is worth reiterating the basic theme, which is that the vaunted ``content'' is not where the action is. The postal system alone collects almost as much money as the entire movie industry, even though the latter benefits from large foreign sales. For all the publicity it attracts, entertainment is simply not all that large, because people are not willing to pay very much for it. The dream of the early 1990s of financing the ``Information Superhighway'' through ``500 channels to the home on the cable TV network'' was an obvious fantasy. Content is not only a small part of the economy, it is often paid for indirectly. Well over two thirds of newspaper revenues, and almost all of broadcast TV and radio revenues come from advertising. Thus content is being given away in order to attract people to goods and services they are willing to pay for. Although spending for content (whether by consumers or advertisers) has been rising, it has been doing so at a sedate pace, and is unlikely to explode. One could object that Table 2.1 proves just the opposite of what is claimed above. After all, this table shows that even if no single content segment collects anywhere near as much money as the phone system, in aggregate huge sums of money are being spent on content. In particular, consumer spending on content is over 50\% higher than on phone services. There is some issue of what one means by content, a question we will return to later. If we take a generous interpretation, we can come up with total content industry revenues comparable to the \$256 billion that the telephone industry collected in 1997. (This would include consumer spending as well as business information services and advertising revenues of broadcast industries.) However, comparing just the total revenues of those two industries is misleading. In the case of the telephone industry, the \$256 billion does include some service revenues as well as yellow pages advertising, but the overwhelming majority of that money is for simple transport of voice and data. The content industry as a whole, though, % just like the cable TV segment, has to use its revenues to pay for content as well as the delivery of the content. The creative souls who command \$20 million per movie know their value, and do manage to appropriate a lion's share of the profits from such enterprises, while keeping their risks lower than those of the investors. Similarly, the professionals who compile the Lexis database, or assemble and monitor Reuters' financial data feeds, have to be paid, just like movie actors and musicians. Even if their average pay is lower, there are more of them, and their payroll, as well as all the equipment and overhead needed to support their work, are not inexpensive. Hence only some of the revenues of the content industry contribute to the communications infrastructure. Since this work is concerned with the future of the Internet, it is the present and potential funding for the network that matters. That only a fraction of the revenues of the content industry go for delivery is an important point that has analogs in the context of ecommerce. Ecommerce is already big, and is exploding. However, what does that mean for the network? A dollar of ecommerce transactions does not mean a dollar devoted to the network. When that sterling example of ecommerce, Amazon.com, sold \$1.6 billion of goods (primarily books) in 1999, it is likely that only a few million dollars of the \$2.3 billion of its costs went for Internet connectivity. Considerably more, but still only around \$150 million, went for the servers, software development, and other information technology products and services that are needed to stay competitive in this rapidly changing field. If the Amazon figure of \$1.6 billion is to represent the ecommerce opportunity for the Internet, then t-commerce (``t'' standing for ``telephone'', and covering all deals that use the phone in any way) amounts to tens of trillions of dollars. (Yes, more than the GDP, since the wonderful accounting of the ecommerce world surely would let us count the same value several times, as it passes through multiple transactions that all use the telephone at some stage.) The contribution of ecommerce to communications is growing, but it has to be kept in perspective. Communications is far from being the largest segment of the economy. It is smaller than cars, housing, food, and especially medicine. It is also about two thirds the size of the primary and secondary education sectors, and comparable to the higher education enterprise. The main point, though, is that communications is huge, and represents the collective decisions of millions of people about what they want. It is also growing relative to the rest of the economy in a process that goes back centuries, as the next section will show. As a fraction of the U.S. economy, it has grown more than 15-fold over the last 150 years. The key point of this section is that most of this spending is on point-to-point communications, not for broadcast media that distribute ``content.'' The predominance of point-to-point communications spending is not new. That has been the historical pattern for ages. For example, in the early 19th century, almost all the revenues of the U.S. postal system came from letters. Yet about as many newspapers as letters were being delivered. % (In Britain, % the fraction of postal traffic that was newspapers was somewhat lower, % around a third, % as is discussed in more detail in Section 12.) The preoccupation of decision makers with content and broadcast communication is also not new. In the early 19th century, the explicit policy of the U.S. government was to promote wide dissemination of newspapers. They were regarded as the main tool for keeping citizenry informed and engaged in building a unified nation. Hence newspaper distribution was subsidized from profits on letters, as will be discussed at greater length in Section 12. The extent of the subsidy may be gauged by the fact that ``[i]n 1832, newspapers generated no more than 15 percent of total postal revenues, while making up as much as 95 percent of the weight'' (p.~38 of [John1]). The policy of the U.S. government to promote newspaper ``content'' at the expense of person-to-person communication through letters may or may not have been correct. It would be a hard task (and one well beyond the scope of this work) to decide this question. However, there are reasonable arguments that the preoccupation with newspapers harmed the social and commercial development of the country by stifling circulation of the informal, non-content information that people cared about. In the 1840s, responding to public pressure, Congress did reduce letter rates, which resulted in increased usage, and changed patterns of usage, as is described at greater length in Section 12. In those days, the government understood clearly that what people were willing to pay for was letters, and that newspapers were being subsidized. The Post Office would have thrived on letters alone, but would have gone bankrupt instantly had it been forced to survive on newspaper deliveries. Thus content was king in the minds of policy makers, but it was definitely not king in terms of what people were willing to pay for. That is similar to the current situation. However, this differential in willingness to pay does not seem to be understood as well today as it was then. Preoccupation with content has historically been common. For example, it was often thought (even by Alexander Graham Bell) that one of the principal uses of the telephone would be in broadcasting [deSolaP1, deSolaP2]. Several substantial experiments in delivering content over the phone were attempted, including the Telefon Hirmond\'o in Budapest that lasted from 1893 past the end of World War I, and the Telephone Herald in Newark, New Jersey, which folded soon after its start in 1911 [deSolaP1]. In the end, though, the phone emerged as the prototypical example of point-to-point communication. Radio initially evolved in the other direction. It started out as a point-to-point communication technology, a wireless telegraph. % an improvement on the telegraph. After about two decades of experimentation, it became primarily a broadcast medium. (For the history of this transformation, see [Douglas, Smulyan].) However, the role of radio in the economy as a content delivery technology is tiny compared to that of the telephone, as Table 2.1 shows. Further, in the last few decades, with the development of cellular services, radio transmission has started to move back to its roots as a point-to-point communications service. The revenues from wireless telephony already far exceed those from radio broadcasting, as Table 2.1 shows (\$33.5 billion versus \$13.5 billion in 1997, with the disparity much greater today). A skeptical reader might say that all this historical stuff is amusing but irrelevant. We live in the 21st century, and our high-tech present as well as our future are on the Web, where content is universally regarded as king. Studies of the Internet regularly find that Web traffic makes up 60 to 80\% of the bytes that are transmitted. % Further, some studies of Web traffic % show appearance of the winner-take-all phenomenon, with % a few sites dominating [AdamicH]. Certainly most of the commercial development effort on the Internet and almost all the attention are devoted to content. Thus even if content was not king in the early 19th or late 20th centuries, it might be king in the 21st. There are three counterarguments to the above objection, all of which support the ``content is {\em not} king'' thesis. All argue that the dazzling success of the Web has created a misleading picture of what the Internet is, or is likely to evolve towards. One argument, to be discussed in more detail later, is that the future of the Internet is not with the Web, but with programs like Napster or (even more, because of its decentralized nature) Gnutella, which allow for informal sharing of data. The second argument is that content is not king of the Web. Most of the traffic on the Internet is corporate (especially if we include internal intranet traffic that is not visible on the public backbones). It is likely that in early 2000, no more than a quarter of the volume went to residential users [CoffmanO, CoffmanO2]. Intranet traffic appears to be much less heavily biased towards the Web than that of private individuals. Furthermore, even the traffic that appears to be Web-based frequently represents a variety of database transactions that are not properly speaking ``content.'' Because browsers are a user-friendly tool that is ubiquitous, a multitude of services have been squeezed into a Web framework, and help perpetuate the image of the Internet as primarily a content-delivery mechanism. The third and final argument is that even if content were king on the Web now, the Web is {\em not} king of the Internet. This may again seem absurd, especially in view of the statistics quoted above, that most of the Internet traffic is Web transfers. However, consider again the U.S. postal system of 1832. Content certainly dominated in terms of volume of data. Newspapers sent by mail weighed about 20 times as much as letters. Further, the density of printed matter is higher than of handwriting, and a typical copy of a newspaper was likely read many more times than a typical letter. Hence newspaper ``content'' was probably delivering at least a hundred times as much information as letters. But volume is not the same as value. Letters were bringing in 85\% of the money needed to run the postal system in 1832. On the Internet in 2000, it is email that is king, even if its volume is small. Today, Web traffic dominates the Internet in volume, with about 20 times as many bytes as email. (Netnews traffic volume is of about the same order of magnitude as email. The fractions of total traffic created by these two services vary from link to link and from day to day.) Even a decade ago, before the Web, email typically accounted for under 10\% of Internet volume. Yet email has been and continues to be the real ``killer app'' of the Internet. The ARPANET (the progenitor of the Internet) was built primarily to connect computers. Yet email quickly emerged as the application that mattered the most to users, even in the early days of the network. This was much to the surprise of the system's designers [LickliderV]. More recent surveys (e.g. [KatzA]) show that email is still the most valuable service. Ask people whether they would rather give up email or the phone, and the responses will typically be split. However, when a similar choice is offered between the Web and email, there is no contest. This is true for both individuals and large organizations. Intranets are all the rage, but it is email that makes enterprises run. The perception that content dominates the Internet is fueled by studies such as [AdamicH], which show the winner-take-all phenomenon, with a few sites dominating Web transactions among residential users. However, one should not read too much into such results. In the early 19th century postal system, studies of usage of information would undoubtedly have reached conclusions similar to those of [AdamicH]. Most data from distant locations that people consumed came from newspapers. Further, circulation figures of individual newspapers probably followed the standard Zipf-type distribution, with the most popular papers attracting a disproportionately high fraction of readers. Yet that does not say much about the value derived from the postal system, which was elsewhere, in letters. To reemphasize the importance of point-to-point communication in the online environment, consider the disappointing reception that WebTV has had. It seems that inexpensive Web browsing is not such a great attraction by itself. Also consider the innumerable failures in teletext experiments (cf. [Ettema, Greenberg, Klopfenstein]), as well as the initially promising start but disappointing end of the French Minitel (dealt with in more detail in Section 16). Their inadequate or even totally missing facilities for point-to-point communication appear to have been fatal errors. Next, consider the fates of CompuServe and Prodigy. Set up primarily for database access and online shopping, respectively, both were forced to emphasize communications and are now basically standard ISPs. Finally, the currently most successful of the public online services, AOL, started out as a game network. The figure in Table 2.1 showing small video game spending explains why that approach was doomed to failure. (The game market is growing, and in any case is larger than the figure in Table 2.1, which covers just the home part of it. It can certainly contribute to profits of large networks, or support some specialized service providers. However, it simply cannot fund anything as large as AOL.) AOL survived and prospered because it was nimble. After several changes in strategy, it partially opened itself up to the Internet. While it has content of its own, and access to the Internet, the majority of the time its subscribers spend online is devoted to email and chat. What this argument suggests is that the Web (and browsers in particular, which made the Web user-friendly) may have created a misleading impression. By focusing attention on centralized delivery of content, the Web may have prevented a proper appreciation of the importance of the often chaotic and generally unplannable point-to-point communications. The Web and the browsers may have played two main roles. One was to force online service providers to accept an open interoperable standard that made the entire Internet accessible for communications for everyone. The other was to introduce a user-friendly graphical interface for email, chat, and netnews, which made such communications easier. However, the Web is probably not as important to the Internet as is commonly thought. As an example of the relative value of content and simple pipes, note that the revenues of the entire cable TV industry in 1997 were only slightly higher than for the cell phone carriers. Furthermore, as Table 2.2 shows, cable TV has been growing far more slowly than the wireless industry. (See Table 3.5 for more detailed historical statistics for wireless communication.) Yet cell phones currently provide primarily simple, low bandwidth pipes. By early 2000, even the gross revenues of radio telephony exceeded those of the cable TV industry. Further, a large chunk (estimated at a quarter to a third) of cable revenues was devoted to paying for content, so in terms of basic network revenues, the cellular industry had pulled ahead even before 1997. The comparison is even more favorable to low bandwidth wireless pipes when we go outside the U.S. In other industrialized countries, cell phones are often much more widely used, while cable TV penetration is almost always far lower than in the U.S. Thus on a worldwide scale, the comparison is skewed even more heavily against content. The cable TV industry does have excellent prospects for faster growth. However, that growth will surely come more from improved communication capabilities, and less from content. \begin{table}[htb] \begin{center} Table 2.2. Revenues of U.S. cable TV and cell phone industries. \\ ~ \\ \begin{tabular}{crr} year & \multicolumn{1}{c}{cable TV} & \multicolumn{1}{c}{cellular} \\ ~ & \multicolumn{1}{c}{(millions)} & \multicolumn{1}{c}{(millions)} \\ ~ \\ 1987 & \$11,563~~ & \$942~~~~\\ 1992 & 21,079~~ & 6,688~~~~\\ 1997 & 30,784~~ & 25,575~~~~\\ \end{tabular} \end{center} \end{table} The general conclusion is that content has been less important than point-to-point communication in the past, including the recent past involving the Internet. Still, the argument that ``content is {\em not} king'' that is presented here should not be taken to an extreme. All it says is that most of the money is in point-to-point communication. It does not say that content does not dominate in volume of data. Historically, as we have noted above, content has often dominated, and probably dominate now. (There is some uncertainty, since there are difficult questions about measuring the volumes of broadcast communications.) There are arguments, to be presented later, that in the future, content will not provide most of the bits traveling on the Internet. However, even if that prediction is wrong, it will not affect the argument, which is about value to customers, and not about volume. That content is not king does not mean that content is unimportant in shaping political or social views. The attention paid to writers, and the political advertising on radio and TV, testify to the influence of content. This also is not a new phenomenon. Over the centuries, millions of people based their opinions of Richard III on Shakespeare's play, just as today millions base their opinions of John F. Kennedy's assassination on Oliver Stone's film. The argument about the value of content says little about the dispute in early 2000 between Disney and Time Warner over carrying ABC channels on cable TV. That issue is about division of revenues between content creation and content distribution. The argument of this paper is that the entire content piece of the economy is not all that large, and its contribution to network costs is much smaller than that of point-to-point communication. It does not deal with how the content piece is divided. Content can be profitable. Numerous media companies are doing very well. Content can also be of value to a network, even aside from providing traffic for the network to carry. However, it is probably best to think of content as either catnip or icing on the cake; something to attract new users, or enhance user experience. That is what broadcast TV programs do for the advertisers who pay for them. That may also have been the main role of the Web and browsers in bringing more people to the Internet. What the argument that content is not king does say is that people are willing to pay far more for point-to-point communication than for the famed content. That is likely to be reflected in what kinds of networks are built, and which companies succeed. It inverts the usual ordering of priorities, making point-to-point communication central, and content secondary. The fights over control of movie distribution may be a distraction from the main business of communication. As a simple example of what this may mean in practical terms, most broadband access links, such as cable modem and DSL ones, are designed to be asymmetrical, with higher capacity on the link to the home than to the network. The expectation is that these connections will be used primarily to pull content to the consumer. However, if the consumer places much higher value on personal communication than on content, the case for symmetrical connections becomes stronger. That may mean that fiber to the home may be justified sooner than expected. In the wireless arena, it suggests (as is explained in greater detail in Section 15) that for much of the next decade, the best strategy will be to emphasize regular voice transmission, supplemented by email and various low-bandwidth data transfers. Music and video services are likely to be delayed until much later. Before continuing, it is worth considering a basic issue, namely, what is content? This word derives from the Latin ``contentum,'' which means ``that which is contained,'' but this derivation is not very descriptive. There is no precise definition, but generally content is used to denote material prepared by professionals to be used by large numbers of people, material such as books, newspapers, movies, or sports events. That is the sense in which it is used in this work. In general, content is distributed by ``mass'' or ``broadcast'' communications systems. Until a few decades ago, such services could be distinguished easily from ``point-to-point'' (or, more precisely, ``person-to-person'') communications, which included first class letters and phone calls, and were specific to the people involved in the transaction. These two types of communications were sometimes combined during distribution, as in the postal system, which carried both letters and newspapers, in an early example of ``convergence.'' However, there was a noticeable distinction in how these two types of communication were prepared, handled, perceived by the recipients, and (a point discussed at great length already) in how much people were willing to pay for them. During the last few decades, the distinction between point-to-point and broadcast communication began to blur. Computers allowed for the mass preparation of personalized letters offering credit cards, say. Answering machines and voice response systems led to machine-mediated point-to-point communications. Individuals were able to reach large audiences through postings to netnews, or, more recently, through their personal Web pages. We can expect this evolution of communications to continue, and eventually to achieve that convergence in which there will be a continuum between point-to-point and broadcast communication. However, we are not there yet, and won't be for a while. In this work I do not classify information services such as weather, directory assistance and airline schedules as content. Many of the standard phone calls access just such services, and the Internet is leading to increasing usage of them. I also do not classify most of ecommerce as content. Somebody going to the Godiva Web site may be exposed to creative work in the ads flashed on the screen, but is interested in purchasing a tangible good. These types of interactions will flourish on the Internet, and some will be merging with content, but they are more typical of the standard point-to-point communications. One of the main lessons from the history of communications is that technologies are often adopted rapidly, but seldom at the astronomical rates that popular imagination associates with the Internet. This applies even to the Internet, where change tends to be less rapid than is often thought. Browsers were adopted rapidly. The first informal release of Mosaic took place in the spring of 1993, and in under two years, the majority of Internet traffic was Web-related. However, that was an exception. Other changes have been slower. Just consider Internet telephony, introduced back in 1995. It is finally beginning to have a noticeable effect, but it is far from dominant. As another example, Amazon.com has had a striking impact on perceptions of ecommerce. Yet after more than four years, less than 10\% of book sales take place online, and Amazon.com's investors are learning the virtues of patience. ``Internet time'' is a myth. In general, the time a new technology takes to become widespread has not changed much in the last half century. \begin{quote} A modern maxim says: ``People tend to overestimate what can be done in one year and to underestimate what can be done in five or ten years.'' \hspace*{+3in}(footnote on p.~17 of [Licklider]) \end{quote} Even technologies with compelling advantages tend to take a decade to dominate markets. Fax machines took about 10 years, % roughly the decade of the 1980s, from the introduction of the first inexpensive models until they became ubiquitous. Cell phones, one of the fastest growing industries, have taken about 15 years to reach their present level (cf. tables 2.2 and 3.5). Cable TV has taken three decades to reach about 60\% of U.S. households. Music CDs and more recently DVDs show similar patterns, taking on the order of a decade to reach dominance. For more examples and a discussion of rates of change, see [Odlyzko3]. Aside from the unusually rapid ascendency of the browser-Web combination, fast change is usually associated with the presence of forcing agents. These can be either governments or a small number of key decision makers who can shift an industry's direction. An example of such forcing agents are the information technology managers at large banks and other enterprises. When they decided that mainframes were obsolete, they did so over a short period of time, and this led to a catastrophic decline in IBM's fortunes. A transition of voice phone traffic from circuit switched transmission to the Internet might occur rapidly for just this reason. The carriers might be able to implement it rapidly, since it could be done essentially invisibly to the end users, and the decisions would need to be made by only a few people. The decade-long diffusion periods listed above for consumer goods and services are due to the inertia of the millions of people who have to individually decide to adopt a new technology. Most new products and services are in that category. Sociological changes are even slower, taking a generation or two. Normal change, with a simple shift in technology that offers serious advantages over older, more established competitor (as with CDs over vinyl records, which provided higher quality sound reproduction, or cell phones, which offered mobility, even at the cost of sound quality), takes on the order of a decade. % An interesting lesson from many technologies, but % especially communication ones, is that they often % decline at a much slower rate than they rise. % For example, the telephone took a long time to % displace the telegraph, and the Internet has yet % to make much of an impact on fax use. The speed with which new technologies diffuse has a direct relevance for the question of whether content might be king of the Internet. It is possible to make a case that even if content is not king now, it might be king in the future, when convergence does occur. Some evidence for this can be derived from the comparison of cable TV and cell phone industries. Their revenues and growth rates were cited as demonstrating that point-to-point communication is more important than content. However, there is another way to look at the data. As was mentioned, about a third of the cable TV revenues go to pay for content. What that means is that two thirds pay for the network. (According to some reports, carriers receive up to 90\% of the revenues from some content, such as hard-core pornography pay-per-view movies.) If convergence moves delivery of most of the content to the Internet, and transport grabs two thirds of the total revenues from content, then the network will get a huge new source of revenue. Total content spending in the U.S. is comparable to that on the phone system, so two thirds of that would make a huge difference to network financing. This scenario is not totally implausible, since the current content distribution system is grossly inefficient. Book authors and musicians typically receive in royalties less than a tenth of the price consumers pay for their creations. It is a striking observation that a participant in Amazon.com's affiliate program can sometimes get more money from a book sale generated by a link from his or her home page than the author does from royalties! The Internet offers a chance to reduce some of the inefficiencies of the current system. For the current dominant content producers, the real threat from the Internet is probably less from piracy, and more from disintermediation. New producers can come in and, unburdened by the high overheads and obsolete habits of established players, can offer better deals to both the creative talent and the consumers. The scenario outlined above, in which Internet transport grabs the lion's share of revenues from content, is conceivable, but very unlikely. One argument against this scenario is based on a simple historical observation. No broadcast medium has ever been replaced by another; despite predictions to the contrary at various times in the past, newspapers were not killed by radio, nor radio by television. However, that argument may not be valid. The Internet is a disruptive technology, it does have unprecedented ability to emulate other delivery mechanisms, and we are already seeing rapid growth in music delivery on it. As an example of what can happen to even the most solid-seeming businesses, just consider the {\em Encyclopaedia Britannica}. (Its problems started even in the pre-Internet days, with CD-ROMs.) In spite of having the greatest brand name and by far the best content in the encyclopedia field, it has been floundering, and has yet to find a viable business model. Thus dramatic changes are indeed possible in the electronic environment. % on the Internet. % (The problems of the {\em Encyclopaedia Britannica} % started even in the pre-Internet days, with CD-ROMs.) % It is a disruptive technology. Still, there are other, more substantial, arguments against the content thesis. The Internet will surely have a major impact on the content industries. However, as was discussed above, consumers % the scenario above, with Internet content discribution % getting most of the revenues, is implausible. % The reason is that, as discussed above, consumers are slow to change their behavior. Even the {\em Encyclopaedia Britannica} has had a decade in which to flounder. Couch potatoes are not going to abandon their TV sets right away, especially when computer and cable TV penetrations in the U.S. are under two thirds of households, and not growing rapidly. Even in the business environment, adoption of new technologies that require a thorough re-engineering of all internal processes is slow. The business-to-consumer dot-coms have discovered this already, and the business-to-business ones are in the process of learning. Although we are living in the Internet era, fundamental change is not all that much faster than a century and a half ago. The adoption of the telegraph by railroads, discussed in Section 13, did lead to huge efficiency gains, but it was slow as well. Slow adoption of new technologies means that convergence will be spread over a decade or more, and there will be continuing competition from traditional media, as well as increasing diversity of delivery mechanisms for content. This may mean that the writers and artists will get a bigger share of the pie. % do better than the delivery industries. That appears to have been the trend over the last few decades, with movie actors and professional sport stars increasing their share of the revenues their work brings in. It is less certain whether carriers will manage to improve their share of the content pie to the same extent. There will certainly be a shift of revenue towards broadband services, but content distribution may not be the largest contributor to it. The main reason to question whether content will ever make giant contributions % be skeptical about content's contribution to network costs is that by the time convergence is likely to occur, at least a decade into the future, content transmission is likely to be a small fraction of total traffic. Further, most content will probably be distributed as ordinary file transfers, not in real-time mode. The various rates of growth that contribute to these predictions are discussed later. Right now we note that if these predictions come true, then it will be hard for networks to charge much for the transport of content. High prices could be charged for content distribution if content made up most of the traffic, or else if content required special transmission quality (such as that needed for real-time streaming traffic). Since neither of these conditions is likely to be satisfied, though, content will probably constitute just some of the huge number of large files, many encrypted, that will be flying around the network. How could carriers pick out the content files for special pricing? Let us next consider the predictions for Internet traffic mentioned in the preceding paragraph. I will treat them briefly, and refer the reader to [CoffmanO2] for more detail. (There is some discussion of this point in Section 4 as well.) Internet backbone traffic appears to be about doubling each year. Further, advances in photonics and additional fiber deployment appear to allow for a doubling of network capacity each year for the next decade. % (Even faster growth, on the order of an annual trebling, % might be possible for a few years.) In early 2000, Internet backbone traffic in the U.S. appeared to be less than a third of voice traffic in volume, but is likely to become larger by about 2002, with the transition point slightly later for the rest of the world. However, if voice traffic were to be packetized, it would almost surely be compressed, and then its volume would already be less than that of Internet traffic (at least in the U.S.). Broadcast TV would still overwhelm the Internet of the year 2000, but at 100\% annual growth rates, it will not be too long before there is more than enough capacity to provide a high quality video channel for every person on Earth. This is because bandwidth that is likely to be required to satisfy any single person's real-time transmission needs will not be increasing fast. % Resolutions of displays are not advancing as rapidly as % processor power or transmission. The versions of Moore's Law that hold in different industries operate at diverse rates. Microprocessors are doubling in computing power every 18 to 24 months, while fiber transmission capacity is doubling every year. On the other hand, display technology is advancing extremely slowly. Broadcast TV resolution has been static for several decades, and even the planned move to HDTV will require only a modest increase in bandwidth. % Even computer display technology is advancing slowly. Thus satisfying the needs for real-time multimedia transmission will not require much of the Internet's capacity. Content is likely to form only a small fraction of Internet traffic for reasons explained above. In addition, real-time delivery of content is likely to be an even smaller factor, for reasons discussed in much greater detail in [CoffmanO2]. Transmission capacity is approximately doubling each year, which is a much faster rate of improvement than Moore's Law for semiconductors. However, hard disk storage capacity is also about doubling each year. Furthermore, that capacity is already huge. At the beginning of 2000, the U.S. Internet backbones appeared to be carrying about 12,000 TB (terabytes) of traffic each month. However, the total world hard disk capacity was 3,000,000 TB. Thus it would take about 20 years to transmit all the data on those disks over U.S. Internet backbones. Nobody proposes to do that (and why would anyone want to send all the duplicate copies of Windows 98 around, in any case?), but this comparison helps in visualizing the technology landscape. The presence of huge local storage capacity in local PCs or cable TV setup boxes will make it much more attractive to send even content as files, not as real time transmissions. There will be a growing volume for real time multimedia traffic, for applications such as videoconferencing. However, such applications are likely to be swamped by ordinary file transfers. The dominant mode of operation is likely to be fast (eventually much faster than real time, but initially often slow) download to local storage, fast transfer to whatever display device one wishes to use (often a mobile information appliance), and then playback. That is already the model we see emerging with MP3, Napster, and TiVo. The advantages of this model include the ability to implement it now, before the Internet can be made ready for real-time streaming media. It also accommodates gracefully the forecasted explosive growth in small mobile devices, which will often have small storage and low bandwidth over wireless links, and thus will be most useful if they can get data from local storage. This model also allows for easy integration with special hardware for intellectual property protection. That real-time multimedia traffic would not dominate the Internet has been predicted several times in the past. It is an obvious conclusion from the rapid increase in traffic. Already the authors of [deSolaPITH] noted that in the early 1980s data traffic was growing much faster than voice communication. They observed that if that trend continued, eventually most transmissions would not be seen by human eyes nor heard by human ears. In a similar vein, in discussing general digital data volumes in 1997, Michael Lesk predicted that ``the typical piece of information will {\em never} be looked at by a human being'' [Lesk]. Bill St. Arnaud appears to have been the first one to predict in the specific context of the Internet that the general expectations for streaming multimedia domination were unlikely to come true [StArnaud, StArnaudCFM]. Further arguments were presented in [Odlyzko8]. The huge volume of local storage will surely stimulate the generation of non-content traffic. Both corporations and individuals so far have had no difficulty filling their disks with data. We can expect this to continue, although predicting the exact source of that data is uncertain. For firms, various databases will likely continue to proliferate and grow. For residential users, pictures are the leading candidates for filling those disks. Ease of use, lower cost, and instant gratification all stimulate use, and digital camera owners appear to be taking many more pictures than they ever did with regular film. % Eventually, as memories grow, we can expect digital video cameras to provide the % bits to help fill those growing hard disks. The same electronic technology that is producing better disks and processors is also producing better cameras. Historically, it appears that privately taken pictures have traditionally been the dominant source of data. An interesting accounting of all the information stored in the world in 1997 by Michael Lesk [Lesk] found that home photographs were the dominant component. They contributed about 500,000 TB each year (even when one assumes that each picture is stored as a modest 10 KB JPEG file). By comparison, all the texts in the Library of Congress amounted to around 20 TB, while the graphics and music in that collection came to about 3,000 TB. Thus even this great library contained less than 1\% of the world's information. (The publicly accessible Web pages currently contain a few tens of terabytes, just a few percent of what the Library of Congress has, but comparable to the text collections in that library.) An obvious comment to the estimates above is that the purpose of a library is to select the most valuable material, and that most of those photographs contributing to the 500,000 TB are of no interest to most people. That is true, but that does not stop those pictures from being taken, and it will not stop an explosion in volumes of data collected this way in the future. A few pictures or video clips will turn out to be of great interest, in spite of amateur production. Just think of the Zapruder film of the JFK assassination, or the Rodney King video. More importantly, many of the pictures being taken are of interest, or might be of potential interest, to at least one person. Most of the world will have no interest in a picture of your newborn baby, but your mother will cherish it. Similarly, in the future you will be taking digital video clips of your children and sending at least some of them to your mother. Many of the video shots will intentionally be made with the hope of not having to see them, as with security monitors. (Note that some of the earliest applications of miniature taperecorders and video cameras has been to snoop on child care providers. There are obvious privacy implications of current and future camera technologies that are not pursued here, but are discussed in [Garfinkel], for example.) Other examples of data that may be filling our disks are suggested by the entry in Table 2.1 for sporting goods. Spending in this area comes to about half of consumer spending on content, showing how highly these products are valued. One can easily imagine future generations of body and equipment sensors that would record precisely all details of a player's movements in tennis, say. These details would then be fed into systems that would analyze the motions, compare them with previous games, and produce high quality graphical displays to help improve the player's game. There is practically no bound on the amount of data that could be generated this way. The data that will be generated is likely to be shared using programs descended from Napster. Email and the Web may not be flexible enough. Napster is currently attracting huge attention because of the threat it poses to conventional music distribution channels. However, that may turn to be less important than its ability to facilitate sharing of files. Napster itself is too limited, as it is designed to handle just MP3 music files, and is also centralized. Yet it has already inspired creation of tools such as Gnutella, which are much more general and decentralized. Given the growth of local storage, and the increasing availability of tools to fill that storage with video clips and other material, it is possible that tools like Gnutella may become more important to the Internet than the Web. The discussion above is futurology. We cannot be certain how the Internet will evolve. However, history teaches us several lessons, which are discussed in greater detail in later sections. One is that the growing storage and communication capacities will be used, often in unexpected ways. (For a careful study of the many early predictions about the future of the telephone, and the actual evolution of that service, see [deSolaP1, deSolaP2], for example.) Another important lesson is that the value of the myriad social interactions has often been underestimated. Only a tiny fraction of the information passing through communications systems has ever been high quality scholarly knowledge. Even in more prosaic transmissions, we have moved from Samuel Morse's solemn ``What hath God wrought?'' to Alexander Graham Bell's utilitarian ``Mr. Watson, come here, I want you,'' to the banal ``How was your lunch?'' that is so common today. The volume of communication has increased, the importance of a typical message has decreased, and the attention we pay to such a typical message has decreased. However, the aggregate value of all these exchanges has increased. As is described in more detail in later sections, sociability was frequently dismissed as idle gossip, and especially in the early days of the telephone, was actively discouraged. For example, a 1909 study of telephone service commissioned by the city of Chicago advocated measured rate service as a way to reduce ``useless calls'' [JacksonCW]. Yet the most successful communication technologies, the mail and the telephone, reached their full potential only when they embraced sociability and those ``useless calls'' as their goal. That seemingly idle chit-chat not only provided direct revenues, but it encouraged the diffusion of the corresponding technology, and made it more useful for commercial and other applications. Such social interaction frequently function to grease the wheels of commerce. This work is largely the story of the development of the communications infrastructure from the point of view of the user. Usually an infrastructure is noted for being unnoticed; it is simply there, something we come to rely upon, do not have to think much about, and are horribly inconvenienced when it malfunctions. Electricity, water, mail, and the phone are excellent examples. Yet they all took much effort to reach this stage. The key failing of the telegraph, as is detailed in later sections, is that it never became a true infrastructure component. It was a revolutionary technology, the ``Victorian Internet,'' as one writer has called it [Standage], but it was too cumbersome and too expensive to attract much usage, and in particular never carried much of that ``useless'' social traffic that pushed the mail and the telephone to their eminent positions. Although social uses are important to the the telephone industry, a glance at Table 2.1 shows that most of the money comes from businesses. Consumer spending on phone service brings in only about a third of the total revenues. (The figures for total revenues, \$256 billion in 1997, and consumer spending, \$85 billion, come from different sources. It is possible that consumers spend somewhat more, especially for cell phones, than is reported in the \$85 billion figure. However, even if one makes the most likely adjustments, it still appears that business spending on telephony is far larger than that of households.) That has been the historical trend, and many communication services, including the phone, were initially devoted almost totally to business uses. Traditionally, commercial users have subsidized residential ones. Sometimes this was done even involuntarily, as in higher rates dictated by carriers or by government regulators, and sometimes voluntarily, as in paying for toll-free 800 numbers. One of the purposes of this work is to examine the history of such subsidies. It appears probable that they will also play a large role on the Internet. We may very well end up with a system in which the largest monetary contribution will come from commercial users, the second largest for households paying for point-to-point communication, and the smallest by the transport component of charges for content. The value of a broadcast network is usually regarded as proportional to the number of users in it. On the other hand, a point-to-point communication network is often said, by Metcalfe's Law, to have value proportional to the square of the number of member. This then leads to the conclusion that eventually, once a single network like the Internet reaches a large enough size, point-to-point communications will provide much higher value than broadcast. There are some problems with this argument. In particular, Section 4 shows, using historical evidence as well as general principles, that Metcalfe's Law does not reflect properly several other important factors that go into determining the value of a network. However, the general thrust of the argument and the conclusion are valid. Certainly all the historical evidence cited throughout this work supports the conclusion that point-to-point communication is what matters the most. Whether content is king or not has direct relevance for the question of whether the Internet will continue to be an open network, or whether it will be balkanized. If content were to dominate, then the Internet would be primarily a broadcast network. With value proportional to the number of users, there would be few inherent advantages to an open network. The sum of the values of several completely or partially separate networks would be the same as of a unified network. On the other hand, if point-to-point communications were to dominate, and if Metcalfe's Law were to hold, there would be strong economic incentives to a unified network without barriers. This is considered more fully in Section 4. The general conclusion there is that even though Metcalfe's Law is not fully valid, the incentives to maintain an open network are likely to be very strong. This will be largely because content is not king, and effective point-to-point communication will demand easy interconnection. An extreme form of the ``content is king'' position, but one that is shared by many people, and not just in the content industry, was expressed recently by the head of a major music producer and distributor: \begin{quote} What would the Internet be without ``content?'' It would be a valueless collection of silent machines with gray screens. It would be the electronic equivalent of a marine desert - lovely elements, nice colors, no life. It would be nothing. \hspace*{+3in}[Bronfman] \end{quote} The author of this claim is facing the possible collapse of his business model. Therefore it is natural for him to believe this claim, and to demand (in the rest of the speech [Bronfman]) that the Internet be architected to allow content producers to continue their current mode of operation. However, while one can admire the poetic language of this claim, all the evidence of this paper shows the claim itself is wrong. Content has never been king, it is not king now, and is unlikely to ever be king. The Internet has done quite well without content, and can continue to flourish without it. Content will have a place on the Internet, possibly a substantial place. However, its place will likely be subordinate to that of business and personal communication. \section{Growth in communications: quantity and quality} \hsp The explosive rise of the Internet is only the most recent chapter in a remarkable history of humanity becoming increasingly connected, the ``annihilation of space and time,'' in a phrase going back at least to the early 19th century. The volume of messages addressed individually to each one of us has increased about a thousand-fold over the last two centuries. Much of this growth has been slow, and accomplished through the power of compounding. However, growth rates rivaling that of the Internet have been observed occasionally before, for example in the development of the telegraph. The wireless industry is experiencing similar growth even now. This section surveys the long trend of increasing communication. The first part is devoted to quantitative measures, the growth in spending and in volume of transactions. The second part considers the qualitative changes in how we approach communications. The role and evolution of pricing are then the subject of the rest of the paper. The communications technologies considered in this paper are primarily point-to-point ones, and generally exclude mass media, such as newspapers, book publishing, radio, and TV. One reason is to keep the size of this work manageable. A related one is to avoid the complexities of trying to measure the volume of information delivered by broadcast media. With point-to-point communication, it has often been true that the entire content was read, heard, or seen by the recipient. With broadcast communication, accounting is harder. If 70 channels come across on a cable TV coax, should one measure the information on all of them all the time, just when the TV set is turned on, or just the channel that is watched when the set is on? Should one also consider the fraction of the attention that the people in front of the set are devoting to the screen, as opposed to talking to each other? It is not easy to come up with a fair measure. (See [deSolaPITH] for an attempt to do this. Also note that, as was indicated in Section 2, similar problems will increasingly arise on the Internet, as a diminishing fraction of the traffic passing through it is seen or heard by people.) The third, and most important reason for concentrating on point-to-point communications is that, as was shown in Section 2, it is far more important than broadcast. \begin{table}[htb] \begin{center} Table 3.1. Growth of the U.S. Postal Service \\ ~ \\ \begin{tabular}{lr@{}lr@{}lr@{}lr@{}l} year & \multicolumn{2}{c}{expenditures} & \multicolumn{2}{c}{expenditures} & \multicolumn{2}{c}{pieces of mail} & \multicolumn{2}{c}{pieces of mail per} \\ & \multicolumn{2}{c}{(millions)} & \multicolumn{2}{c}{as percentage} & \multicolumn{2}{c}{(millions)} & \multicolumn{2}{c}{person per year} \\ &&& \multicolumn{2}{c}{of GDP} \\ ~ \\ 1790 & \$0&.032 & ~~~~0&.02\% & 0&.8 & 0&.20 \\ 1800 & 0&.214 & ~~~~0&.05 & 3&.9 & 0&.73 \\ 1810 & 0&.496 & ~~~~0&.09 & 7&.7 & 1.&.07 \\ 1820 & 1&.161 & ~~~~0&.18 & 14&.9 & 1&.55 \\ 1830 & 1&.933 & ~~~~0&.21 & 29&.8 & 2&.32 \\ 1840 & 4&.718 & ~~~~0&.28 & 79&.9 & 4&.68 \\ 1850 & 5&.213 & ~~~~0&.20 & 155&.1 & 6&.66 \\ 1860 & 14&.87 & ~~~~0&.39 \\ 1870 & 24&.00 & ~~~~0&.33 \\ 1880 & 36&.54 & ~~~~0&.35 \\ 1890 & 66&.26 & ~~~~0&.51 & 4,005& & 63&.7 \\ 1900 & 107&.7 & ~~~~0&.58 & 7,130& & 93&.8 \\ 1910 & 230&.0 & ~~~~0&.65 & 14,850& & 161 \\ 1920 & 454&.3 & ~~~~0&.50 & \\ 1930 & 803&.7 & ~~~~0&.89 & 27,887& & 227 \\ 1940 & 807&.6 & ~~~~0&.81 & 27,749& & 211 \\ 1950 & 2,223& & ~~~~0&.78 & 45,064& & 299 \\ 1960 & 3,874& & ~~~~0&.77 & 63,675& & 355 \\ 1970 & 7,876& & ~~~~0&.81 & 84,882& & 418 \\ 1980 & 19,412& & ~~~~0&.70 & 106,311& & 469 \\ 1990 & 40,490& & ~~~~0&.70 & 166,301& & 669 \\ 1998 & 57,778& & ~~~~0&.68 & 197,943& & 733 \\ \end{tabular} \end{center} \end{table} The communications industry as a whole has been expanding for centuries. Not only that, but most of the prominent services have continued growing. The tables in this section provide more eloquent testimony to this historical fact than words can. Even the venerable mail, often derisively called ``snail mail,'' is still growing. The remarkable story of this ancient yet still widely used technology is summarized briefly in Table 3.1, which shows key statistics in the development of the U.S. Postal Service (USPS). (For most of its history it was known simply as the U.S. Post Office.) Although there have been many predictions that mail would decline as a result of competition from the telegraph, the telephone, and more recently from email, it has continued to expand. It went from fewer than a million pieces of mail in 1790 to almost 200 billion in 1998. The mix of traffic has changed. It was about half first class mail, and half newspapers in the early 19th century, if we count separate pieces delivered. It is about half first class and half ``junk mail'' today, and even a large fraction of the first class mail is targeted advertising. (It is estimated that only about a tenth of the first class letters are sent by households, see for example [Wolak].) Various new services have been added, such as money orders, home delivery, and parcel post. Even without the new services, the volume of letters alone has been increasing. Overall, total mail volume and spending have grown at impressive rates. Prices, discussed in Section 12, have decreased. The decline in prices was dramatic in the 19th century. Even in recent years prices have been decreasing, if we adjust them for increases in earnings. Moreover, this was not an isolated case of a frontier economy getting civilized. The French postal system in the 19th and early 20th centuries had similar growth patterns, as can be seen in Table 3.2. The increasing role of communication was common to all countries as they industrialized. (Table 12.1 shows data for the British mail system that behaved similarly, although there were major differences that are discussed in detail in Section 12. Analogous trends can also be seen in the statistics of the Swiss postal system presented in Table 21.1.) \begin{table}[htb] \begin{center} Table 3.2. Growth of the French postal system \\ ~ \\ \begin{tabular}{cr@{}lr@{}lr@{}l} year & \multicolumn{2}{c}{pieces of mail} & \multicolumn{2}{c}{pieces of mail per person} & \multicolumn{2}{c}{} \\ & \multicolumn{2}{c}{(millions)} & \multicolumn{2}{c}{per year} & \multicolumn{2}{c}{} \\ ~ \\ 1830 & 104& & 3&.2 & & \\ 1840 & 147& & 4&.3 & & \\ 1850 & 254& & 7&.1 & & \\ 1860 & 444& & 11&.9 & & \\ 1870 & 733& & 18&.8 & & \\ 1880 & 1,231& & 33&.0 & & \\ 1890 & 1,763& & 46&.3 & & \\ 1900 & 2,433& & 63&.4 & & \\ 1910 & 3,758& & 96&.1 & & \\ 1920 & 4,162& & 108&.1 & & \\ 1930 & 6,281& & 153&.2 & & \\ 1940 & 4,354& & 105&.7 & & \\ 1950 & 4,050& & 98&.1 & & \\ 1960 & \qquad6,093& & \qquad133&.9 & & \\ \end{tabular} \end{center} \end{table} USPS revenues went from 30 thousand dollars (in current dollars, not adjusted for inflation) in 1790 to almost 60 billion today, and are still increasing. Such nominal dollar measures ignore the effects of inflation and economic growth. Hence it is more informative to compare the postal system to the whole economy. We find that spending on mail in the U.S. grew from 0.02\% of the Gross Domestic Product (GDP) in 1790 to almost 0.9\% in 1930. Since then, it has declined to about 0.7\%. However, this decline is misleading, since the statistics in Table 3.1 do not include competing private delivery services. UPS has taken the lion's share of the parcel business from the USPS. (Parcel delivery does not count as communication by the standards of this paper, but it is hard to disentangle it from regular mail. Parcel post was introduced by the USPS in 1913, and contributes significantly to its budget, although not to the extent it did before UPS made its inroads.) The annual revenues of UPS are around \$27 billion. Further, express delivery companies such as FedEx, DHL, and Airborne Express are largely in the message communications business. FedEx, the largest by far, has annual revenues around \$17 billion, with about \$12 billion of that from U.S. traffic. If we added up the contributions of all these express services, we would likely find out that the fraction of U.S. GDP devoted to basic mail services is still growing. The USPS, like other countries' postal systems, and even the express delivery companies, does face challenges, particularly from the growth of email and the Internet in general. Those challenges also represent opportunities, though, as ecommerce still requires physical delivery of most goods. % Even in the purely electronic realm, % digital certificates do require some physical verification, ... \begin{table}[htb] \begin{center} Table 3.3. U.S. domestic telegraph industry \\ ~ \\ \begin{tabular}{cr@{}llr@{}lc} year & \multicolumn{2}{c}{revenues} & \multicolumn{1}{c}{revenues as} & \multicolumn{2}{c}{messages handled} & messages per person \\ &\multicolumn{2}{c}{(millions)} & \multicolumn{1}{c}{percentage of GDP} & \multicolumn{2}{c}{(millions)} & per year \\ ~ \\ 1870 & \$6&.7 & \quad\quad\quad0.11\% & ~~~~~~9&.2 & 0.23 \\ 1880 & 10&.6 & \quad\quad\quad0.09 & ~~~~~~29&.2 & 0.58 \\ 1890 & 20&.1 & \quad\quad\quad0.15 & ~~~~~~55&.9 & 0.89 \\ 1900 & 22&.8 & \quad\quad\quad0.12 & ~~~~~~63&.2 & 0.83 \\ 1910 & 30&.7 & \quad\quad\quad0.09 & ~~~~~~75&.1 & 0.81 \\ 1920 & 124&.4 & \quad\quad\quad0.14 & ~~~~~~155&.9 & 1.46 \\ 1930 & 148&.2 & \quad\quad\quad0.16 &~~~~~~212&.0 & 1.72 \\ 1940 & 114&.6 & \quad\quad\quad0.11 &~~~~~~191&.6 & 1.45 \\ 1950 & 178&.0 & \quad\quad\quad0.06 &~~~~~~178&.9 & 1.18 \\ 1960 & 262&.4 & \quad\quad\quad0.05 &~~~~~~124&.3 & 0.69 \\ 1970 & \quad402&.5 & \quad\quad\quad0.04 & ~~~~~~69&.7 & 0.34 \end{tabular} \end{center} \end{table} Postal system revenues are more than three times larger as a fraction of GDP than they were 150 years. In spite of this, mail is no longer the dominant communication technology, as it is dwarfed by the phone industry. Comparison of tables 3.1 and 3.4 shows the telephone passed the mail in revenues by 1920, and today is about four times as large. Adding the contributions of these two industries, we find that their role in the economy, measured as a fraction of GDP, has increased almost 20-fold over the last 150 years, from 0.2\% to over 3.5\%. % The Internet and the telephone before it have often % been accused of dehumanizing our interactions. Yet most % of the spending has been on person-to-person contacts! % Technology has changed the nature of our interactions, % as will be discussed later, but they remain largely % personal ones. Even on the Internet, the ``killer app'' % remains email. \begin{table}[htb] \begin{center} Table 3.4. Development of U.S. phone industry \\ ~ \\ \begin{tabular}{lrlr@{}lr@{}lr@{}l} year & \multicolumn{2}{c}{revenues} & \multicolumn{2}{c}{revenues} & \multicolumn{2}{c}{phone calls per day} & \multicolumn{2}{c}{phones calls} \\ & \multicolumn{2}{c}{(millions)} & \multicolumn{2}{c}{as percentage} & \multicolumn{2}{c}{(millions)} & \multicolumn{2}{c}{per person per day} \\ &&& \multicolumn{2}{c}{of GDP} \\ ~ \\ 1890 & \$16 && ~~~~0&.12\% & 0&.0015 & 0&.00002 \\ 1900 & 46 && ~~~~0&.26 & 7&.8 & 0&.1 \\ 1910 & 164 && ~~~~0&.46 & 35&.6 & 0&.4 \\ 1920 & 529 && ~~~~0&.58 & 51&.8 & 0&.5 \\ 1930 & 1,186 && ~~~~1&.32 & 83&.5 & 0&.7 \\ 1940 & 1,286 && ~~~~1&.29 & 98&.8 & 0&.7 \\ 1950 & 3,611 && ~~~~1&.27 & 171& & 1&.1 \\ 1960 & 8,718 && ~~~~1&.73 & 288& & 1&.6 \\ 1970 & 19,445 && ~~~~1&.99 & 494& & 2&.4 \\ 1980 & 59,336 && ~~~~2&.13 & 853& & 3&.8 \\ 1990 & 133,837 && ~~~~2&.33 & 1,272& & 5&.1 \\ 1998 & 246,392 && ~~~~2&.88 & \qquad 1,698& & \qquad \quad6&.3 \\ \end{tabular} \end{center} \end{table} The tide of increasing communications volumes has been powerful enough to lift most services. The telegraph did not stop the mail from growing. The telephone took a long time to eliminate the telegraph. Wireless telephony and the Internet have been growing explosively, yet wired telephony is still growing. Even fax use is increasing. Yet if anything should have been eliminated by the Internet, it surely should have been the fax. Most messages sent by fax are typeset on a computer, printed out, and then fed into a fax machine. Why not eliminate some of the redundancy, and send the message via email? Well, that is beginning to happen, as is faxing directly from a computer. Still, the fax is ubiquitous, inexpensive, easy to use, and provides some features, such as signatures, that the Internet can and will eventually provide, but does not offer now. \begin{table}[htb] \begin{center} Table 3.5. Growth of U.S. cell phone industry \\ ~ \\ \begin{tabular}{cr@{}ll@{}r@{}r} year & \multicolumn{2}{c}{number of subscribers} & \multicolumn{3}{c}{revenues} \\ & \multicolumn{2}{c}{(millions)} & \multicolumn{3}{c}{(millions)} \\ ~ \\ 1985 & ~~~~~~~~~~0&.20 & \$ &354 &~ \\ 1986 & ~~~~~~~~~~0&.50 & &667 &~ \\ 1987 & ~~~~~~~~~~0&.89 & &942 &~ \\ 1988 & ~~~~~~~~~~1&.61 & &1,558 &~ \\ 1989 & ~~~~~~~~~~2&.69 & &2,480 &~ \\ 1990 & ~~~~~~~~~~4&.37 & &4,061 &~ \\ 1991 & ~~~~~~~~~~6&.38 & &5,076 &~ \\ 1992 & ~~~~~~~~~~8&.89 & &6,688 &~ \\ 1993 & ~~~~~~~~~~13&.07 && 9,009 &~ \\ 1994 & ~~~~~~~~~~19&.28 && 12,592 & ~\\ 1995 & ~~~~~~~~~~28&.15 && 16,461 & ~\\ 1996 & ~~~~~~~~~~38&.20 && 21,526 & ~\\ 1997 & ~~~~~~~~~~48&.71 && 25,575 & ~\\ 1998 & ~~~~~~~~~~60&.83 && 29,638 & ~\\ 1999 & ~~~~~~~~~~76&.28 && 37,215 & ~\\ \end{tabular} \end{center} \end{table} A technology that survived for longer than the fax (at least so far) is the electric telegraph. One can speculate that had the telephone not been invented, the telegraph might have spread more widely and become less expensive. Eventually it might have developed into something like our Internet, going from digital transmission directly to digital, bypassing the analog phone stage. There was intense development of new applications of telegraph technology in the second half of the 19th century for stock tickers as well as fire and home alarms [TarrFG]. However, these developments came to nought. The telephone started encroaching on the bread and butter business of the telegraph, and the latter could not compete effectively. Still, as Table 3.3 shows, usage of the telegraph kept growing almost to the middle of the 20th century. (The actual peak year for sending domestic telegrams in the U.S. was 1945.) Had Table 3.3 combined statistics for telegraph and telex traffic (since telex was to a large extent a substitute for the telegraph, and drove the latter into the ground), the catastrophic decline in volume of messages that is visible in the data would have been postponed. Instead of showing up during the 1950s and 1960s, it would have been visible only in the 1980s. It was only the fax that provided a way for the phone system to emulate telegraph and telex services, and eliminate those. The long survival of the telegraph (and the fax) suggests many lessons about rates of technological change. One that is particularly relevant for this paper is that efficiency often plays a minor role when relatively inexpensive goods or services are involved. The Internet can emulate other communication technologies much better than the phone could emulate the telegraph, say. Still, the power of inertia should not be underestimated. After all, why hasn't the Internet eliminated the fax yet? Where did the steadily increasing spending on communication that is documented in the tables come from? The general progression for new technologies % communication services has been to start as an expensive and limited service. Then, as a result of technological improvements as well as economies of scale, prices usually declined and utilization increased. Growth in usage was accomplished by development of new types of usage. For some very expensive services, governments were the first customers. As Samuel Johnson said, ``when a man knows he is to be hanged in a fortnight, it concentrates his mind wonderfully.'' The French revolutionary government of the early 1790s was fighting for its life, and did face the prospect of being hanged (or perhaps guillotined, were its enemies to turn its own invention on it) should it lose the war. This undoubtedly increased its willingness to fund Chappe's optical telegraph, mentioned briefly in Section 13. Even earlier, mail systems started out in most cases as government communications services. % Public use of communication services % After government use, the next step % in the evolution of a communication service % was usually public service. (For some technologies, such as the % electric telegraph and the telephone, this was the first step.) Some communication technologies, such as the electric telegraph and the telephone, went into public service right from the start. Initial usage was typically limited to special high-value commercial communications, since prices were high. In the Introduction, the sending of a letter in early 19th century England was compared (on the basis of earnings) to the purchase of a book today, costing about two hours of work. Table 13.1 shows that the first reliable commercial telegraph communication between New York City and London, established in 1866, cost \$100 for 10 words. That was several months' wages for a laborer then, comparable to the work time needed currently to pay for a round trip on the Concorde supersonic plane. By 1880, prices had decreased 20-fold, so that relative to earnings, the cost of a similar 10-word telegram became comparable to an exceptionally good airplane excursion fare from New York to London today. Thus prices were still very high, and careful consideration of the need for each telegram was called for. Services that expensive could only be afforded for social purposes by the very rich. Otherwise they were restricted to commercial purposes, where the savings justified the expenditures. Over shorter distances, prices were lower, but still high compared to either mail or earnings, and so usage was also low, and restricted to urgent needs. As Table 3.3 shows, at the end of the 19th century there wasn't even one domestic telegram per person per year in the U.S.. (Today there is slightly more than one airplane round trip per person per year.) Telephone service initially was also extremely expensive. For example, a century ago, unlimited local service in New York City cost \$20 per month. That is comparable to \$2,000 per month currently, when one considers prices in relation to average earnings. How many people would be willing to pay \$2,000 per month for Internet connectivity today? That such prices were paid testifies to the value of communication services as perceived by users. The precise economic benefits of communication are a complicated subject beyond the scope of this work. However, they have consistently been high, often demonstrably so. (At a high level, this claim is proved by the statistics in the tables throughout this work. Since nobody was overtly forced to pay for any of these services, yet usage and spending went up, customers must have been getting their money's worth.) As one example, the first transatlantic cable to carry traffic was completed in 1858. It worked only sporadically, and broke down completely after a few weeks. It never carried commercial traffic. However, when the news that the Indian Mutiny was being brought under control arrived in London, the British government was able to use it to send a telegram to stop the departure of two regiments of Canadian troops for India. That single message was claimed to have saved \$250,000. Similar, although usually smaller, savings were often attained in commercial settings. As a simple example (others will occur throughout this work), the first telegraph connection to Australia was opened in 1872. ``In Sydney the softgood firm of Prince, Ogg and Co., doing business of between [\$2,500,000] and [\$3,000,000] a year, reduced the value of stock held on the premises from [\$1,300,000] to [\$500,000] once they could telegraph to London for supplies'' [Inglis]. The main reason telecommunications services are being liberalized and privatized around the world is to spur innovation and greater investments. Many careful studies have been conducted that documented the contribution of communications to speeding up economic growth. These benefits are easily appreciated by individuals. We read stories of farmers in Bangladesh renting time on cell phones to get the most recent market prices. A century ago, the highest penetration of phone service in the U.S. was in the Midwestern farming states. Even when there were no direct monetary savings, the savings in time from not making unneeded trips or arriving right on time were immense, and easy to understand for anyone. Once mail and phone service prices became low enough, social uses exploded (a topic mentioned already in Section 2, and treated in greater detail in Section 4). This enabled those two industries to increase their shares of the economy. The telegraph never moved beyond the high-priced and infrequent commercial use stage, and eventually died. % In its heydey, though, % xxx debunk many myths: death of distance to be done in Section 21, % here deal with homogenization, peace, (ITU first international % organization), diverging accents % New technologies brought with them social, political, and % economic changes. New crimes and new crime fighting methods % developed. Yet the main reason that new technologies did % develop is that they provided value that people were willing % to pay for. The usual process was that initially % This paper avoids technological issues as much as possible. % Development of postal systems does not appear to have % occasioned any substantial negative comments. There were % many complaints about the cost, slowness, and unreliability, % but mail was perceived in general as a good institution. % (There are probably more negative comments today, about % the volume of junk mail.) The large and often immediately noticeable economic effects of communications meant that any changes in availability or pricing of services could have a large effect on the economy. Even regular mail played such a role. Much of the maneuvering in the U.S. Congress in the first half of the 19th century over postal rates had to do with the competitive positions of newspaper publishers in small towns versus those in large ones, and those in the South versus those in the Northeast. (See [John1, Kielbowicz2] for details.) The interaction of mail with economic life developed gradually. This was largely because mail evolved gradually, and was a familiar service, simply performing better and less expensively what everyone could do, namely carry letters at the speed people could travel. Postal services did play a major role in the economy even in the early 19th century (as is visible in tables 3.1 and 12.1, and is discussed in greater detail in Section 12). However, mail was apparently not associated in the public eye with dramatic changes, at least not negative ones. The electric telegraph introduced a novel phenomenon. The Internet is sometimes portrayed as a uniquely potent disruptive technology. Yet it is worth keeping in mind that the electric telegraph had a similar effect. In many ways, the telegraph was a greater innovation than the Internet. It was the first technology that effectively separated communication from transportation. It had a dramatic effect on science and technology, and it facilitated huge economic changes. The role of the telegraph in transforming the economy of the 19th century can be illustrated briefly by considering the history of the Rothschild banking dynasty [Ferguson]. The Rothschilds were definitely not technophobes. They were the world's largest financiers of railroads in the middle of the 19th century, and much of their wealth derived from that source. On the other hand, they were hostile to the telegraph, since it eroded their competitive advantage. % stimulated competition. In the first half of the 19th century, the Rothschilds attained unprecedented dominance of international finance because they alone had a cohesive organization covering all the major financial centers of Europe. The closely knit family could trust each other, and in addition they had a private communication system that provided them with news in advance of rivals. These advantages were much more important to Rothschilds' pulling ahead of their competitors, such as the Barings, than any early coups during the Napoleonic wars. The telegraph threatened their dominance. \begin{quote} Throughout the 1850s, [James de Rothschild] repeatedly complained that ``the telegraph is ruining our business.'' The fact was that the telegraph made it much easier to do what the Rothschilds had managed so ingeniously before, namely to conduct financial business between affiliated houses over long distances. ... ``It appears,'' James complained in April 1851, ``that yesterday a great many German scoundrels sold [French] railway shares in London with the telegraph...Since the telegraph became available, people work much more. Every day at 12 they send a despatch, even for trivial deals, and realise [their profit] before the bourse closes the same day.'' Once, the Rothschilds had been able to steal a march on their rivals with their unrivalled system of couriers and carrier pigeons; but now ``anyone can get the news.'' James could see that there was no alternative but ``to do the same,'' but it still struck him as a ``crying shame that the telegraph has been established.'' ... Such complaints were still being echoed by James's son as late as the 1870s: although the Rothschilds had no option but to make use of the new technology, they always regretted the way it tended to broadcast financial news... \hspace*{+3in}pp.~64-65 in [Ferguson] \end{quote} The effect of the telegraph on the Rothschild banking business was an example of Schumpeterian ``creative destruction.'' New competitors sprung up, eroding the Rothschild's competitive advantage and profit margins, and leading to greater economic efficiency. Most businesses in the mid-19th century had a choice of using the telegraph or dying, and using it effectively often meant large changes in modes of operation. This new technology was closely associated with new industries, especially the railroads, and new styles of management. It did lead to much more centralized control in both government and business affairs. \begin{quote} The establishment of a direct telegraph line between England and India in 1870 was an event of far-reaching importance. The delay in communication was a great advantage to the Government of India in so far as it of necessity left the initiation of policy in urgent matters to its own hands, and enabled it to confront the Secretary of State with accomplished facts. But all this was bound to change when the Secretary of State had to be kept constantly informed of the course of events in India, and was in a position to issue immediate orders. Henceforth the Secretary of State exercised a far more effective control over the administration of India than was the case before, and the Viceroy really tended to be a mere ``agent'' of the Secretary of State. \hspace*{+3in}p.~848 of [MajumdarRD] \end{quote} For the role of the telegraph and the telephone in shaping commercial management operations, see [Chandler]. There is not doubt that the telegraph and the telephone had a great role in transforming society. However, many of the claims of dramatic changes caused by new communication technologies should be treated with caution. There were changes, but how big were they, and to what extent were they offset by other changes? Corporate management was centralized, but there was an explosion in ranks of corporations, so some general measure of autonomy by managers may well have increased. % Furthermore, often new communications technologies % were developed in order to satisfy society's needs Many communication technologies led to predictions that they would lead to a new era of universal brotherhood and peace. As we know, that did not happen with the telegraph and the telephone, and is unlikely to happen with the Internet. New institutions did arise. For example, the International Telegraph Union (which has since transformed itself into the International Telecommunications Union, the ITU), % formed in xxx, was the first permanent international technical organization. (The Postal Union followed later, % in xxx, since the urgency of standardizing mail communication was not as great as that of the much faster telegraph signals.) % telegraph a tail on the large railroad dog, although % sometimes the tail wagged the dog. There have been many predictions that the Internet would doom large corporations. After all, the generally accepted theory, due to Coase [Coase1], says that firms exist to minimize transaction costs. In the absence of such costs, the factory worker, the garbage collector, the secretary, the software developer, and the salesman could just contract for their contributions to any given job in a decentralized market, without the need for a cumbersome bureaucracy. Since the Internet reduces transaction costs, it has often been predicted that it would eliminate the need for large organizations, and reduce the world to cottage enterprises. However, while some large companies are divesting units, and there is a continuing growth of small new companies, we are also experiencing the greatest wave of mergers and acquisitions in history. Coase himself has explained the reason for this mixed picture. \begin{quote} If transaction costs were the only factor, then we would [become an economy of individual entrepreneurs]. But the fact of the matter is the cost of organizing is decreasing too. The question is whether the costs of transacting decreases as fast as the costs of organizing. My guess is that sometimes it does and sometimes it doesn't. \hspace*{+3in}[Coase4] \end{quote} The same phenomenon occurred in the second half of the 19th century, facilitated by the lower transaction and coordination costs offered by the telegraph. There does seem to be a strong tendency today to move away from vertical integration. That mode of operation appears to be replaced by enterprises concentrated in narrower horizontal layers of the economy, but operating on a worldwide scale. Similar tendencies can also be discerned in the 19th century and early 20th centuries, facilitated by the telegraph and the telephone. This might seem absurd, especially in view of Henry Ford's vertically integrated River Rouge plant. However, even that enterprise was not anywhere near as vertically integrated as say the British and Dutch East India companies. Those ran shipping lines, plantations, banking, and sometimes entire countries or provinces. The general trend since then has been towards greater specialization. There are persistent fears that the Internet will homogenize the world's societies, turning them all into slight shades of the American culture. Such fears are not new, since they were also associated with the telephone, and later with radio and television. Yet this has not happened, and linguists are even discovering that regional accents in the U.S. are diverging. The Internet has stimulated a series of sociological studies, some of which (e.g., [KrautLPKM]) claim that it decreases human contact, while others (e.g., [Kanfer]) come to the opposite conclusion. This is just what happened with the telephone (see [deSolaP1, Fischer2]). It is certainly true that proportions of different types of interactions have changed. It appears hard to categorize them easily, though. \begin{quote} For better or worse, I expect these changes to facilitate a continuing transformation away from interaction in solidary communities and workgroups and towards interaction in far-flung, sparsely-knit and specialized social networks. \hspace*{+3in}[Wellman] \end{quote} There has been a measurable growth in physical distances over which we interact, facilitated by improvements in both transportation and communication. However, we do not have a ``death of distance'' in its extreme form, where we could live anywhere we wished and do any kind of job, and are unlikely to reach such a state. Sections 21 and 22 have further discussion on these topics. The general conclusion is that it is hard to make any categorical statements about communications having caused any great social or economic transformation. New technologies have often facilitated changes, but they were seldom the sole factors. Precise predictions have seldom been correct. Society has adopted the tools that were offered to it in a variety of often unexpected ways. The one factor that is undeniable, though, is that apparent from the tables in this section. Historically communication has been extremely highly valued. Individuals and enterprises have been devoting increasingly large fractions of their resources to it, and consuming rapidly rising volumes of it. The growth in volume of information has led to a change in how we react to it. Over the last two centuries we have moved from scarcity to surfeit. There have always been some complaints about too much information. The Bible says that ``[o]f making many books there is no end, and much study wearies the body'' (Ecclesiastes 12:12). John Winthrop's heavy correspondence from Massachusetts with many people in England led Richard Saltonstall to remark in 1632, ``I fear he hath too many letter'' (p.~217 of [Cressy]). One of the complaints that James de Rothschild had about the telegraph in the 1850s was that it generated too much work. ``[The telegraph] meant that even when he went to take the waters for his summer holiday, there was no respite from the business: `One has too much to think about when bathing, which is not good''' (p.~65 in [Ferguson]). However, on the whole such reactions tended to be rare. % xxx Even Winthrop (?) talked wistfully of how ... has many fine books. People used to be information-deprived, and eager for any scraps they could get. The books [John1, Kielbowicz2] have many illustrative examples of the eagerness with which mail was received in 19th century America. A great visual testimony to the hunger for current information is provided by a photograph reproduced on p.~64 in [John3]. It shows the Post Office in Rochester, New York, on a Sunday in 1888, full of businessmen picking up their mail and sharing information. Today, our task is to cope with the flood of information. There are many worries that restrictive copyright laws will limit our access to data. However, in practice what we see is an unprecedente