Preface

This book was inspired by the recognition that risk has new dimensions in electronic commerce and pushed forward by my experiences with the those who are evaluating and assessing these new dimensions using inappropriate traditional business practices as models.

Each month produces a new model for success on the Internet. First Yahoo! was a search engine and now it is a portal. First it was marketing and now it is service. Yet somehow many consultants and businesses offer a single approach to all these opportunities and all the various businesses. I had a most illuminating conversation with senior consultant from a major consulting company. I asked her to enumerate the times when, after reviewing their own work, the consultant's solution was found to be wrong. I asked her to explain the company actions in response to this discovery of error. There were none. Such flawlessness for a single integrated solution in the face of rapid change of businesses and wide variances within the business community was stunning, too stunning to be believable.

Thus this book is based on some unmentionable facts in business: Internet commerce is a change in business; some businesses will do everything "right" yet be destroyed; some businesses will make mistakes yet thrive. There is no single right model for Internet commerce. There is no single right answer that can be bought with certainty by paying the most expensive consultants. For every business there are different choices. Although some businesses have some obvious price advantages (e.g. Big Lots) and advantages in consumer confidence (e.g. Crayola), these businesses are not destined to succeed. When the business is in code ,not in concrete, there are an infinite numbers of forms. There will be an explosion of business models in the near term.

There are of course some excellent consultants, who offer many solutions and endeavor to create not a marginally customized product, but a truly tailored solution. As with all that is excellent, these are rare. The vast majority of customers get a set of viewgraphs and documents prepared for the mythical generic firm. The generic firm is as mythical as the unicorn. The unicorn was a transmutation beginning with the rhino and resulting from the textual drift inherent in of hand-copied scribal documents. As scribal copying embodied human error and inherently resulted in slow transformation of documents , so mass production embodies the understanding of economic actors as generic units. The mythical firm is a creature of technological limits, beginning with common institutional constructs and resulting from the requirements of mass production. The mythical firm results from the need for a standard compatible with the mass-production model of business and business consulting. In the information age everything is malleable, customized, individual. Thus the texts that propose generic solutions are flawed. Texts that educate and enable individuals to make their own tailored choices are needed for the post-print age. Thus I have endeavored to create one of these: a text that explains some basic parameters.

There are some stable factors in electronic commerce, the towering power of the browser bookmark being one, but every business has its own model. The browser bookmark promises that a site will be the first place checked for news or shopping. But if that site is badly designed, offers bad service, or is unreliable it is unlikely to get a second look.

Expecting a single monetary form to emerge form Internet commerce is a reasonable as expecting a single paper currency to come off the printing press. Movable type created fundamental changes in knowledge production: standardization, the ability to compare systems of knowledge, and specialization of intellectual labor. With the printing press, complex markets for intellectual goods developed, contracts proliferated, paper money expanded, and the age of quantification took flight . (Crosby, 1997; Eisenstein, 1979).

At the beginning of the age of print there was no clear single path on which to direct those beginning at vastly different positions. Similarly there is no single path forward through the age of electronic information. Every business person and consumer has a risk profile that is a function of (to name a few variables) market, market position, and risk aversion; because of this diversity this book is meant to be not prescription but rather descriptive. Just as there was no single way to re-organize business is response to the wealth of paper, forms, and currencies made possible by movable type, there is no single way to minimize risk for every party in light of digital information.

The interdisciplinary nature of this book hints at the magnitude of the changes ahead. The current disciplinary structure was built on the print mode of learning and teaching. Just as a Mater's of Information Science degree is a creature of modern change, a future student might get a Bachelor's in Trust. Similarly some modern business structures will be as useful as the medieval guild as the next century approaches and passes. This magnitude of change requires that a detailed book be focused on the near term. Thus this book inherently has a near-term focus, especially with respect to the Internet commerce systems examined. Trust, risk, privacy, security, and reliability are as fundamental to information commerce as Arabic numerals are to paper modes of commerce. Thus trust and risk are the core of this book.

This book is meant to empower individuals to be their own contractors when shopping on the net, constructing an information business, or building a virtual addition onto their current business structure, to encourage shoppers to tread on the Internet instead of in the mall, and tell them how to keep their hands tightly on their virtual purses. The Internet has a power to intimidate that is unfathomable for someone who has seen the vast bulk of digital silliness that was the early days of the Web. This book should remove any residual intimidation. Should it fail to do so, a quick tour of Usenet should eliminate any residual awe for the denizens of the modern Internet. Decades ago, the Internet was inhabited only by researchers, intellectually engaged gentlepersons with shared norms of behavior and common interests.. Now, it's everyone, all the myriad human foolishness, wisdom, joy, and grief flows through the wires every day. There was a widely used acronym on the Internet, IRL, which stood for "in real life". Now the Internet is real life. Sign up or miss it.

It is my contention that Internet commerce will truly come of age in the Christmas season of 1999. Allow a digression into personal experience to explain this entirely qualitative, rather unfounded projection. First I find that I tend to be a moderate early adopter, the first (or third) to try out new technology. Second, I am rare among technical researchers in Internet commerce as I am the one who actually does the family holiday shopping. I attribute this to gender role differences. During the holiday season of 1998 I did my shopping one Saturday morning while my children played downstairs. I had the list, my credit card, and dogpile. (Dogpile is a metasearch engine. That is, it is a search engine that searches other search engines.) As I have been shopping on the Web for nearly four years I was one or two years ahead of the curve. Thus I predict that in the next holiday season working parents and the elderly across the globe will discover this saver of trouble and time, leading to a more relaxed holiday season for everyone (except, of course, the retailers who have not adapted to Internet commerce.) I found in my shopping no price (dis)advantage, as the difference in price tended to be absorbed by shipping costs. Shopping on the Internet gave me a price equivalent to the discount store, with no taxes paid, and home delivery.

This leads to the second, more mundane, inspiration for this book. Three years ago at the First Usenix Workshop on Electronic Commerce, I realized I was perhaps one of three people in the room, by a combination of gender, class, and age, who actually shopped. I was the representative of every parent who has the experience of holiday shopping. I was the single person there who understood at a visceral level the need for shopping without catalogues, phone calls, or expensive personal assistance. I live in the gap between mythical SuperMom and actual working parent. That is, I am an actual working parent who needs life to be friction free to meet the demands of the mythical SuperMom. The time crunch and the need for schedule -friendly remote shopping that is oblivious to interruptions will drive Internet shopping. The aging of the population makes a trip to the mall less an effortless jaunt and more a day's event. The reorganization of the modern family demands, and the technology allows, Internet commerce. Together these forces point to inevitability. This is the ideal moment to thank my family. First the incomparable Shaun McDermott, a truly wonderful man. A patient and supportive man most supportive in that he is a wonderful father. y daughters, Adonica and Amelia, who have made their own contributions to this book by immeasurable contributions to my life. And finally, Wilson, who taught me many lessons I will not forget.

Certainly my early academic mentors deserve acknowledgment. I would never have started the program of study, much less the book, without the support of Michael Feldman. Early on Hudson Welch and James Morris were endlessly intellectually engaging. I am deeply indebted to Granger Morgan for following his own dreams and beginning the department where I had the honor of studying.. Pam Samuelson provided irreplaceable insight into the subtleties of the law, and despite a schedule that is frightening even in retrospect, always found time to provide detailed comments. Mary Shaw has offered valuable time and insights from her technical and personal wisdom. Bennet Yee has given both professional counsel and patient consideration. I wish his office were still across the way, rather than across the continent. Finally my dissertation advisors, without whom this text would not have come to fruition, Marvin Sirbu and Doug Tygar.

To my friends who started virtual and ended up more than actualized: Phaedra Hise, Charlotte Chen, Robin Schoelenthaler, and of course Pip. Laura Painton and Tse-Sung Wu: Thank you. Rosy Chen shared her heart, wisdom and office. Milind Kandlikar provided passionate occasional doses of perspective. Indira Nair for whom mention is necessary but not sufficient. Donna Riley shared her rare gifts of strength and kindness, bestowed with a discerning wit. Ian Simpson provided continued intellectual engagement. Richard Field offered his very relevant expertise and the kindness of his heart in reviewing and commenting on my work. Cathleen McGrath offered engaging debate or empathy, as appropriate, over uncounted cups of tea. Phoebe Sengers reminded me to like myself, and hold my work just dear enough.

Barbara Slater, Andrew Russell, Denise Murrin-Macey, Patricia Steranchak, Janice Trygar, and Victoria Massimino assisted in a many ways, the greatest of which has been in the sharing of their company and friendship.

At Harvard, Jane Fountain, Susan Cooper, Rob Jensen, and Lewis Branscomb have provided moral support and given me the gift of their time. Harvey Brooks was kind enough to be a reader, and gentle in communicating his sharp insights.

Trust in American monetary instruments is not an eternal national constant. American commercial instruments were marked by early failures; the Continental being the obvious example. 1 In Internet commerce people are once again embarking on a long-term trust commitment. Internet moneys are both unlike and like the dollar. It is one thing to build on the trust of generations past on a monetary instrument and, another to be among the first to take the risk that trust implies. The adopters of the Continental were not made whole by the eventual global adoption of today's greenback.

Internet moneys are like the modern green and historical Continental dollar in that all are based on invisible trust bindings. The trust binding value to the dollar depends on the physical difficulties of reproducing the paper monetary instrument and a centuries-old governance system; Internet commerce depends on the difficulties of calculating mathematical functions and decades-old networks. An Internet commerce system may require trust in the merchant's goodwill as well as his technical competence. Another system may require only faith in risk management of major financial institutions.

In this text the trust relationships in electronic commerce are examined and illuminated. The focus is on trust, but it is equally on risk. Trust is the positive view of exposure: whenever there is trust, there is risk. I focus on these two interrelated topics: trust is risk.

The focus here is on trust as well as risk not only to stress the continuity of the evolution of money from gold bars to bytes but also to provide the most broad explanation of Internet commerce. This focus further distinguishes this study from a consultant's, who might consider risks in a specific scenario to the mythical generic firm.

The determination of risk can be found in an examination of who trusts in Internet commerce transactions. Who will pay, in terms of both money and data, if trust is misplaced? When the inevitable early failures occur, who will be at risk? Who is liable? In many commerce systems there is a trusted third party. Who is this trusted third party? Why is it necessary to trust this party? What exactly is this party trusted to do? Answering these questions means understanding risk allocation in electronic commerce. Answering these questions requires understanding security, record-keeping, privacy, and reliability.

There is no single currency or transaction system which is certain to dominate the future Internet. The answers to the previous questions vary across the multitude of protocols proposed for electronic commerce on the Internet. However, an examination of a broad range of these protocols makes clear that in electronic commerce, there is considerable opportunity to lose both money and data. Customers can lose money and privacy. Merchants can lose money, proprietary information, and reputations. There is much to be gained. It is worth the necessary risk, but only that risk which is necessary. It worth extending trust , but narrowly.

In this text I translate from the technical protocol to the financial risk. There are three basic sources of risk: security failures, data misuse, and reliability failures. This book placed to illuminate the space defined by these three axes. I do not attempt to address every possible risk inherent in electronic commerce. Electronic funds transfer can magnify the weaknesses of cash control systems (Fischer, 1988; Mayland, 1993). If a company has problems with cash control mechanisms and misplaced trust, electronic commerce can make it worse. This is obvious, and is not the focus of this text. The purpose of this book and set of system evaluations is to illustrate risk allocation when a customer, merchant, Internet Service Provider, or commerce service vendor misplaces trust in others, not within their own organizations. (Note that I refer to sellers of all goods but Internet commerce systems as merchants; I refer to those who offer commerce systems as vendors.)

Vendors, banks, consumers, and merchants have different interests. Market and legal mechanisms will assure that the needs all are met in the long term. But one takes risks in the short term. Today the legal environment is uncertain. The market requires information to function, and many are functioning without any better sources of information than the vendors themselves. Thus there are systems which place risks on participants that might better be left with the vendor. This text should provide the tools to determine the sources of risks, what risks are of greatest concern in a few specific systems, and how to evaluate other similar systems.

Understanding risks in Internet commerce requires integrating an understanding of money, network technologies, information security, and the potential for data appropriation and misuse. Thus this books begins with definitions and discussions of money, the Internet, security, and privacy.

In this book I consider the Internet as a framework for commerce. Much of the argument for Internet commerce is essentially information on the growth and population of the Internet. The history of the Internet is included, as it is more than academic. There was at one point an alternative vision of the Information Highway -- citizens as consumers of 600 channels with feedback limited to a single button labeled "BUY". Instead the open Internet has prospered. With respect to shopping and selling, the open nature of the Internet creates trust issues. An open Internet with millions of "channels" has far different trust implications than a centralized broadcast model with orders of magnitude fewer choices.

In short this text addresses the terrain of Internet commerce, rather than trying to lay out a specific path or roadmap. Here are identified the avoidable hazards which are likely to be found on the road to Internet commerce. And thus we begin by considering the nature of the Internet.

What is the Internet

This text focuses on protocols suitable for commerce on the Internet. Why the Internet? The complete answer to that question depends on the set of questions here: What is the Internet? Where is the Internet? Who's out there? Why Internet commerce? What distinguishes Internet commerce from telephone and mail order commerce?

The Internet is a set of networks which are connected using protocols which are open, portable, and enabled the entire research community to share information. That the protocols were open means that the there were no secrets about how the software works. That the protocols were portable meant they could function on more than one operating system.

Software under the corporate tradition is protected by patents, secrecy, and licensing prohibitions against reverse engineering. Software under the Internet model is very different, these differences have important implications. Open software progresses faster than proprietary software. This is because the body of developers is larger. The code or protocols is available to all hobbyists, academics, and every person who can who can study the code, improve it, and share the results. The code has an installed base and is available to all start-ups who would add functionality. Thus, those solutions which are most likely to keep up with the rate of change on the Internet are those that are as open as possible. Thus a popular innovation will not leave your site behind.

Notice open does not imply a lack of security; in fact the opposite is true. More closely controlled code requires a greater extension of trust than open code. Because software is open, and any one can examine or modify it, it is often presented as less secure. But no one modifies the code that a particular site using. Rather modifications extend the menu of options for the software. The modification of the code is relevant for upgrades. The ability of anyone to examine the code there is not likely to be widespread disagreement about its functionality or features. Open code is examined for security flaws by a community of impartial but expert observers. Greater transparency means a lessor need for trust, both in the stock market and the software business.

What are its Origins?

What is the Internet? And who's out there? For those for whom the Internet has exploded onto the scene in the nineties it may come a surprise that the Internet has been developing for decades. The Internet began as the ARPANET, a United States government project for connecting scientific research sites.

The tools for networking networks of computers were developed by scientists and researchers for use in their own non-hierarchical heterogeneous computing environments. The techniques developed were designed for distributed support, using an iterative process which included seeking and considering comments from the user community.

Although the ARPANET connected only a couple of hundred computers at that time, it created the core of compatible inter-networked computers that became the Internet. By 1983, all the networks connected to the ARPANET used the same protocols (TCP/IP) for communication.

After the release of Berkeley UNIX 4.2, TCP/IP was included in every UNIX workstation. The UNIX standard created a commercial opportunity for network products. Although the vast majority of these machines were not initially connected to what we now know as the Internet, the ability to inter-network networks became a standard feature for high-end operating systems.

In 1986 ARPANET became NSFNET. Eventually the protocols that ran over networks existing at the same time, e.g. the IBM/VMS-based BITNET, ran over the Internet wires as well. The students, researchers and librarians were all now connected.

The purpose of NSFNET was to connect all the supercomputers. As part of connecting the supercomputers the regional networks were also connected. The T1 lines connecting these machines were the first Internet backbone.

In 1990 the first commercial email provider, MCI Mail, was connected to NSFNET. Also in the nineties the National Science Foundation began to reduce subsidies, and gave the responsibility of the NSF backbone to commercial providers, thus enabling a commercial Internet without the limitations borne of Federal funding. As long as the Internet was funded from tax dollars its primary purpose should reasonably be research and not the enrichment of corporations or domain name speculators. As the Internet became increasingly commercial the support for the Internet from research funds became increasingly inappropriate.

Along with commercial email providers, commercial information providers came onto the Internet. Early adopters of Internet technology for information marketing included Dow Jones and Dialog (Cerf, 1993). Thus began Internet commerce.

By 1990 the growth of the Internet was too profitable for information providers to ignore. However, the market remained primarily technical individuals, as access to information on the Internet required either some understanding of UNIX or proprietary software provided by an Internet service provider (ISP). The figure below, based came from annual Internet Domain Survey, illustrates extensively the user community has expanded. (Internet Domain Survey, 1998.)

Note that the left hand axis represents millions of Internet hosts2. Thus an estimate of forty-five million Internet users is a reasonable lower bound because it assumes that no one shares computers and that the survey located every host.

Figure 1.1

Exponential Growth of the Number of Computers Connected to the Internet

A year before the connection of MCI Mail, a European researcher, Tim Berners-Lee, became concerned with effectively transporting the images, postscript files3, text and data files necessary for collaborative physics throughout Europe. The protocol he developed for collaborative physics is the underlying technology for the World Wide Web. The Web allows consumers to search for information on the Internet with a straightforward easy to use interface developed first by Mosaic (i.e. the browser) . Easy access to information has been the greatest driver of Web growth.

The World Wide Web is a critical element in emerging markets. With the Web any person could access information easily. Mosaic made it as easy as point and click, Lycos made searching as easy as point and search. These tools dramatically lowered the threshold for technical knowledge to connect to the Internet, send, search, and obtain information. Although the Internet began as a specialized US Government project, it is now global. The Internet domain survey has expanded to include ninety countries.

Where is the Internet?

The Internet is on the desktop and in South Africa. The Internet is global and American. Determining the scope and population of the Internet with any certainty is both an art form and an open research question. Attempts to determine profiles have included hosts counts, mailed surveys, phone surveys, and voluntary Web surveys. This section is limited to the number of host. Because actual machines are easier to count than users. The user-to-machine ratio may vary between and among institutions and households.

Another way to investigate who users are and what they are doing is to consider domain names. These are relatively easy to count and their growth is clearly exponential. A domain name is the part of a URL that is to the right of "www" , or the part of an email address that is to the right of the @ sign. A domain name is a mnemonic for humans who would rather not remember emails by Internet Protocols address, such as: Sue_Smith@128.196.93. Domain names are entirely for user interface. Because domain names are the only ubiquitous form of identity information on the Internet, a detailed discussion of domain names is included in Chapter 2.

As described in the following chapter IP runs on IP addresses. When a device needs to communicate with, for example, a web server and knows only the domain name of the server, the device must get the corresponding IP address before communication can begin. Domain names are not limited in number as IP addresses are. The number of IP addresses is limited by the design of the system. The number of domain names is limited by the human ingenuity ( and, from the evidence, human silliness).

Any number of many domain names can point to a single IP address, so that a single IP address can represent many domain names. Although each domain name must point to exactly one IP address. These IP addresses maybe of a single machine, or of a class of IP address that represents an entire network.

A domain name consists at least of two parts: the top level domain name and the second level domain name. Top level domains are .com, .net, .org, .mil, .gov and .edu. The second level domain is immediately to the right of the top level domain, e.g. "harvard" in harvard.edu; "chicken" in chicken.com; "despair" in despair.com and "slashdot" in slashdot.org.

Conflicts occur most often at the second level, where for example an early adopter might own mcdonalds.com by virtue of having this as a last name. Then the fast food chain would find itself pre-empted. One of the major issues in electronic commerce today is ownership of domain names. If a person starts a business and makes the business grow, can another take away their domain name by previous ownership? There is no definitive ruling on this topic. A domain name may be a extension of intellectual property whether the company owning the corresponding second-level phrase (e.g., "mcdonald's") has registered the domain or not.) Domain names may be a raw material, subject to "gold" rushes. Domain names may be important political speech, as in the case of http://www.gwbush.com/. In the case of gwbush.com the domain name could be considered important political speech, the property of the Bush campaign which was stolen by the commentator, or valuable electronic space which was claimed first by an innovative entrepreneur.

There has been something of a domain name rush to get the best, and sometimes the worst, domains names. Some of the domain name selections seem rather strange. For example, if a confused user searching for the meta search engine www.dogpile.com accidentally typed www.dogpatch.com the page loaded will be http://www.nwnexus.com/. This means that the domain names for Northwest Nexus and dogpatch point to the same IP address.

Domain names are assigned. Only the assignment of IP addresses and the Domain Name system are centralized. In all other ways the Internet and the protocol on which it depends are decentralized.

There are three top level international domain names: net, com and org. Addresses in these domains are currently assigned by Network Solutions Inc. (NSI) of Virginia. The domain name system is the only It costs $50 to register a domain name, and $30 of that goes to a fund controlled by the National Science Foundation to support the Internet for the public interest. There are three top level domain names that are US-specific: mil, edu and gov. Assignment of second level domain names in the mil and gov domains are controlled by the Department of Defense.

It is likely that assignments in the edu domain will go to EDUCAUSE (http://www.educause.edu/). EDUCAUSE is a non-profit consortium of higher education institutions that encourages the use of information technology in higher education. It was formed by the merger of EDUCOM and CAUSE.

There are also many top level domain names that are geographically bound, these are called country code top level domain names (ccTLD). Each nation that cares to have its own two-level country code top level domain name may have that domain. Examples of these include "fr" for France and "uk" for the United Kingdom. Domain names are registered by continental or national entities.

Every domain name must correspond to an Internet address. IP addresses must be unique for the Internet to function. IP addresses are assigned but there is no authority which requires that these assignments are honored. In Asia domain names are assigned by the Asia-Pacific Network Information Center (www.apnic.net) . In Europe domain name assignment is handled by the "Reseau IP Europeens " (www.ripe.net/). In the United States IP addresses are assigned by the American Registry for Internet Numbers (www.arin.net).

At this point it seems possible that the .us domain name will be supervised by the U. S. Postal Service. The us domain is still being handled by the original research-support institution, ISI. The us domain name appears to be used primarily by K -12 schools, which do not qualify for edu domain names , and municipalities. One reason its use by municipalities is popular is that many of the big city com domain names were bought in the domain name gold rush in the1990's. Some big city names, for example, boston.com, were bought by location-specific businesses before the city registered.

The graph below shows that the distribution of the purposes of the users on the Internet has changed over time.

Figure 1.2

The graph above presents the percentage of domain names registered in the in the different top level domains since January 1995 through January 1999 as reflected first by public registration levels and then by the Domain Name Survey.

The org domain is non-profits, for example, sierraclub.org. When looking at this figure it is important to keep the previous figure of absolute group in mind. For example, clearly the number of universities has not declined. Yet the percentage of domains on the Internet that are universities has decreased. Similarly, the number of domain names registered to non-profit organizations have more than doubled in absolute number over the time period. The graph shows that the number of nonprofit organizations has expanded exponentially with the number of overall domains as is necessary to keep the appearance of an approximately constant percentage in the figure.

The mil domain consists of addresses for the US military. The military's share of total domain names has not significantly decreased, staying at roughly 4.5%. Given the rate of growth of the Internet this illustrates that the military has built upon its early commitment with aggressive deployment of Internet technologies in the Internet.

The edu domain is populated by universities. The number of registered educational domain names appears not to have changed dramatically over the period depicted in the figure. It has declined in percentage terms. In the years covered by the graph the number of edu domains more than quadrupled, rising from 1,133,502 in January 1995 to 5,022,815 in January 1998. The phenomenal growth of international, network and commercial domains over the same period accounts for the create the relative percentage decline in the educational domain.

The predominant commercial domain is com. The network domain was originally for network service providers: the IP address registrars above; ISPs; and providers of other network services. Shortly after the inception of the net domain it was discovered that net domains provided more fertile hunting grounds for ideal domain names for late adopters. Particularly for companies which missed the chance to obtain the .com name of choice the net domain provided a second chance. The net domain now serves three markets: traditional businesses moving onto the net, modern net-created business opportunities (e.g. Web hosting) and personal interest domains (e.g. http://www.momspace.net/).

The top area of the figure gov (United States government) and geographical domain names. The increase in regional names, including .us, Asian and European domain names, is reflected by the sharp drop in the total percentages of the other domain names. After July 1995 registration of net domains increases so dramatically that there appears to be a leveling of regional domain name use. Again considering absolute growth registration in international domain names continues at an exponentially increasing rate as shown in the table below.

The growth of hosts on seven continents from the Internet Domain Survey.

Region	Hosts in Jan. 94	Hosts in Jan. 95	Hosts in Jan. 96	Hosts in Jan. 97	Hosts in Jan. 98	Hosts in Jan. 99
North America	1,685,715	3,372,551	7,088,754	11,216,036	20,302,652	33,702,867
Western Europe	550,933	1,039,192	2,699,559	4,352,152	5,537,049	9,300,942
Eastern Europe	19,867	46,125	168,142	238,580	443,191	694,723
Middle East	6,946	13,776	44,484	58,930	103,925	211,824
Africa	10,951	27,130	84,715	104,838	199,958	284,912
Asia	81,355	151,773	672,495	1,006,664	1,661,034	3,089,659
Pacific	113,482	192,390	475,505	647,948	916,538	1,066,398

The customer base on the Internet grows as the number of countries and connections grows: exponentially with time. Although the coefficient of growth varies across the continents, the shape of the growth curve remains the same for each region. It is the expectation of the future of these growth curves as much as the current magnitude that so excite the providers of content and commerce services.

Who is on the Internet?

Profiling the domain is no trivial task; and the profile of the typical Internet user is more difficult. Thus any survey-based discussion of Internet users is subject to gross generalization. With this caveat I now consider just such gross generalizations about users. Keeping in mind the tendency of Americans to deviate wildly from any norm, take this discussion for what it is: an examination of trends by a long time user with an academic bent.

First, the number of female users continues to be smaller than the number of male users. However rate of growth of the number the rate the the rate of growth for the number of female users is growing faster than the rate of growth for male users. Cultural and economic factors appear to drive this gender imbalance. As the percentage and numbers of female users increases, as they will, eventually gender distribution will stabilize at a level that reflects the differences in incomes and free time between men and women. This will have a number of effects aside from the obvious one of increasing the importance of Internet commerce since women still do most of the shopping in America.

Certainly men and women, as well as any gross generalizations can apply to billions, will use the Web differently. Will the threshold at which men and women choose to trust their purchases and data to the Internet be different? At what point will Internet commerce break the threshold of the average shopper? How will the purchasing and marketing decisions of men and women differ and how will theses decisions be similar?

A survey by the Pew Research Center for the People and the Press noted that the Web is multimedia not only in the technical sense but also in the sense that it used differently by different people. This survey classified Web users into four basic types according to what they did while on line: researchers, political expressives, home consumers, and party animals.

Researchers use the Web for professional purposes. They augment the workplace with news and radio, but the primary interest in their Internet is for work-related purposes. (Having been in computer science environments I might point out that an simple count of the pure number of packets to researchers might place them in the party animal department, but this is only an artifact of the bandwidth required by multimedia applications. Video and audio require far more packets than text.) The early markets for Internet commerce are for researchers: books, computer hardware, software, and educational opportunities. Researchers will find that the combination of Internet commerce and the service economy enables them to spend less time at the mundane tasks of life and more time in the lab. If not for issues of privacy, researchers would be the ideal target for integrated Web service sites, which lower the overhead of managing one's life by offering house-keeping, grocery shopping, and delivery services. However integrated services currently sell data about the customers as part of their revenue. Researchers, and others aware of the resale of data, are less likely to embrace such services.

Political expressives go on-line primarily for the political information and opportunities for organizational and discourse that the Internet offers. Political merchandise can be obtained readily on the Web. For example, posters from Nelson Mandela's original presidential campaign can be ordered off the Department of Communications page at the South African government's Web site. The political season is now accompanied by political Web sites that offer information, candidates schedules, and sell goods. Political expressives are committed individuals. The time-saving qualities of Internet commerce are an advantage for this group. Another significant advantage is the ability to evaluate a company or product according to the political information readily available on the Web; searches for products can be easily correlated with evaluations of company performance. That labor practices can be reviewed immediately before an athletic shoe purchase could prove a beneficial to New Balance shoe merchandisers who are well-known for their fair labor practices. Political expressives need mundane goods too.

Home consumers are the obvious point of interest for Internet commerce. The Web's advantages for home consumers are time and convenience. Shopping can be done while the children are napping, or even absorbing their television ration downstairs.

Party animals are a major target of entertainment sites on the Web. It is a reasonable supposition that suck.com is not aimed at the research audience. Similarly reams of sites exist for even the most obscure show, cult movie, and bad habit.

The Internet offers different advantages to different groups of users. For young people, for example, it provides endless playmates and a variety of games and chats. It allows them to discuss potentially frightening topics, such as sex, religion, politics, and drugs, in the safety of their homes. The anonymity of the Internet allows young people to explore new personalities, cultural subcultures, and new roles. Users can change their gender if they wish to. Teen-agers can safely hurtle obscenities, and debate adults with heady feelings of anonymity and equality. All these things draw high schoolers and young collegiates on-line. Once there, they can shop at all hours; without the need for permission or transportation to go to the mall. It is simple to implement a site aimed at teens so that shopping and chatting can co-exist. (A parallel disadvantage is , of course, that they can also impersonate their parents and order merchandise that is later innocently disavowed by cardholding adults. )

Users change their profiles over time. Early users will play, because young people play and there is no reason young people should become suddenly serious when presented with a keyboard. Some of these young people will go to college, and if previous trends prevail they will become both more intense party animals and political expressives. Perhaps they will even manage to graduate from college in the process. This will lead to a move into the researchers group at first jobs or graduate schools. When people have families the extra time will gained from use of Web-enabled service (one hopes) be spent playing with their families. These people will then move to the home consumers group in terms of Internet commerce, although they may generate the bandwidth demands of researchers at work.

How does the Internet work

Why Internet commerce? Why play now when the hazards are undetermined and systems untested on a large scale? Certainly the obvious answer is, "That's where the customers are" as illustrated in the previous discussion of Internet growth. But Internet commerce offers the potential to greatly reduce transactional overhead and remove the constraints of geography and time.

Understanding how the Internet supports varying information markets requires understanding the layers of the Internet. Discussing Internet transactions requires understanding of different network applications (news, Web browser and clients, chat) as well as layers of protocols underneath these applications.

Changes in markets, especially information markets, depend on the nature of the Internet. When publishers and advertisers pay to provide information, they are paying for attention span, increasingly referred to as mind share. In the information economy attention span is going to become an increasingly valuable commodity. University Professor and Nobel Prize winner Herb Simon stated that the most valuable products in the coming years will be those that decrease information flow: filtering, rating, organization and evaluation products. At the time of that statement the Web was still an obscure engineering feat, but its power in organization of information has since become apparent.

The Internet Protocol

A protocol is a communications standard. A protocol defines a series of messages and the syntax for evaluating those messages. The beginning of any datastream identifies the protocol used for formatting the data for transmission. With humans , for example, the greeting will denote the tone of the following conversation. With Internet connections the protocol will define the nature of the connection: streaming high bandwidth content, store and forward text, chat, etc. The receiving machine identifies the protocol and therefore knows how to parse the rest of the data. People use standard sets of exchanges and closures for conversations that are not all that different in function than protocols. When a greeting in a human conversation is businesslike, or friendly, or aggressive the participant who receives each of those types of greetings has information on what is to follow and how to proceed. When a network protocol is described each message has a purpose and a form.

Consider if a human greeting was defined as if it were part of a formal protocol. Included might be standard forms for identification, mood evaluation, and topic introduction. For example, the mood evaluation query, a.k.a. friendly greeting, might be defined as:

query-> How are

e.g. How are you today, Professor Lia?

Protocols may look complex but are only abstractions of simple, and at the best, graceful, underlying standards.

Image greeting the Queen of England with a non sequitur, such as, "FISH!" She would not know how to respond. Essentially this is what happens when network protocols are not interoperable. When systems lack interoperability, the connection is there, but neither machine can make sense of what the other is saying. The machines are, in a sense, speaking different languages.

The fundamental technology of the Internet is the Internet protocol . The Internet is the network of networks that are connected using IP. There is only one Internet as distinguished from intranets or internets of which there are many.

IP is a connectionless protocol. That is, in IP the routes by which each part of a message will travel to reach its destination is not predetermined and the resources for message delivery are not reserved. In contrast, telephone networks have traditionally been connection oriented. Connection oriented protocols establish a point-to-point connection, from one phone to another, when communication is requested. This gives the connection-oriented protocol the ability to ensure quality of service before the connection is established.

Sending information using only IP is not unlike sending a postcard. A postcard is excellent for a discrete message. Of course it is no way private --despite legal controls it could be read by many. IP provides only the addressing, and a best effort in delivering the data.

The figure below illustrates the analogy between packets and postcards. On the Internet data is broken down into packets. The packets are dated and each sent independently. The packets are not encrypted; there is no virtual envelope protecting the contents. Each packet is addressed. Other protocols, including cryptographic protocols and the transmission control protocol, add other features as illustrated by analogy in this figure.

Uses of Protocols

IP addresses have two parts: a netid and a hostid. The network identifier is like the state and city in a postal address: it identifies an area. The host identifier is like your house number: it identifies a specific destination in a general region. IP addresses look like this:

An Internet Protocol Address (IP)

Considering at the sheer length of an IP address it would seem virtually impossible that there would be a shortage. After all, the address are 32 bits long, suggesting that there are 232 possible Internet addresses. In fact, the addresses are separated into different classes of networks: A, B and C. The higher the letter designation the shorter the network ID. This means that different size networks could be easily connected to the same network.

Network ClassNumber of Networks
A	127	16,777,214	0	0/126
B	16,383	65,534	10	128/191
C	2,097,152	254	110	192/223

Classes of Internet Protocols Networks

The table above shows how many networks of each class it is possible to have. Looking at the table and then returning to the postcard example it is clear that Sandia National Laboratories has a class A address and Lawrence Berkeley Laboratory has a class B address. This is because the address is within the appropriate minimum/maximum range shown in the table. The number of networks and the number of hosts is limited by the size of the IP address; it is 2(number of bits in netid). The number of nodes is 2(number of bits in hostid) which is the same as 2(32 - number of bits in netid).

All this means that the number of Internet network addresses is a little over two million, rather than the excess of four billion that the simple length of the address suggests. Since the network addresses were separated into classes so that all machine could have individual addresses the number of networks which can have addresses is limited. So, despite the simple observation of the length of the address, it is possible to have a shortage of IP addresses.

Considering that IP was developed in 1974 (Cerf & Kahn, 1974) when there were sixty-two computers, not networks but computers, on the ARPANET allowing two million networks to easily connect shows foresight and the grace of fine design. The next version of the Internet Protocol, Ipv6 will have mechanisms to address the shortage of IP addresses.

An IP address provides a guess on the size of a network. Of course such a guess does not always prove correct. A network may be connected to the Internet at only one point, so that it really needs only one address, and host routing can be handled behind this point of connection. This is very common with commercial sites that have firewalls 4and extensive local area networks. Or a connected institution may use dynamic IP addressing5 so that they need fewer addresses than machines. So a large internal network does not require a correspondingly generous IP address.

Coincidentally with the expansion of the ARPANET there was BITNET. BITNET was an early network that consisted of dial-up terminals and IBM mainframes. (The UNIX-based NSFNET grew to embrace and obliterate this concurrent network.) By 1985 BITNET had its first exponential growth in mainframes, to seven "conference machines". Mainframes ran distribution list and forerunners of Usenix groups and could be considered ancestors to today's servers. In addition there were hundreds of machines which could connect to BITNET including machines at Yale, University of Maine, State University of New York -Stony Brook, Brown University, Harvard, MIT, and Tufts University. BITNET allowed users from hundreds of machines to run terminal emulators to come together and chat synchronously. Daniel Oberst offered the followed evaluation of the network in the BITNET monthly newsletter: "BITNET is still by and large a voluntary, cooperative network that only exists to the extent that people work together..."

That this comment came from BITNET illustrates an important point: networking is connectivity, is sharing, is trust. The following rough description of Internet routing shows why this previous evaluation remains arguably applicable. Routing illustrates is an excellent example of trust. Routing is how a machine, given that a machine has an IP address, connects to others. Routing is, specifically, how a packet (small information chunks, like postcard greetings) gets from machine A to machine B. Routers are special -purpose machines which direct packets. Routing is also a function of a general purpose desktop machine. Here the primary focus is on the mechanisms, not the machines. However, envisioning only a router as performing the function of might make this explanation easier to follow.

Routers keep a list of all the machines or hosts to which the router is directly connected, and a list of machine to which all the physically adjacent routers are directly connected. A network to which neither the router itself nor none of the physically adjacent routers are directly connected is a remote network. For remote networks, the router keeps a continually updated list of the first step of what it believes to be the shortest path to that remote network. Routers do not store complete paths to remote destinations, they store only enough information to send the message to the next router, under the assumption that the next router will direct the message properly, and so on until the message reaches its destination. The trip between two routers is called a 'hop' regardless of the physical distance between two routers the distance between them is still one hop. Thus routers trust the message to other routers.

Routers are always updating their beliefs about the network. At any time a router may receive a broadcast from another router about that routers stored information. The receiving router always trusts the received information about routes and updates itself appropriately. Because of the constant updating it is quite possible that in a message consisting of several packets will travel on different paths and possibly arrive out of order.

Machine	Connected To these Networks	Connected To these Machines
122.46.77.32 (me)	122.46, 113.22	122.46.77.31, 122.46.77.35
128.22.36.81	128.25, 126.14, 115.22	128.25.233, 126.14.122, 115.22.004
128.14.46.98	114.7	128.14.56.33, 128.14.77.22
113.22.88.45	default

Router Database

Notice the addresses on this table are imaginary numbers, the point is to illustrate the type of knowledge a router would have.

Consider the network of routers on the Internet as analogous to a social network. Imagine a world with no central information for people in various regions, in other words no telephone directories. Imagine searching for a person, say Gene Eric Person in San Francisco, in the way a router searches. First you would go to your address book. Consider a region a subnetwork and a person a machine to make this analogy function. You would know all the regions to which you can send directly --Chicago, Dallas, Charlotte. You who also know the regions to which your direct contacts can send to. Here is how a page in your address information would work if it was analogous to a router. In this case you would send your message to Gene Eric Person in region San Francisco to Cathy. She would look at her address book and send it to someone in San Francisco, and that person would send it to Gene Eric.

NameLocation	Can Connect To
Carlos	Chicago, Milwaukee	Pittsburgh (1) Philadelphia (2)
Cathy	Dallas, Austin, SF	San Francisco (3) San Juan (8), Austin (1)
Catlin	Charlotte, Atlanta	Jacksonville (2), Miami (3)
Carter	Oklahoma City	almost anywhere

Your Address Book

If you did not have him in your address book you would send the message to Carter, knowing Carter is well-connected and is likely to be able to get a message to any location. Carter is your default router.

Imagine later you get an updated address book from your friend Carlos. Carlos notes that it takes him one friend, one hop, to get to San Francisco. You notice that it took only three hops to get to San Francisco through Cathy. You would update your address book as so:

NameLocation	Can Connect To
CarlosChicago, Milwaukee, SF	Pittsburgh (1) Philadelphia (2) San Francisco (1)
CathyDallas, Austin	San Juan (8), Austin (1)
CatlinCharlotte, Atlanta	Jacksonville (2), Miami (3)
CarterOklahoma City	almost anywhere

Your Address Book

You may find also that Carlos is two hops separate from Carolyn who lives in Portland. You would not add Carolyn to your address book because you are not trying to build a complete and global database of every location. You just want to update your information on the best way to get to any location from your location. But the next time you looked for Gene you would send your postcard to Carlos instead of Cathy.

The address book is your local routing. If the person you wanted to contact was not in your address book you would call someone who would be likely to know him or her because of their location. That is, like the router, you would make your best guess from your most recent information about your network of peers as to the shortest path to the person you want to reach.

There are three critical observations from the above discussion of routing: there is no optimal physical location, there is no single point of failure, and routing is a cooperative exercise in trust.

First, there is no optimal physical street corner on which to reside. Customers will come from all locations, and the appearance of a web presentation depends upon the path between a browser and the information. It is not possible to be adjacent to every browser - there is no ubiquitous next door location. Yet there is valuable real estate in the Internet: not on the network but on the customer's desktop. A good place to be on the Internet is in the Web surfer's bookmarks. The ideal place to be is in the bookmarks that the user trusts.

An ancillary implication of this lack of optimal physical location is that there is limited possibility for monopoly control of distribution6. In traditional media markets there is limited competition. Consider newspaper, television, and radio markets. The ownership concentration results from expensive or exclusive distribution channels. Most towns have one newspaper because the start-up costs are too expensive in a market with an existing newspaper. The newspaper chicken-and-egg problem is that one must have a subscription base and a distribution network to have a paper. Only one radio or television station can exist at one wavelength. On the Internet there is no single advertising venue. There is no reason that multiple competitive search engines as well as multiple meta-search engines can not continue to thrive and compete. Because there is no center to the Internet, because of the routing, there is no way to ensure that every person entering a market sees one product first.

Second, in a connectionless network each packet is delivered independently, so that there is no single point of failure. So if packets are not getting through on one route the following packets will try a different more likely route. Packets are routed independently of each other, and therefore are not stuck repeating previous mistakes. Because there is no central directory of addresses, there is no single point of failure. This means connectionless networks are survivable, that is, hard to disrupt. One business implication of this is that such networks are reliable.

Third, routing is an exercise in social cooperation. Social networks break down from lack of cooperation. Routing could similarly break down. The widespread routing failures that I know about (there may be classified information but a routing failure tend to be noticeable) have thus far resulted only from errors in router configuration, not from malevolence.

Transmission Control

Layers of Protocol & Stages of a Transaction

The hypertext transfer protocol (HTTP) is a protocol that provides seamless delivery of different types of data, and since the Mosaic project, through a user-friendly graphical interface. HTTP is the data formatting protocol of the World Wide Web. It allows users to easily publish and obtain information on the Internet. Browser used with HTTP provide a simple user interface which highlights other files using color or graphics. Browsers using HTTP catalog locally available applications for file display, and automatically provides the text, sound, or graphic using these local applications.

ProtocolConnects	By Providing
Internet commerce Protocols	Consumer to Merchant	payment, possible delivery verification
Hypertext Transport Protocol	Application to Applicationlocation and presentation
Transmission Control Protocol	Machine to Machine	reliable delivery of multiple packets
Internet Protocol	Network to Network	delivery of packets between networks

Hierarchy of Protocols on the Internet

With the development of the Web, the Internet became fully capable of supporting user-friendly distributed commerce, just as previous protocols had enabled functionality from simple communication to file transmission. Table 1.6 above illustrates how Internet commerce protocols have built on previous protocols, which has in turn expanded the pool of possible merchants and consumers. Without the ability to locate goods, consumers would not shop on the Internet. Without the ability to easily present goods, merchants would have difficulty selling their wares on the Internet even if they could be located. Of course, Internet commerce does not depend entirely on HTTP as some protocols include options for users with only email with no HTTP capacity.

The Internet supports a range of business functions, not simply payment. Every transaction, on or off the Internet, has multiple phases: discovery, price negotiation, final selection, payment, delivery, and dispute resolution. The Internet can support many types and all stages of Internet commerce (Sirbu and Tygar, 1995). Understanding how necessitates understanding why.

HTTP works on a simple client/server request and response mechanism. The Web is indeed the "killer app." Not only is it the killer app in business terms, it could have "killed" the Internet by having many short, and therefore potentially ill-behaved, connections ill-suited for TCP/IP. The investment to ensure that the Internet will succeed and thrive has been made despite the sub-optimal design of HTTP offers promise that the informal governing structure of the Internet can handle future problems that may arise.

Internet commerce has increasingly become possible with the advent of the World Wide Web. The Web is growing at many times the rate of overall Internet host growth. The Web allows the consumer to locate information of interest on the Internet without requiring any technical expertise.

All Internet commerce protocols can be used with the Web. In addition, some commerce protocols (Mastercard, 1995; VISA, 1995) are comprehensive and include the ability to transfer funds using only email. (For a detailed discussion of network protocols see Schwartz, 1987 and National Center for Supercomputing Applications, 1995).

Commercial Transactions

Evaluating Information On-line

Product discovery is the greatest current commercial use of the Internet. Yet the lack of validation of services and the uncertainty of the quality of information are serious issues in discovery and shopping. How trustworthy is information provided during discovery?

When a business is presented on the Web there is no tangible information about that business. Slander, fraud, and misinformation are not confined to the Web; but relative anonymity and the lack of a need for physical presence makes misinformation easier. Those who lie can hide behind Web sites with noble faces, including words like "Justice", or "Consumers for Freedom" in their names. Because of the importance of open discourse, American judicial traditions of respecting speech, and the patchwork of jurisdictions over the Web the reader, not the state, must take the responsibility for detecting what is false, and not accepting information on the Web at face value. From the other side is the responsibility of the creator of a Web page to show why that page should be believed.

Misinformation is the hazard in any medium of communication. In this section I discuss ways to evaluate a site to determine whether it is reliable. There is no certain test to determine from afar the quality of a product or trustworthiness of site. However, some factors can signal falsehood or identify misrepresentations.

A competent computer science student who has been insulted, been ill-served, lost money or had his or her affections dismissed can easily put up a web site illustrating how the previously beloved, the ex's business or the ex's employer is an enemy of good -- and with some small skill can make that site look quite believable. On a larger scale an activist, during the 1996 Presidential race (I believe a Forbes supporter) made bogus Clinton, Dole and Buchanon attacks through sites at www.clinton96.org, www.dole96.org, and www.buchanon96.com, sometimes clearly mocking, sometimes subtle and vicious The casual Web browser was likely shocked by the policies and quotes found at these sites; understandably as these fabrications were beautifully and professionally presented. In the 2000 Presidential election the Bush Campaign has requested that the FCC force removal of the www.GWBush.com site on the basis that it has the same look as the official Bush Campaign site and may confuse voters. The URL for searching on domain name ownership is http://www.networksolutions.com/cgi-bin/whois/whois. This should provide contact information for the domain name owner.

Companies and organizations have so far been able to respond to angry disavowals such as www.netscapesucks.com (which includes the Sucks 500). Just as the Web has been the mechanism for angry students in the past, it will increasingly become the mechanism of angry customers and employees in the future. In learning how to deal with this companies can take a note from those who have been subject to harassment on the Internet long before the explosion in the com domain.

The first thing to consider when evaluating information is the source. Can you determine the source? In my hobby space, Mom's (http://www.momspace.net/) I clearly identify myself as the creator. In the Presidential race examples above; however, only a search of registered domain names would identified the source of the bogus political pages. Is it really a nonprofit organization fighting for right, or a talented undergraduate at a technical school? Who is the source? Look for the ability to contact a physical person, not through email but with a street address or phone number that connects to an actual human. This information provides jurisdiction information; that is, there is a forum in which to sue should things go wrong.

When evaluating information look at the domain name. Does the it end in edu, com or org? This may identify an irate undergraduate pretending to be an organization. Technical schools are particularly hazardous (cmu.edu, mit.edu, etc.). A real world non-profit organization, such as the Sierra Club, will have a domain name ending in ".org". A real company domain name will end in a ".com", or less likely a ".net" (or even less likely a geographic name). The absence of a top level domain may be a very good indicator of a bogus organization - but its presence is meaningless.

When evaluating an evaluation, consider the tone. Especially look for loaded link names. If a link says "Evil Smith Hobbies, is owned by the vile Bob Smith: he who stomps worms, hates flowers" and connects to Bob's page, look for evidence in Bob's own words. That the link naming him as a worm-stomping flower-hater goes to his page is no evidence that he does in fact loath flora & fauna. Is the "evidence" of Bob's practices fake email from Bob? It is very easy to write text and pretend it is an email, or to edit an actual email. Which Bob Smith is it? Search the Web for Bob's own words. If an organization advocates a truly hateful idea or truly loathsome policy it is probably documented somewhere, if not openly advocated. If a contribution to a hate organization results from a purchase there will likely be mention of it elsewhere, certainly as a tribute or a shopping suggestion on the hate organization's page.

Always look for links to independent sources. Who links to this page? Is this just a page with many headers that it will get selected by search engines often? Or do verifiable organizations link to this page? If such an organization does link, does that constitute a general endorsement of the information contained therein (e.g. "more information here"), or a specific statement of a single example of cooperation (e.g. "Smith Hobbies Haters & The Society to Loathe Bob joined us in this lawsuit for free roses.")? None of the Clinton, Dole or Buchanon pages linked to other organizations. They were self-referential. They should have linked to political parties and, more important, the political parties should have been linked to them -- and would have if they were real! Who links to pages that evaluate businesses? Virtual Better Business Bureau stickers should connect to the Better Business Bureau, and references to Consumer Reports positive evaluations should link to the appropriate story in Consumer Reports. The ability of watchdog groups to use bozo filters7 makes the link far more important than the image (which is trivial to copy).

Of course consider the content. Information too ridiculous to be believed should not be believed. If the things claimed are too bizarre to be true, they may well be false. The alternative: make claims reasonable and believable. Being outrageous to lure customers is not as effective on the Web. A huge $20.00 Levi's Jeans sign and no $20.00 will probably not work as well on the Web as in real life. There is no price to leaving the virtual store. There are no large plastic "Going out of Business" banners on the Internet, because these would decrease trust and not lure customers.

Beyond evaluating businesses and organizations, this checklist can serve to determine if an irate consumer or angry gadfly is presenting reliable information. Consumers can effectively provide information and companies can respond. There are social and technical mechanisms with which to respond top harassment or mere editorials.

First, the mythical Smith Hobbies can copy the Meta Data of the complaint page so that searches that result in the complaint page will result also in Smith Hobbies response page. Meta Data is the information about the Web page to help spiders, search engines, and bots to classify pages. To view the Meta Data for any page just select 'view source' from the browser menu, for example a page on parenting might have the following Meta Data:

Of course, this also works for those who are unhappy about a business. If a business treats a consumer badly that consumer can take action to ensure that all searches that find the company find the irate consumers page as well.

When a browser hits a page it is a trivial matter to determine the referring page. If the referring page is a complaint page, use simple commands and direct that browser to a response page. This is a way to respond to complaints without pointing them out to individuals who would not be aware of them.

Pricing and Quality in Internet Commerce

As discussed in the last section, quality of information is a critical element on the Internet. This is an issue both for customers evaluating merchants and merchants evaluating customers. Customers want to be certain.

Companies that use the Internet are often attempting to capture consumer surplus: the difference between the amount a consumer would pay and the price actually paid. Companies can come closer on the Internet than off to charging every customer the most that customer would conceivably pay for any item the customer purchases. Anyone who has ever found a bargain has experienced the joy of consumer surplus.

A current leader in real-time price discrimination is books.com, which offers customers the ability to easily compare prices. If Barnes & Noble or Amazon offers a lower price on a particular item and the buyer chooses to use books.com automatic price comparison feature then books.com automatically matches the price. On average books.com has a slightly higher price than Amazon or Barnes & Noble. The consumer who does not bother to compare will sometimes pay the higher price. (Often; however, the prices are the same.) The consumer who shops at books.com and always compares pays the lowest price if the other servers are immediately available. Of course sometimes the Barnes & Noble or Amazon sites are not available. In this case books.com charges its usual price, regardless of whether it has searched for the object before and has some knowledge of a lower price. Books.com offers price discrimination between those who compare prices and those who do not.

There are social as well as business implications to pricing in Internet. First, discrimination in markets is not inherently bad. For example, in clothing, upscale stores discriminate against bargain buyers by pricing them out of their stores. Bargain buyers go to TJMaxx instead. This better suits both buyers who will pay top dollar for selection, timeliness, and atmosphere and buyers who want lower prices. Similarly high feature or brand name Web sites can charge more, as Amazon's continued success in the face of books.com strategy illustrates.

Second, pricing discrimination on the Internet cannot be based on the demographics on which socially destructive price discrimination is based. For example, in traditional markets with price discrimination, women pay more for cars. This argues for women to shop on-line for cars. Mortgage approval rates vary based on ethnicity. Yet Web pages offer the same mortgage rates to all. If the perceptions expressed by Fukuyama (in Trust, 1995) are widely shared, the variance in mortgage prices is a function of trust. That is, the lender and seller have less trust in certain demographic groups. This is expressed as higher rates, higher frequency of credit refusal, and higher prices.

The trust of the customer in the merchant is as much an issue on on-line markets as off-line market. On-line; however, trust is likely to be based on browsing habits and credit lines with no gender or ethnicity being an issue.

In Internet commerce customer trust is the critical variable. The more a customer trusts a site the higher the price the site can charge for what it sells. This does not suggest that price discrimination is a matter of customer betrayal; rather price discrimination is a function of merchant reliability. Customer trust is belief that the merchant will fulfill the terms of the transaction (e.g. deliver quality goods in a timely fashion). Any customer on or off the Internet has a price/reliability sensitivity. Thus the lower price of quality second-hand goods. Price discrimination may mean offering a lower price to obtain a sale rather than offering a targeted higher price, based on preferences exhibited by the customer at the site at the time or purchase.

Mistaken attempts to capture consumer surplus can lead to lost sales, sometimes very sour business relationships that can last a lifetime. Consider a real life example. I went shopping intent on buying a new car but ended up not doing so. I very much wanted a yellow Geo Metro convertible. (This should not reflect upon other recommendations in this book.) I investigated the price. I was willing to pay slightly above dealer cost and sign a long term maintenance contract. The offer I was making was fair. When I found no takers for my offer, I bought a used VW Beetle. This is an example of imperfect price discrimination based on mistaken trust. The salesmen whom I met had greater trust in the veracity of their gender-based evaluation of me than in my ability to set a price. I had the option of sending in a male friend to make the deal, in fact a possibility I investigated. Yet I decided the business relationship was too sour. This has negative social implications, of course, but also has negative business implications for the merchant. I would not seek the friction involved in attempting to buy from a GM dealer again, based on my almost arguably unrepresentative experience. On the Web it is even easier to leave the lot. On the Web there will be some merchant to take a fair offered price.

Internet users can respond to price discrimination by using the power of the Web to search; thus the market in mortgage rates and offers for mortgages in which there is a perception of discrimination. In any such market the Internet will have a distinct advantage in terms of customer trust over traditional marketing mechanisms. There is similarly a significant market for information on automobiles.

The user who would rather have a listing of selected books rather than the option of a second price comparison will go to Amazon. The most price-sensitive user willing to do price searches on every purchase will go to books.com. Consumers will respond to price discrimination by changing how they use the Internet. Pricing will become increasingly dynamic. With Internet purchases, as with automobiles, there is much competition. If a customer experiences ill-suited price discrimination on the Internet, it is likely that the customer will never return to the site where the error in pricing was encountered. Thus sellers must make price discrimination decisions carefully on a case-by-case and product-by-product basis.

Convergence and the Internet

Internet commerce is a subset of telephone and mail order commerce. In a few years Internet commerce will be distance commerce because of the technologies of convergence. (Here mail order and telephone order are referred to as "distance" commerce to distinguish them from emerging models of electronic commerce.) Packet telephony, advanced television, and cable modems are all artifacts of digital convergence.

What is convergence? Previously, technology has provided for policy makers three distinct platforms for speech: print, air, and wire. This resulted in the creation of four media types: publisher, distributor, common carrier, and broadcaster. These types began to converge with wireless telephony, multi-media services, and television delivered through cables. Now all tradition media types exist on the Internet: the Wall Street Journal is a publisher; Amazon is a distributor; ReMix Radio is a broadcaster; and AT&T MediaOne is a common carrier.

All media types will ultimately converge onto a single network of networks using IP switching. All these media play different roles in distance commerce. Here I compare and contrast the uses of traditional media with the Internet.

Broadcasting is especially useful for advertising and information distribution (e.g. discovery.) Obviously its one-source-to-many-recipients model makes it unsuitable for purchasing; rather broadcasting is used to encourage a purchase. Discovery is supported in multiple modes on the Internet, as previously described. Now everyone has NTSC 8televisions and will slowly adopt digital high definition television. With high definition television the television image will be as good as the image on a computer screen so WebTV will be truly useful and possible.

With "Internet broadcasting" companies need to be cognizant of the recipient's capabilities in a manner that is not necessary with broadcasting. Disney provides an excellent example of a failure to understand the distinctions between traditional broadcasting/advertising and the Internet. On its Web site Disney offers much paid content and that seems reasonable, given its market power. However, to view the free part of the Disney site requires fast hardware, a very high speed connection, and multiple helper functions. The Disney site contains every form of content: video, audio, animation, etc. This probably looks wonderful at Disney Studios. To hazard a guess, I will say based on no information but my experience as a Web surfer that Disney used its regular internal team to develop its state-of-the-art Web page. But exclusive use of the best high-end graphics which are excellent for television are an error on the Web because of the resources they require. Disney's site is so state-of-the-art that it is time consuming if not impossible to use over a 56.6 modem. Children do not like waiting for downloads and will likely not install multiple helpers. There is little or no easy-to-download content; for example, pages to print and color. The use of video and animation is truly excessive. Clicking around the Disney page is an experience in frustration for users that do not have the latest equipment and at least cable-modem speed connections. Finally, Disney does not have the dominance on the Web it has on televisions so frustrated users can easily visit the sites of competitors. Thomas the Tank is as easy to locate on the Web as is Disney. In broadcasting, and in movies, Disney can dominated distribution. That is not the case on the Web. There are multiple Thomas sites with free coloring book pages, and simple interfaces that are suitable for a wide range of connections and machine, but nonetheless very entertaining. Disney does not understand that the distribution dominance it has on television does not map perfectly onto the Web. Thus Disney has used a flawed publishing model for its Web site.

Distributors are the category of media most perfectly replaced by the Internet. In media terms, distributors refers to "distributors" of information. The Internet is likely to greatly reduce the need not only for bookstands and newsstands, but for all types of distributors. Desktop computing reduced the number of middle managers, previously needed to watch the books and handle the paperwork. Sales forces will be the most notably reduced population by Internet commerce. As the Disney example illustrates, dominating the distribution channel has been to this point critically important in selling information goods and obtaining mind share. The music and movie industry structures are built upon the assumption of expensive distribution channels that tend to be controlled by a few major players (i.e. natural oligopolies). This will continue to change.

Publishers of material that is not broadcast are the second category to undergo fundamental change due to the Internet. In the case of newsprint, such publishers have monopolies in most cities. The Internet promises democratic in that there is no natural monopoly in distribution. Limited competition in newspaper, television, and radio markets results from expensive or exclusive distribution channels. Most towns have only one newspaper because start-up costs are too expensive for potential competitors. One must have a subscription base and a distribution network before one has a newspaper -- clearly a bootstrapping problem. Only one radio or television station can exist at one wavelength.

In contrast, on the Internet anyone can be a publisher. The network is the distribution channel and it cannot be monopolized. Monopolizing content control would require software at all user endpoints; for example, built into the operating system. IP provides only transport -- only distribution. IP does not distinguish between sources and destinations. Any user can be both.

Traditional publishers, of course, find this unnerving. However, traditional publishers will also dominate on the Internet if they fulfill consumer's trust criteria. Users will go first to established sources of information because they have some preexisting trust in these locations. This fact will serve the interests of established institutions who provide product and advertising information along with the opportunity to purchase. Outrageous claims and ill-considered priorities will; however, decreases this trust and evade their advantage.

The rise of the on-line magazine Salon (http://www.salonmagazine.com/) in the midst of the Starr investigation illustrates the role of trust in the established media outlets. The traditional print and broadcast media had consistently chosen to honor the privacy of Rep. Henry Hyde by not disclosing what they knew of his sexual conduct while simultaneously publishing the details of President Clifton's personal life. Salon broke with the pack by reporting the sad tale of a family destroyed by Rep. Hyde's sexual infidelity. Such reporting increased consumer trust in Salon and decreased trust in the traditional media. Certainly major media players have much trust remaining among consumers but the vaulted Watergate press corps of the seventies has already become the Lewinsky press corps of the nineties, with an corresponding decrease in trust. This leads to the question: how much of market control is based on trust and how much is based on established distribution patterns? Only long-term Internet use will answer this question.

Common carriers transmit any material, regardless of the message. Telephone companies and the U.S. Postal Service are common carriers. Clearly the Internet can provide the services of common carriage. Currently the mails are used for both transactions and discovery. On the Internet mail can be more tightly targeted, by for example, asking people to sign up at a Web site for a mailing list. Because of the low cost of sending email, some companies send spam. Spam is as likely to result in impassioned recipient's refusals to be customers as it is to result in customers. Internet commerce enables tightly targeted requested email.

Digital convergence usually includes broadcast televisions, radio, telephony, and cable transmissions. But more than traditional communications signals run across the wires -- payment also goes through the Internet. As video, voice, debates and newspapers converge payment for these will converge also. Internet commerce illustrates that money itself is also converging onto a digital, Internet-transmitted form.

Functions of Money

What is money? As defined by its three elemental functions, money is a store of value, a standard of value, and a medium of exchange. Ensuring that electronic commerce maintains money's functions as store and standard of value is not a trivial matter but certainly manageable. In contrast, ensuring that electronic commerce maintains money's function as a medium of exchange is difficult. The Internet's power lies in its lack of need for physical presence. This creates a difficulty for electronic commerce and however, in that because there is no physical presence there are also no handing of papers, no tactile examination of goods, and no certainty of receipt.

Money as a store of value requires durable storage. For money to be a store of value, it must not be easily destroyed or created. If money decays or is destroyed in storage, then it obviously does not succeed in storing the value it represents. In contrast, hyperinflation illustrates the failure of money as a store of value when it can be too easily created. Under hyperinflation, entire nations are forced to abandon money and return to barter. Durable storage is a critical factor, but one that is not difficult to achieve, in electronic commerce. Unlike physical money, electronic money is merely bits, and thus can be trivially duplicated. Note that this duplication of money differs is the same as the creation of only when the duplicates can be spent. Ease of duplication eases durable storage but also can simplify fraud. Thus ease of duplication is a double-edged sword. Durable money storage is necessary for electronic money to fulfill the functions of paper money, but durable storage is simplified in electronic commerce.

Money as a standard of value provides a simple triangulation for all transactions. Consider the sheer number of transactions that may be required to obtain a desired good in a barter economy. An entire series of trades may be required so that some final barter could be arranged. That is if one person had some good, say flour, wanted software there would be a series of transactions until the person who had flour had the goods desire by the person who has the software. Essentially this is how the Internet ran for many years. People wanting software contributed to the common good by working on the software they used and identifying bugs and offering additional features. This is the freeware culture: a social network of a closed community. Clearly that is not going to work for widespread consumer commerce because consumer markets are open and the consumers follow a wide variety of value systems who would value coding contributions in a substantially different way.

Exchanges with no Standard

Consider the series above. A has code and wants services. B can provide services but has no need for code. A has to go through a series of trades to get what B wants, in this case a new modem. D has a modem and wants a the latest Java shirt. So A has to go to C to get the Java shirt and then process a series of trades to finally be able to obtain the desired services. When A and B can both trade one common item which is wanted by both, rather than a series of exchanges of goods and B can just trade money. The comparison of the lines in the boxes and the simple exchange below illustrates why this is called triangulation. Electronic money must provide for this triangulation by setting standard of value. This is currently being done by pegging electronic moneys to a particular currency or set of currencies, so that electronic mechanisms can be used to trade dollar equivalents over the network.

Triangulation with a Standard of Exchange

Money as a standard of value also requires interoperability;9 that is, to serve as a standard of value, any specific form of money must either be itself widely used (a standard), or readily convertible to another form that is widely used.

Money as a medium of exchange requires special transactional properties. As a medium of exchange, money must have transactional durability; that is, money must be conserved in transactions, not created or destroyed. Monetary transactions must be consistent: the amount received by the seller must be the same amount paid by the buyer, with no change in that amount occurring during the transaction. The transactional properties that enable money to serve as a medium of exchange amount to transactional reliability. Reliable transactions in electronic commerce are necessary to the proper functioning of electronic money as a medium of exchange.

What are the properties of a reliable electronic commerce protocol? The study of distributed databases has defined the characteristics of reliable database transactions as atomicity, consistency, isolation and durability. I will address these in detail in Chapter 9; for the moment consider how physical transfers of money illustrate the properties of a reliable transaction.

Please note that during this, and all future analyses, I take advantage of gender-specific language to simplify my discussion. The customer is assumed female; the merchant male; and the bank neuter. This allows me to use she, he and it without worrying that the reader may confuse the noun referenced by the pronoun.

Consider a customer's physically handing a dollar bill directly to a merchant and how it exemplifies each of the properties referred to above. Atomicity means that the transaction fails or succeeds completely. Consistency means that both parties know the outcome of the transaction. Isolation means two payments do not interfere. Durability means that the transaction cannot be undone without the consent of both parties.

Atomicity : The dollar bill will not be lost as it leaves the customer's hand and is transferred to the merchant. There is always exactly one dollar; it is never duplicated or destroyed. If the dollar is dropped, then the customer can pick it up and return the transaction to its previous state.10

Consistency : After the transaction the merchant knows he has one dollar more; the customer knows she has one dollar less. At no point in the transaction is there ever any confusion over who has the dollar.

Isolation : That dollar bill will not be confused with a previous dollar bill, so the merchant cannot falsely claim failure to have received payment, and the customer cannot escape her obligation to make payment.

Durability : After any party receives the dollar bill, he or she retains the dollar bill until he or she transfers it in another transaction.

None of these simple physical safeguards to reliability necessarily holds in an electronic transaction. Like purchases with a dollar bill, some Internet commerce transactions are anonymous. When a merchant receives a anonymous payment using an anonymous system, it is as if the customer has thrown a dollar bill across a dark room. Who should the merchant credit with this payment? How can this payment be linked with a specific purchase if there is no customer standing in front of the merchant? Who should receive the goods? In this case, the electronic dollar cannot be identified with a specific purchase or purchaser. So the trivial issues in a face-to-face anonymous purchase are significant problems in a networked anonymous purchase. Overcoming these problems depend on a cryptographic public key to verify identity for a promise of payment (as described in detail Chapters 3 and 4). A public key is a mathematical way of proving identity and signing messages. If a public key is used to sign for a payments, who is at risk if the key was not valid: the verifier, the merchant, or the consumer? Right now the answer to that question depends on the physical location of the transaction. However, questions of jurisdiction are far from simple on the Internet. Knowing that the rule of the law varies between Florida and Utah does not determine which law is binding. Furthermore, neither the Florida nor the Utah statute has been tested.

In electronic commerce, the payment message must travel over an open network (that is, a network without security) from the customer to the merchant. Without verifiable acknowledgment as part of the protocol, the customer has no way of knowing whether the merchant received the payment message sent by the customer. Under the Internet's standard transmission control protocol, a payment message may be duplicated if the communications protocol believes the packet containing the payment message has been lost on the network. This happens, for example, when there is congestion and messages get dropped at the congested router. Moreover, network failure may destroy a payment message. If a payment message is lost, delayed, or destroyed, confusion may result. If forced or faked network outages can create confusion profitable to someone then such outages are sure to be created. (In the system analyses the last chapters include examples of how creating or falsifying network failures may enable fraud and theft. This varies with every system.)

In sum, ensuring transactional reliability is not a trivial matter in electronic commerce. Thus, the provision of reliable transactions is a critical issue in the analysis of risk in electronic commerce protocols. Difficult technical matters involving reliability assurance may obscure business decisions and risk allocations.

Digital Information Money

In its form money reflects the economy in which it is found. Cowry shells are the money of an inland society that trades with the coast. Tobacco is the money of an agricultural nation, in which producing something that is not food illustrates wealth. Internet or electronic money is the money of an information economy. Those left holding only paper moneys, those who do not make the evolutionary leap, in business practices and consummation of transactions, will be passed by.

Historically early money in most economies commodity money. A commodity is something that can be consumed. Commodity money fulfills the role of money as a standard of value but is less useful as a medium of exchange. After all, who could carry around an estate's worth of grain or tobacco? Commodity money is also a poor store of value. Very few commodities can be stored, especially agricultural commodities that will rot or be consumed by rodents. Commodity money is also subject to inflation on the basis of quality of goods. For example, after tobacco became the standard of exchange in colonial times, the streets of Virginia were flooded with quantities of inferior leaf. This dismayed the holders of money as much as the smokers. Leaf or grain may conceal impurities, such as stones, which have relatively great weight and no value.

A return to commodity money usually results from hyper-inflation or economic collapse. A well-documented case of returning to commodity money is the use of cigarettes as money after World War II in Europe. Two decades later public phone tokens were used for money following hyper-inflation in Israel. In this case the commodity was arguably a service - a phone call.

Metal money typically displaces simple commodity money in most economies. Metal money is the interim money, between money which can be consumed and money which has no intrinsic value. It can be argued that metal money is a commodity money because metal can be turned into instruments of warfare (e.g. bronze) or used in decoration (e.g. gold). Metal money can fill all three monetary functions previously identified: it can serve as a store of value, a medium of exchange, and a standard of value. It is more easily transported in large amounts than commodity money. Metal money can be transported over large distances --another improvement over commodity money. In particular, metal money can be transported on ships and through inclement weather in circumstances where grains and other consumables may rot. Metal money was particularly useful for international commerce in the nineteenth century.

Like paper money, metal money is subject to inflation, although certainly with stronger constraints. To wage war and finance empires, rulers through history have lowered the purity or weight of the coins of the realm. Yet not even the most creative ruler using only metal currency could not produce the hyper-inflation possible with paper moneys.

The recognition that there is no actual need to hold the metal money itself in order to possess the value it represents has given rise in the modern economy to convertible paper money. Convertible paper money began with merchants and banks, who would write out notes of credit for customers declaring that they had adequate deposits to enter into a particular debt. The money specified in the note could then be converted into metal at the trustee institution. Yet the holder of the money still had the certain feel of paper in hand. These paper guarantees of deposit were the first Western trust money. The holder of the money trusted the buyer not to abscond with the gold represented by the paper money, and similarly trusted the guarantor of the deposits to hold sufficient gold in the depositor's account to cover the notes when presented.

The concept of symbolic money was a necessary (and not uniquely Western) invention to go onto the next step - intangible money. Some vendors currently offer forms of electronic money that can be converted into tangible money, or greenbacks. Others offer money that can be changed into notational credits, on credit cards, for example, where changing to greenbacks can have a high overhead.

Fiat money is paper money with is no guarantee that it can be converted to any other form. The U. S. dollar has been fiat money since 1971, when President Nixon took America off the gold standard. Fiat money is trust money on a larger scale. Some vendors offer electronic fiat which works in a closed environment where it can be exchanged only for goods sold for that currency or for credits within the system of the vendor which issues it. For example, proposals for sharing computer code based on ratings of individual contributions are fiat money. In that case one could use code based upon what was previously distributed but on a token system instead of a reputation-based system. As I describe later on in detail (see Chapter 11) First Virtual credits are fiat money for a matter of weeks, but there is a guarantee that if both parties (customer and merchant) act in good faith the money will eventually be altered to notational money in the form of credits on a depository account.

Arguably one consistent trait in all these forms of money is their difficulty to produce: the labor associated with creation of the money is appropriate to the economy it serves. Thus when agricultural production was the standard, agricultural goods were standard money. As wealth increased with trade, the standard of harder-to-produce metal came to replace consumable goods. After the Industrial Revolution, when steel could be rolled out and sliced like cookie dough, the creation of detailed paper money was required. Now, anyone can produce photo-quality paper. Paper is too easy to produce, ill-suited for remote commerce, and risky to carry. Thus for moving money in an information economy only bags of bits will do.

Social scientists would argue that all money except immediately consumable commodities are networks of trust, not just fiat money (Coleman, 1990). This argument would suggest that Internet commerce is nothing new. I believe that Internet commerce is something new; never before has there been the need to establish a trusted currency for so many on the basis of such intangible connections. Further the implications for supply of money; for a private creator of global money; and global commerce cannot be foreseen. Yet the historical examples of shipping, private banking, and long distance commerce offer some guidance. Essentially history offers the lesson of caution. Even the most trusted entities may fail. Early Internet monies may hold their value no better than Dutch tulips. Putting the technical ability to prevent risk in the same hands as the contractual ability to distribute risk calls for careful consideration and oversight. That is, those that can best prevent risk should be the ones to bear it. And as the regulation of credit cards tells us, limiting the ability to exploit the customer may be a precondition for an explosion of the next generation of commercial instrument. So, with that in mind, now consider the risks of selecting a vendor for an electronic commerce protocol.

Money Vendors

Today the creation of money is seen as a national right - an inherent function of the nation state. Yet this was not always the case and there is no reason it should continue to be in the long term. As the phrase "not worth a Continental" reminds us, bankers and state governments were more broadly trusted with the ability to uphold commitments to convert money than the U. S. government in the early days of this republic. The U.S. government was able to successfully declare its national monopoly on coinage only after a century of repeated instances of financial speculation became intolerable to the public at large.

Who will offer electronic money? It appears that the first parties to offer successful electronic money have been multi-national financial services corporations. These companies have several advantages. One of the greatest is that they have already established trust, or at least customer relations, in many nations. Second, they have the diversity of resources to protect their entry into enter the potentially risky market of electronic commerce: other financial services they provide will remain profitable while the budding Internet commerce market unfolds. They have scale in number of users; that is, there are already many consumers who use their services and the corporations have the ability to manage all these accounts. And finally, they can offer easy interoperability between national currencies through their current international market relationships.

Another potential player the setting of standards for electronic money is Microsoft. There is an efficiency argument for integrating the wallet into the operating system; the same argument for integrating the browser, ftp, and many other functions into the operating system. By integrating the standard for electronic commerce into the operating system, Microsoft can set the default standard for Internet shoppers. Integrating the wallet into the operating system has many potential advantages. The wallet can be seamlessly integrated into the browser, enabling every user to browse and buy. This will allow Microsoft to control the risks as well as set the terms. Of course, if any single company holds the coin of the information realm, consumers may find post-hoc negotiations about the distribution of risk inadequate for self-protection. In any case merchants and consumers must take care in selecting their options and reading their software licensing agreements.

Computer chip manufacturers may also set standards for Internet commerce, or at least influence those standards that function optimally. Why is integrating the wallet into the hardware such an obvious next step beyond the browser? First, Moore's Law says that chip density doubles every eighteen months. But what will be the use of all those transistors now that every decent desktop machine end can handle multimedia? Security remains computationally intensive and slows even the best desktop machines. Security is therefore the next obvious application for the denser more powerful chips predicted by Moore's Law. A fast encryption chip could determine which systems works fastest, and is ther