PROPERTY AND INNOVATION IN THE GLOBAL INFORMATION INFRASTRUCTURE
1996 U. CHI. L. F. 261
.

Henry H. Perritt, Jr. [FNd1]

 At least since Blackstone and John Locke, legal and political commentators have recognized the central role that property plays in market economies and democracies. Property preserves personal autonomy and provides an incentive to produce. [FN1] At the same time, public uses of resources also are necessary to well-functioning modern societies. Intellectual-property law creates property interests in commodities that derive their value from the information contained therein, rather than from tangible raw materials. But when information is subject to property interests, a variety of competing interests enter into the balance between private ownership and public use.

  New information technologies disrupt historic balances between these competing interests. Just as print technology gave rise to copyright concepts, just as phonograph recording and film technology required adaptation of copyright concepts developed for printers, [FN2] just as the Xerox machine necessitated reassessment of mid-twentieth-century copyright concepts, [FN3] so also does the Internet require rethinking the role and form of intellectual property in stimulating a healthy information infrastructure, in which adequate incentives exist to create and deliver quality information.

  Threats to established balances do not, however, suggest that balance is no longer necessary. Owners of intellectual property should not be granted enlarged economic protection merely because they would like it. Users and exploiters of existing intellectual property should not be free of traditional intellectual-property restrictions just because they would like to be. Rather, intelligent appraisal of the role of property in an Internet-shaped information infrastructure should focus on the specific free riding and piracy risks created by the new technology. For example, traditional electronic-information services like CompuServe and Lexis do not present the same problems as the Internet's World Wide Web ("Web"). Traditional electronic-information services involve closed systems, within which it is relatively easy to arrange licenses between creators of content and electronic publishers. Such systems can enforce copyright by excluding potential or actual infringers. Neither traditional licensing nor infringer exclusion, however, is easy in the open-network environment of the Web. Moreover, the publisher/intermediaries in the Web usually only point to content; they do not reproduce, distribute, perform, or display the content themselves. [FN4]

  Unless one carefully probes the implications of specific technologies and network architectures, one may be drawn into a fundamental and unnecessary alteration of the historic balance between property ownership and socially productive uses of information resources. The White Paper issued by the Clinton Administration in 1995 [FN5] falls into this trap. This Article suggests how such a trap can be avoided.

  One must do more than merely appreciate the technology. One must also understand the role that the law plays in conjunction with other forms of protection for intellectual property. Intellectual-property production flourishes in many environments where legal protection is uncertain or entirely absent, such as the early days of desktop computer software development and radio and television broadcasting. Once one understands the specific threats posed by specific technologies, and the complex of protections available from law and other means, one may suggest legal changes, if any are needed.

  This Article begins by exploring the nature of intellectual property in order to understand the difficult balance between competing private and public interests. It then explains how the Internet and other open network architectures bring new challenges to the positions of both content originators and intermediaries. Next, the Article considers a number of solutions for protecting the legitimate interests of intellectual-property owners while avoiding inappropriate threats and costs for entities that facilitate legitimate use and exploitation of information in an open information infrastructure.

  This Article concludes that a combination of existing intellectual property and contract protections can sufficiently prevent gross piracy, especially when reinforced by anticipated changes in pricing and new product concepts and packaging. In addition, if Congress makes statutory changes, some new statutory protection for intermediaries may be appropriate. The Article emphasizes that copyright collectives can play a useful role in reducing transaction costs for the deployment of new types of protection based on technological limitations on unauthorized uses, technological detection of copyright and license violations, and electronic payment systems. In addition, collectives can make it feasible to express and enforce specific-use licenses.

  The Article does not explore the important question of patent protection for processes in the Global Information Infrastructure ("GII"). It only mentions in passing the role that trademark and unfair-competition law can play in protecting the good will and marketing investment of those that produce information value. The Article assumes that basic governmental information, such as statutes, judicial opinions, agency rules and orders, and land records, is ineligible for intellectual-property protection. [FN6]

  Professor Trotter Hardy's article in this volume applies the Calabresi and Melamed [FN7] transaction-cost framework to property in cyberspace.  [FN8] Professor Hardy concludes that lower transaction costs resulting from cyberspace technology justify a shift toward property instead of liability concepts for protecting originators of information value. [FN9] This Article, however, concerns a different facet of transaction costs in markets for information: the transaction costs that justify legal intervention in the first place.

I.  Intellectual Property as a Limited Property Right

  Property enjoys a special place in the ideology and law of market economies and democratic political systems. Indeed, the concept of property rights dates back to the earliest Western thinkers. For example, Aristotle expressly linked man's normative duty to pursue the "happy life" with man's ability to possess enough property to "make . . . life desirable and lacking in nothing." [FN10] John Locke argued that ownership derives from nature itself, so that one must work on property in order to divorce it from raw nature. [FN11] Because man's relationship with nature, not with other men, imbues ownership, no one person or group of persons possesses the power to interfere with that right of ownership. [FN12]

A. Moral Rights in Property

  Professor Margaret Radin's theory of property and personhood argues that a fundamental purpose of property is to give property owners a sense of identity, a feature so critical to an individual that rights in property must be protected by some legal framework. [FN13] The paradigm of this connection between property and personhood is an heirloom or a house. [FN14] Yet despite this intrinsic value that property presents to individuals, the possibility that some property can be easily exchanged for an identical good negates the inference that all property confers a moral or natural right.  [FN15] To accommodate the needs of individuals to maintain their sense of personhood through strong property rights and the need of society to allow the free flow of fungible goods, Professor Radin suggests that property should be thought of as a continuum. [FN16] On this continuum, property for personhood would receive the most protective rights, and correspondingly, fungible property would receive lesser or no rights. [FN17] Where a particular piece of property rests on this continuum depends on social consensus. [FN18]

  Professor Radin's theory accommodates Europe's and America's traditional legal approaches to intellectual property. [FN19] Europeans have long recognized the existence of a moral right to intellectual property which resembles Radin's property for personhood. [FN20]

  The American tradition is utilitarian, [FN21] concentrating on the market for information and seeking to maximize the amount of content available to society. Under this concept, the fear is that, if authors have no property rights in their works, the incentive to create such works will be destroyed and the overall amount of information available to society will necessarily diminish. But, if authors have such strong property rights that almost no unconsented-to use is ever legal, then the information will be so inaccessible that it might as well not exist. The proper equilibrium is achieved, not by examining the author's sense of identity, but by designing legal rules to create the largest, most accessible pool of information.  [FN22] This utilitarian standard avoids the social consensus and personal/fungible ambiguities of Radin's theory by providing a quantifiable method for balancing the competing public and private property interests.

B.  Coase Theorem

  Utilitarian analysis argues that intellectual property may be entirely unnecessary. Thus, it relieves the law of the need to strike any balance. The Coase theorem argues that, in the absence of transaction costs, it does not matter where law places entitlements. Parties will bargain and allocate legal rights and duties according to their economic preferences. [FN23] If there were no transaction costs, intellectual property would not matter. While the originator of valuable information would not have a copyright, he nevertheless would produce the information because someone who wanted to consume the information product would pay him to create it.

  Transaction costs exist, however, in the real world of information-product creation, distribution, and consumption. [FN24] One type of transaction cost arises from the public-goods character of information. Public goods are those demonstrating the characteristics of nonrivalness or nonexhaustiveness, and nonexcludability. [FN25] One person's consumption of a public good neither interferes with nor excludes another person's consumption of the same good. [FN26] Hence, a public good is nonrival. It is also impossible (for a pure public good) to exclude any one person from benefiting from that good. [FN27] The classic example of a public good is a lighthouse. [FN28] The use by one ship of its benefit, its light, does not reduce the amount available for other ships. Nor can any ships be excluded from the benefits of the lighthouse if any ship has access to the light. Any ship in the lighthouse's vicinity may use its light to keep from running aground.  [FN29]

  A private good is both rival and exclusive. [FN30] Consumption of the good excludes others from consuming the same good, [FN31] and relative to a public good, it is much easier to exclude consumers from the good's benefit.  [FN32] Most of the goods and services bought and sold are private goods.  [FN33] Gasoline, for example, is a private good because each gallon can be used by one consumer only to the exclusion of another consumer.

  The nonexcludability and nonrival features of public goods threaten the ability of an original supplier of goods to recover her investment. [FN34] Nonexcludability means that she cannot systematically refuse to supply the good to nonpayers while supplying it to payers. Nonrivalness means that each customer becomes a potential competing supplier. [FN35] Thus, public goods, and goods that have some public-good characteristics, have a higher free-ride potential. [FN36] Conversely, the stronger the characterization of a product as a private good, with its commensurate increase in exclusivity, the lower the free-ride potential. A drive-in movie, for example, has public-good characteristics. It is nonrival and it is difficult to exclude viewers. Nonpaying viewers may not be able to hear the movie, but they can still see it and free ride to an extent. The same movie takes on private-good characteristics when shown in a private theater. It then becomes possible to exclude people from enjoying the movie altogether, ending the free-rider threat. Thus, as the movie takes on more private-good characteristics, its free-ride potential decreases.

  A dichotomous private/public good distinction may not be helpful in determining free-ride potential since few goods are pure public goods.  [FN37] An information product is not a pure public good, but rather a public good with private-good characteristics. The consumption of information does not necessarily reduce the amount available for another user's consumption and is therefore nonrival--a characteristic of a public good. Reading the information contained in this Article, for example, does not reduce the amount available for the next reader. On the other hand, a consumer can be excluded from using an information product until he pays--a characteristic of a private good. The same reader may be excluded from reading this Article until he, or a library, has purchased the book. To determine its free-ride potential, a product's public/private good features must be evaluated carefully.

  Most information products are not perfect public goods because excludability is possible at some cost, depending on the technology. Indeed, even the light from the lighthouse--the classic example of a public good--may be handled as a private good in some circumstances. [FN38] When exclusion is feasible, it often represents a transaction cost both to producer and consumer. The originator must erect barriers, such as fences--electronic or otherwise--to keep nonpayers out, and consumers must deal with these barriers in order to strike a bargain. An example familiar to lawyers is the interposition of login routines before one can access Lexis/Nexis. Lexis/Nexis had to pay the cost of establishing and maintaining this login routine as a kind of fence around its information products. Consumers must pay the cost of remembering their login names and passwords and logging in each time they wish to consume some information.

  The nonrival feature also represents a transaction cost, but only to the producer. The first consumer, having paid only for his consumption, can transfer a copy of the information to another consumer--or would be seller-- without losing his own ability to consume. This means every consumer becomes an alternative source of supply at only the cost of copying. This transaction cost presents a challenge for the initial negotiations between originator and initial consumer to impose conditions on the consumer's transfer of a copy of the information.

  Together, the transaction costs arising from nonexcludability and nonrivalness have justified the establishment of property rights, represented by traditional copyright, patent, and trademark rights. [FN39]

C.  Information and Tomatoes

  As the preceding section noted, the features of information technology and networks complicate the utilitarian calculus. They are Coasian transaction costs. For example, suppose someone devotes much effort to discovering all of the servers on the Internet that contain information about China, believing that such information would be useful to businesses desiring to develop Chinese markets. He carefully collects the universal resource locators ("URLs") for the servers and the particular locations (files and directories) on the servers where the Chinese information may be found. He assembles the URLs into a logically organized set of Web pages so that anyone with access to his Web server can easily find and obtain copies of specific information about China.

  The entrepreneur occupies a different economic position from someone who has, for example, grown a crop of tomatoes. Like the farmer, the Web entrepreneur can deny access to his product until he is paid. Unlike the farmer, however, the Web entrepreneur faces competition from his customers. The farmer's customer, having bought one tomato, has only one tomato. While she may resell that tomato, her resale does not deprive the farmer of anything. The farmer has already been paid. This characteristic of the tomato is "rival," signifying that two persons may fight over the tomato, but that they cannot both have exclusive possession of it.

  Unfortunately for the Web entrepreneur, his products may be nonrival,   [FN40] depending on how he packages and sells them. If he transfers the entire collection of Web pointers to each customer, the customer easily can make copies of the set of pointers and sell the copies. The nature of digital technology makes the cost of copying very low. Thus, the customer can keep what she bought and also sell one or more copies. [FN41] The economic reward gives her an incentive to engage in the copying. The customer is better off if she sells the copies for any amount of money greater than her cost of copying.

  The nonrival character of digital information has two implications for our Web entrepreneur, both of them bad. First, the possibility of copying means that he is deprived of the sales made by his customer. Second, because his customer did not have to incur the costs of discovering the information and assembling the pointers, she can price them much more cheaply and still cover her costs--essentially, the mere cost of copying. Either the Web entrepreneur will not be able to sell anything because he maintains a higher price, sufficient to cover his higher costs; or, in order to compete effectively, he must price below his total cost, thus eventually driving himself out of business. This teaches the entrepreneurs a lesson--do not invest your energy in creating products like this because you will not be rewarded.

  The foregoing parable explains the justification for the reproduction right in copyright law. In some respects this reproduction right seems similar to the traditional property interest that the farmer has in his tomato. Just as taking the farmer's tomato without his permission constitutes the tort of conversion, so also reproducing the Web entrepreneur's set of Web pointers without his permission may be copyright infringement.

  There are, however, two difficulties with the analogy. First, unless the reproduction right is carefully limited, it may represent a significant barrier to the very activity it seeks to encourage. [FN42] The consumer of the information product may need to reproduce the product to use it. Some reproduction of information products may be necessary to build other information products.

  Second, a special reproduction right may not be necessary. A conventional property interest, sufficient to protect the farmer against the theft of his tomato, also may be sufficient to protect the Web entrepreneur against the theft of the fruits of his labors. The exclusivity characteristic concerns means employed by a producer to exclude a consumer. A theater owner, for example, excludes consumers by showing his movie only within an enclosed theater and only admitting paying patrons. The movie is nonrival, but excludable. A recent study for the Australian government identified several features of publishing markets that could allow returns for a first publisher to remain high enough to cover the costs of producing the original, thus obviating the justification for intellectual property under the public-goods rationale. [FN43] These features include lead time, costly copying, less than perfect copies, inability to use copies without support and assistance from the originator, and collusion between potential copiers and originators.  [FN44]

  Consider the efficacy of conventional property interests. The farmer benefits from these interests by fencing his tomato patch and asserting control over his stall in the marketplace. Anyone who takes tomatoes from the patch or the stall and sells them without the farmer's permission commits the torts of trespass to land, [FN45] trespass to chattel, [FN46] or conversion,  [FN47] and the crimes of theft, [FN48] burglary, [FN49] or defiant trespass. [FN50] The rationale for these torts and crimes is the prevention of conduct inconsistent with the farmer's property interests in the tomatoes. Trespass, conversion, larceny, and burglary are applicable to information resources. If one fences off the information wares in one's computer, unauthorized entry may be remediable by property-oriented torts, such as trespass or conversion, and property-oriented crimes, such as burglary or larceny. [FN51] Some new form of legal protection or property interest is necessary only when conventional property interests are insufficient. Historically, protecting against theft of the tangible object in which intellectual property was embodied, such as a book, protected the author and publisher to some degree, but did not protect them against the prospect that someone would copy the contents of the book and resell it. In the electronic context, it is important to identify the difference between theft of the tangible artifact and free riding on the intangible intellectual property. If technology makes it more difficult to steal the content of the "book" after buying or stealing the book, the gap to be filled by intellectual property, as opposed to conventional property, may be smaller. [FN52]

  Regardless of the respective roles that conventional and intellectual- property concepts play, all forms of property are limited by the needs of the larger society to do things that in some sense diminish the economic expectations of the property owner. For example, owners of real property never have enjoyed the privilege of engaging in conduct that constitutes a nuisance.  [FN53] Early in the twentieth century, the common-law courts established that owners of real property do not enjoy the right to restrict flights over their property at reasonable altitudes. [FN54] More recently, environmental law has imposed further restrictions on property owners' rights and privileges. [FN55] Taking of private property, though entitling the property owner to compensation, similarly represents a limitation on uninhibited exploitation of private property. In all of these traditional examples, the law strikes a balance between the utility of private property and the utility of public use of what otherwise would be within the scope of private property.

  The same balancing occurs in an intellectual-property regime. Absolute ownership of information potentially blocks use of the owned information. Since all information in some sense is derived from other information, ownership of information must be limited in order to avoid preemption of the information base from which public debate and various economic activities can take place.  [FN56]

II.  Economics of Free Riding

A.  Coase's Caution

  Professor Coase appropriately cautioned that policy judgments about his theorem and about transaction costs should be based on real-world facts, not on theory:

  (H)ow is it that these great men have, in their economic writings, been led to make statements about lighthouses which are misleading as to the facts, whose meaning, if thought about in a concrete fashion, is quite unclear, and which, to the extent that they imply a policy conclusion, are very likely wrong? The explanation is that these references by economists to lighthouses are not the result of their having made a study of lighthouses or having read a detailed study by some other economist. Despite the extensive use of the lighthouse example in the literature, no economist, to my knowledge, has ever made a comprehensive study of lighthouse finance and administration. ( [FN57]) The lighthouse is simply plucked out of the air to serve as an illustration. The purpose of the lighthouse example is to provide "corroborative detail, intended to give artistic verisimilitude to an otherwise bald and unconvincing narrative."

  This seems to me to be the wrong approach. I think we should try to develop generalisations which would give us guidance as to how various activitiesshould best be organised and financed. But such generalisations are not likely to be helpful unless they are derived from studies of how such activities are actually carried out within different institutional frameworks. Such studies would enable us to discover which factors are important and which are not in determining the outcome and would lead to generalisations which have a solid base. They are also likely to serve another purpose, by showing us the richness of the social alternatives between which we can choose. [FN58]

  Professor Coase's caution is particularly appropriate when one thinks about new information technologies. The application of the Coase theorem to markets for information is debatable. [FN59] New information technologies may be misunderstood by policymakers and commentators. And, even if policymakers and commentators understand these technologies, the use of simplifying metaphors may obscure important realities. This Article argues that it is useful to begin with the actual transactions that occur in new cyberspace technologies. Those technological realities suggest that there may be important sources of excludability and rivalness for information products in the new environments that undercut any arguments in favor of expanding intellectual-property protection. As a recent report commissioned by the Australian government said, "(t)he . . . onus is on those advocating any expansion of copyright protection to show on a case-by-case basis that such extension would benefit producers and consumers." [FN60]

  Of course, one must be careful not to build policy on transitory technology applications. For the foreseeable future, however, it is safe to assume that computers will be used to process information, that they will be connected in networks, that the networks will use open standards for their connections, and that the capabilities popularized through the Web will continue to be exploited. Thus, the Web model is not likely to become irrelevant any time in the near future.

  Technology in the Web plays two important roles in analyzing the role of intellectual property. First, it reduces the nonexcludability and nonrivalness characteristics that justify intellectual-property protection. In this role, features of the technology that impose zero or only slight transaction costs between originator and initial consumer, while imposing very high transaction costs on potential free riders and pirates, reduce the overall transaction costs justifying intellectual property. Second, because the transaction costs resulting from the Web technology are asymmetric, they may not justify new configurations of property and liability entitlements in the evolution of intellectual property.

  One justification for special property rules for information is the inability of the originator and initial owner of intellectual property to recover the cost of creation because of the potential of free riding. [FN61] This argument suggests that, absent legal protection, the market will not produce enough information. Pirates do not have to repeat the investment of the creator, and thus can sell the creator's product at the marginal cost of copying, forcing the creator to price at the marginal cost of copying in order to retain any market share. [FN62] At this price, the creator will never recover her fixed costs of creation.

  Intellectual property law remedies this problem by increasing the pirate's costs. [FN63] Under a regime of intellectual property, the pirate faces not only the marginal cost of copying, but also legal costs, which represent his expected liability. [FN64] A real pirate, moreover, faces additional costs. He must acquire the pirated information, transform it into a form that he can resell, and market it.

  The costs of originator and pirate can be expressed in equation form. The originator's costs are:

  co=cc cct cm cr

  where cc is the cost of creation, for example, payments to the author; cct is the cost of chunking and tagging, or otherwise preparing the information for publication; cm is the cost of marketing, including promotional expenses, distribution costs, and costs of billing and collection; and cr is the cost of copying or reproduction.

  The pirate's costs can be expressed as:

  cp=ca ct cm cr cll

  where ca is the cost of acquisition, for example, finding and downloading the material a pirate intends to resell; ct is the cost of transformation;  [FN65] cm is the cost of marketing; cr is the cost of copying; and cll is the pirate's cost of legal liability. There is no reason to expect that the pirate's marketing or copying costs will be lower than those of the originator. [FN66]

  A free-ride problem exists only if the pirate's costs are less than the originator's costs, in other words if

  cc cct>ca ct cll.

  Professor Hardy's article, [FN67] the White Paper, [FN68] and other commentaries about property in cyberspace assume that cc cct is much greater than ca ct, and therefore that the value of cll must be increased to maintain the current level of free-riding risk.

  This Article argues that that may not be so. In particular, it argues that products can be designed so as to increase ct to the point where it counterbalances any reduction in cll that results from cyberspace technology. Product design can increase the benefits of the originator's lead time, thus increasing the probability of recovering fixed costs before the pirate can establish a market presence. Finally, the Article argues that payments can be transferred from those who gained from a relatively frictionless cyberspace  [FN69] to creators who otherwise face free-riding risks.

  When fixed investment is large compared with variable costs of production, free riding on the first producer's investment becomes a possibility. The imitator free rides by producing the same product at marginal costs close to those experienced by the innovator, but avoids the fixed costs incurred by the innovator. [FN70] It is not necessary that the free rider have no fixed costs; only that the free rider's fixed costs be less than the first producer,  [FN71] and that his marginal cost not be so much higher than the first producer's [FN72] as to cancel the advantage of lower fixed costs. Accordingly, the free rider faces a lower average total cost and enjoys a price advantage over the innovator. The greater the free ride, the greater the cost advantage.

  But there are countervailing considerations. The monopolies created by intellectual property may be unnecessary where significant economic barriers to entry exist. [FN73]

  First, the potential free rider may not be able to avoid fixed costs to any great degree. [FN74] An example outside the information industry is the imitation of a new aircraft technology by another producer. The detailed engineering design and tooling required by the copier potentially overshadows the original producer's investment. [FN75] Second, there are delays before any copier can realize earnings from free riding. [FN76] Third, the first innovator gains reputational advantages, which assist in differentiating that producer's product from those of imitators. Fourth, the first innovator has a head start in taking advantage of learning-curve cost advantages which can deter entry and result in supranormal profits while the technology matures. Finally, firms already in a market use existing market structures to deter new entrants who lack production facilities, managerial experience, and channels of distribution. [FN77] In one survey, being first with an innovation, moving quickly down the learning curve, and having superior sales or service efforts exceeded patents as means of appropriating the benefits from innovation.  [FN78] Similar barriers to successful entry can operate in the information industry.

  In addition to the differences between fixed and variable costs, other barriers to entry by pirates may exist. Some fixed costs may represent transferable rather than sunk costs and thus be less at risk. Publishers can reduce the incidence of sunk costs by automating certain publishing processes. Lower sunk costs reduce the amount of fixed investment available for free riding.

  Fixed costs may be broken down into two separate components: sunk costs and transferable costs. [FN79] Sunk costs are product specific and the producer cannot recover them through sale or lease. [FN80] If an investment can be used for other purposes, it does not represent a sunk cost. [FN81] Only the sunk portion of that investment is free rideable, because only it represents irreversible investments. [FN82] Transferable costs, however, are not free rideable because the initial producer can sell the investment, recover its cost, and negate any cost advantage held by the free rider. Thus, free rideability depends on whether a significant part of a first producer's costs are sunk. Investment risk to an innovator decreases commensurately as it moves from irreversible to reversible investments [FN83]--from sunk to transferable costs. A first producer will be much more willing to enter into a venture with lower investment risk represented by lower sunk investment. Labor, a major sunk cost, illustrates this point. Technological progress in electronic publishing has transformed labor-intensive activities into capital- intensive activities. The widespread use of computers in information production has made this apparent. When production is more labor-intensive, sunk costs are more likely because, once expended, labor cannot be used for another product.  [FN84] However, as production becomes more capital-intensive, sunk costs become less likely because capital may be more easily sold or transferred for another use. [FN85] Accordingly, as information production--that is, electronic publishing--becomes more capital-intensive, transferable capital costs replace sunk labor costs. As these sunk costs decrease, free rideability decreases accordingly. [FN86]

B.  How Traditional Electronic Publishing Tamed the Technological Risks to Intellectual Property

  The introduction to this Article acknowledged that new information technologies disrupt historic balances between competing interests in intellectual property, but urged that appraisal of the disruptions focus on the specific free-riding and piracy risks created by the new technology. Acceptable balances among competing interests have been worked out reasonably well in electronic technologies used by proprietary, host-based electronic publishing systems. Before considering new challenges presented by open network environments like the Internet and its World Wide Web, one should consider how the balances have been struck in the more traditional electronic environments. This section considers the scope of legal protection for intellectual property, and the role of pricing, product design, bottlenecks, and enforceability, in the context of well-established, host-based, remotely accessible electronic publishing technologies.

  Digital information technology emerged as a possible threat to intellectual property during the Second World War, when the first digital computer cracked the first enemy cipher or aimed the first bomb. [FN87] The technology reduced the cost of reproduction nearly to zero and facilitated incorporation of one work into another. These characteristics concerned the National Commission on New Technological Uses of Copyrighted Works ("CONTU") [FN88] and made the position of creators of digital works somewhat weaker than the position of creators of works on traditional media in terms of overall protection. These technological threats, however, have been dealt with in pre-Internet environments through pricing and product design, as well as modest adaptation of intellectual property law, especially copyright law.

  Although the case law is limited, [FN89] it is reasonably well accepted that a combination of copyright and contract protects the intellectual property of creators who sell information [FN90] on physical media like magnetic and optical disks, and through centralized host-based dial-up services such as CompuServe, America Online, Westlaw, and Lexis. [FN91] In the market segment involving remote access--the one containing CompuServe, Lexis, and their competitors--the basic product relationships among content originator, electronic "retailer," and customer are fairly well standardized, although pricing varies considerably. In the typical set of relationships, the content originator either transfers its copyright or grants an exclusive license to an enterprise that maintains the host computers and the dial-up communications facilities. For example, John Wiley & Sons, a conventional publisher, owns the copyright and grants a license for certain copyrighted works to West Publishing Company ("West"), in order to make the works accessible through Westlaw.  [FN92] West, in turn, acquires whatever copyright interests the publisher transferred to it under the license. [FN93] West also establishes contractual relationships with all of its customers, limiting what the customers can do with works accessed through Westlaw and obligating them to pay West according to its price schedule for access to Westlaw. West, under its contract with Wiley, pays a royalty. In this arrangement, West is the "retailer," and Wiley is the "originator." [FN94]

  Figure 1 illustrates the typical centralized-host electronic- publishing concept. The server not only provides pointers and other finding and retrieval value, [FN95] but it also publishes content. [FN96] The client deals only with the server, which looks like a conventional publisher. The host ensures payment for its intellectual property by denying a connection to those who have not paid or made payment arrangements.

  A variety of common pricing arrangements exist. [FN97] For example, some works are available through Westlaw based on payment of the basic Westlaw fee, a combination of monthly and hourly charges. Access to other works results in a surcharge in addition to the basic Westlaw subscription fee. The royalty arrangements between originator and retailer also vary. For example, the retailer may keep track of the number of times a particular work is accessed,  [FN98] and make royalty payments to the originator based on numbers of accesses, or it may simply obtain rights based on payment of a fixed fee for a license of a defined duration.

  In current online electronic-publishing arrangements, the single bottleneck of the host-based retailer makes intellectual-property rights sufficiently enforceable to allow creative and entrepreneurial activity to flourish.  [FN99] If a customer does not have a subscription relationship with Westlaw, the customer cannot obtain any information through the Westlaw service, because the first step in any dial-up connection is verification of an active account through an exchange of account name and password information with the host computer. The originator depends on this tollgate, defined in an express contract and enforced technologically, to protect its economic interest and to prevent piracy. The tollgate makes the information product excludable, thus weakening its public-goods character. The tollgate increases a pirate's ca cost.

  To be sure, the nonrival nature of the digital format permits free riding and piracy by customers of the retailer. Once someone accesses a work through Westlaw, the accessor can duplicate the work and sell it in competition with Westlaw. [FN100] Such conduct, however, would not only violate the copyright interest of both originator and retailer, it also would violate the contract between the subscriber and the retailer. Moreover, the granularity [FN101] of most works accessible through Westlaw and other host- based services makes it costly to download large amounts of content, as would be necessary for one wishing to engage in large-scale illegal redissemination.  [FN102] Fine granularity in remotely accessible databases may not only improve performance of the system for a consumer, [FN103] but it may also increase a pirate's costs to the point that piracy is unattractive.

  The existence of a single tollgate makes violations of the copyright and license agreement relatively easy to prove. Records on access exist, and unobtrusive codes can easily be inserted in content to detect unauthorized copies. [FN104] The absence of much case law on license or copyright violations by subscribers to host-based services is consistent with an acceptable level of compliance with copyright and license terms. If widespread piracy existed, one would expect a greater incidence of civil and criminal enforcement actions. [FN105]

C.  Why Internet-Like Architectures Threaten Intellectual Property

  The Internet is a model of an open network architecture that will likely dominate electronic publishing in the future. The transaction costs and the resulting possibilities for excludability and rivalness are significantly different from those of host-based architectures.

  The Internet is an international network of computers and computer networks connected to each other through routers using the TCP/IP protocols and sharing a common name and address space. One can communicate with any computer connected to the Internet simply by establishing a connection to an Internet router or node. The Internet is not a corporation or administrative arrangement; it is a method for connecting computer systems. Voluntary cooperative bodies, such as the Internet Engineering Task Force ("IETF"), discuss and formulate standards and protocols through documents called requests for comments ("RFCs"). The Internet began in the 1960s with federally subsidized connections among universities and government research laboratories. An "acceptable use policy" limited traffic unrelated to research and education. By 1990, the Internet's potential as a model for a National Information Infrastructure ("NII") had been recognized. The federal government began to reduce the subsidy and to encourage private entities to take over responsibility for basic communication and traffic management functions. By 1995, most of the traffic on the Internet involved unsubsidized facilities and private traffic. The Internet is the archetypical open network.

  Widely used Internet applications (in addition to e-mail) include telnet, a method of establishing a remote terminal connection to another computer across the Internet; file transfer protocol ("ftp"), a means for transferring files between computers linked together by the Internet; gopher, a user-friendly menuing system for making files and text available (now largely replaced by the World Wide Web); news and newsgroups, means for electronic discussions in which anyone connected to an Internet node has access to posted messages and replies to them; and the World Wide Web.

  The World Wide Web ("Web") is a particularly popular application for the Internet. Reflecting the client-server model of computer-program design,  [FN106] the Web makes use of two kinds of software: server software and client or "browser" software. The client and the server work together during a Web session, communicating with each other through messages and files conforming to the hypertext transfer protocol ("http") [FN107] and hypertext markup language ("html") standards. [FN108] Documents formatted in html are displayed by browser software that presents colors, typeface styles and sizes, and hypertext links as highlighted text, all according to html instructions or tags.

  An increasingly popular form of electronic publishing involves taking documents, graphical images, sound files, or a combination of these, and placing them on a Web server--usually a small or medium-sized Internet server with several directories devoted to Web files and the server application software. The Web is a hypertext system, meaning that a typical Web document has pointers to other html documents. These other documents may be located anywhere in the Internet--on a Web server, on the same server on which the pointer is located, or on other parts of the same document that contains the pointer.

  For example, one could take this Article and organize it for electronic publishing on a Web server by substituting internal hypertext pointers for all of the cross references and by substituting external hypertext pointers for all of the reference information in the footnotes. When a Web browser displays the document, each of the pointers, internal or external, would appear as highlighted text or as a small symbol. A user interested in the highlighted information would click on the highlighted text with his mouse. This would cause the Web to retrieve the identified information and to automatically display it on the client screen.

  Thus, the Web organizes information distributed across the Internet. It facilitates unbundling because editors or publishers interested in collecting resources related to a particular subject need not obtain or maintain actual copies of the content of the resources; they can make their knowledge available simply by writing a Web document that contains pointers to the identified references and information about the significance of the resources. The clearest example is a typical law review footnote or citation in a legal brief. One can make the cited case or statute available simply by pointing to it in the law review article or brief. A user reading the law review article or brief can retrieve the full text pointed to simply by clicking on the footnote or citation.

  Figure 2 illustrates how the Internet and similar architectures permit value-added products to be unbundled, with different elements of the bundle being supplied by different entities. The bundling of these elements occurs according to the desires of a particular user at the time the user wants the complete value-added information product. In this architecture, suppliers of information content [FN109] supply their content to anyone who wants it simply by putting files on computers connected to the Internet, called servers or "content servers."

  Those value-added features are supplied by others. [FN110] It is entirely possible with Internet applications like the Web for an entity to supply only index or table of contents-type value in the form of pointers to content. The pointers are implemented through Web pages and lists or Gopher menus on an Internet server that offers no other kinds of value. Someone else can provide user-friendly interface software through another server. Yet another person can provide connection services that permit connections through dial-up telephone lines or through higher speed dedicated links.

  When a user wishes to identify and obtain a particular type of content, the user interacts with several Internet servers operated by different entities. First, the user establishes a connection to the Internet through a connection- services provider. The user then establishes a connection to an index provider. From the lists, tables, and menus provided by the index server, the user identifies one or more items of interest. The index server uploads the pointers (not the content, because the index server does not have the content) through the Internet to the user's client computer. Then the user's client computer executes the pointers, which automatically downloads the identified content from the content server into the client computer. [FN111]

  This is a kind of assembly line for pieces of information value. It produces the product just in time, rather than producing it in bulk according to someone else's design. It lets the user design the product on an ad hoc basis.

  Such an infrastructure is not only two layered; in many cases a pointer points, not directly to the full information resource, but to another collection of pointers, which may point to still other collections of pointers, and so on, collectively marking a trail to the complete resource. The computer programs involved assemble a trail from the three pointers and then retrieve the desired content from wherever it resides, directly into the computer of the requester, without the content having to traverse all the intermediary computers. Whether an intermediary points directly or indirectly to the desired resource is inherently an engineering decision driven by performance considerations. In many cases, the computer automatically decides to maintain copies of a particular information resource without any human intervention. A clear example of this is the caching of recently retrieved resources within a Web browser, such as Netscape.

  Figure 3 illustrates a typical set of relationships among Web servers. Web server X (the consumer's intermediary) never possesses the content. Itpoints to server Y (the pure intermediary), which in turn points to server Z (the originator's intermediary). Z, in turn, points to the content on the originator's server. [FN112] Transactions 1-6 with servers X, Y, and Z are requests for, and transmissions of, pointers only. The eventual request for, and

   retrieval of, the content transpires entirely between the client and the content originator, in 7 and 8. [FN113]

  One must ask two sets of questions. First, what features of the technology increase excludability and/or rivalness and thus reduce the need for legal grants of entitlements to stimulate allocative efficiency? Second, how should one describe the transaction costs confronting various participants in the Web and draw conclusions about property versus liability-based entitlements and the identity of beneficiaries? This Article addresses the first question below and discusses different forms of protection for content. It addresses the second question in Part III and discusses intermediary liability.

  One of the core concepts of the Internet and other open architectures is that information storage and processing functions are distributed rather than being centralized in a single host. Another core concept is the absence of proprietary protocols or technologies that isolate relationships among particular originators, retailers, and customers. Instead of an originator like Wiley having a single, exclusive relationship with one retailer like West as in the preceding example, an originator on the Internet is expected to have a multiplicity of relationships with several retailers. The open and distributed characteristics greatly weaken the techniques developed in host-based structures for protection of intellectual property.

  The Web is the best example of how a content originator is threatened. But the Web is also the best example of the potential of Internet-like architectures for new forms of electronic publishing. In the Web, a content originator publishes his work simply by placing a file containing the content on an Internet Web server and allowing access to that server and file through the standard Web protocols, http and html. Anyone who wishes to view or otherwise use that content can do so simply by establishing a connection to the originator's server and pointing to that file

   with a standard pointer in the URL format. [FN114] A potential pirate's ca is low.

  Because search costs would be too great for potential users of information published in this fashion to know about all potential sources, the Web depends on a variety of more or less independent intermediaries who add value to the system by supplying pointers. In other words, someone wishing to access information from law reviews typically starts with an Internet node known for identifying legal resources on the Internet [FN115] and finds on such a service a pointer to a particular law review file on another Internet node. Frequently, a user goes through two or more such intermediaries, perhaps beginning with one identifying a broad range of legal resources, which points to a particular law school, which in turn, points to a particular law review issue and article.

  The breadth of the market reachable by the content originator depends upon a large number of intermediaries pointing to its works. This is good, for the content originator. His revenue potential is much greater than if he were limited to the market reached by a single distributor. [FN116] Usually, the content publishers do not have prior arrangements with the pointer intermediaries; indeed, ordinarily they do not even know about and cannot limit who points to their works. [FN117] This is bad for the content originator. He cannot make contractual arrangements for royalties or fixed-fee licenses with retailers he does not know. In other words, the transaction costs of effective licensing are high.

  This does not change copyright law, and thus does not change the copyright protection for the content originator, except insofar as placement of its work on a Web server impliedly licenses certain uses [FN118] or, less likely, places the work in the public domain. [FN119] Effective contract protection for the content originator, however, is more problematic. [FN120] While the content originator can post various terms and conditions in its work, or can post various messages on its Web server, there is no assurance that an intermediary or an end user will see any of these limitations until it actually accesses the content.

  Intermediaries are even less likely than end users to be bound by posted limitations on use. The intermediary may be entirely unaware of such posted terms of conditions, never seeing the content to which it points. [FN121] The likelihood is somewhat greater that the end users might be bound by terms and conditions because end users are more likely to read the contents at some point, and thus become aware of terms and conditions bound to the content. But even if end users are bound, they are far less satisfactory defendants, from the originator's perspective, than intermediaries, as in conventional host- based arrangements. In other words, the bottleneck that is the heart of intellectual-property protection and host-based networks does not exist naturally in open network architectures. [FN122]

  Moreover, contract protection is unavailable except when privity of contract exists. Thus, the originator's contract position is even weaker with respect to pirates who get access not directly from the originator, but through others, such as intermediaries several steps removed. Privity requires either specialized communications, through, for example, Electronic Data Interchange ("EDI") transaction sets, that satisfy requirements for contractual offer and acceptance without human involvement or contemporaneous human assent to contract terms.

  The possibility exists, of course, of refusing intermediary connections unless prior arrangements have been made with that particular intermediary. The Internet and Web offer the possibility of allowing access to particular files or directories only to preauthorized users who give the correct password.  [FN123] The exchange of user-authorization information can be automated to establish a kind of trusted intermediary network. This technology permits electronic publishing through the Internet to rely on the same user account and password protections as host-based services. The possibility of transforming the Internet-like architectures into something close to host-based architectures, however, is not a fully satisfactory answer. It vitiates many of the advantages of an open architecture for new kinds of publishing and blocks the flow of information, the goal of intellectual property. The transaction costs of making prior arrangements are significant, and a content originator shrinks his market considerably by allowing access only through such prearranged intermediaries.

III.  Intermediary Liability as a Countervailing Problem

  At the same time that open architectures may make it more difficult for content originators to protect their intellectual property, open architectures also may increase the exposure of intermediaries to liability for intellectual-property infringement.

  In the Web, multiple intermediaries may interact to allow a consumer to retrieve information directly from a content originator. [FN124] An information object allegedly infringing another's copyright may move through several intermediaries in this distribution system. The potential liability of the various intermediaries for copyright infringement is a matter of obvious interest both to the intermediaries and to the copyright owner alleging infringement. The copyright owner may perceive the intermediaries as having deeper pockets, or as being more amenable to personal jurisdiction than the originator of the allegedly infringing object. [FN125] Targeting the intermediary for a copyright infringement action is natural, considering the traditional role of the intermediary as a bottleneck on which intellectual- property protection focuses.

  It may be relatively easy for the person claiming infringement to establish a violation of one of the exclusive section 106 rights [FN126] by the originator of the allegedly infringing item. The originator (pirate) almost certainly has reproduced the copyrighted work and also probably helped to distribute it to the public. [FN127] It is somewhat less clear whether the intermediaries have infringed any section 106 rights; this depends on the network architecture.

  Intermediary liability is more likely in traditional bulletin-board arrangements [FN128] than in the Web. Consider the case of a dial-up electronic bulletin board, where any item selected and retrieved actually resides in the form of a file on the intermediary computer. [FN129] When the user selects a file, including an infringing one, the intermediary computer makes a copy and downloads it to the user's computer. When the intermediary makes the copy, it violates the section 106 reproduction right; when the intermediary downloads, it may violate the section 106 distribution right if it is aimed at the "public." The owner of the bulletin board may argue that it did not place the allegedly infringing item on its computer and did not itself cause the copy to be made. Rather, one copy was made by the third person placing the item on the bulletin board, and another by the user retrieving it. Such arguments failed in Playboy Enterprises, Inc. v Frena, [FN130] where the district court found that the operator of the electronic bulletin board on which third parties placed digitized images of Playboy centerfolds infringed Playboy's distribution rights as the content originator. [FN131] Moreover, such a bulletin-board operator may have publicly performed or displayed the work [FN132] when a user browsed material on the bulletin board. [FN133]

  Intent is irrelevant when conduct falling within section 106 occurs. Playboy Enterprises held that "(i)t does not matter that (a defendant) may have been unaware of the copyright infringement. Intent to infringe is not needed to find copyright infringement." [FN134] In other words, intermediaries are subject to no-fault liability. [FN135]

  Technology matters. The infringement case against intermediaries who provide anonymous ftp sites is essentially the same as the case against electronic bulletin-board operators. On the other hand, the case against Gopher and Web servers is more tenuous. They never have possession of a requested item, which makes it less plausible that they have reproduced, distributed, displayed, or performed it. [FN136] Rather, they are more like providers of bibliographies to works, some of which may be infringing.

  Nevertheless, copyright owners may argue that the scope of the section 106 duties should be extended to intermediaries as well, because they constitute links in a chain that necessarily results in the conduct covered by section 106. Copyright infringement is a statutory tort. The causation requirement in tort law long has been satisfied by the doing of an act intended to and substantially certain to result in the injury constituting the tort. [FN137]

  Moreover, even if Gopher and Web-type services do not expose their providers to liability for direct infringement, as in Playboy Enterprises, they may expose their providers to liability for derivative or contributory infringement. [FN138] In Sony Corporation of America v Universal City Studios, Inc., [FN139] the leading case on contributory infringement, the Supreme Court held that manufacturers of video cassette recorders were not liable for contributory infringement of televised works recorded on their machines by consumers. [FN140] The Court held that systems capable of substantial noninfringing uses cannot produce contributory infringement.  [FN141] In order to apply this test to the Gopher or Web server, one must ask whether the particular server facilitates access to noninfringing information objects. If its purpose is not infringing, the fact that consumers may occasionally use the server to retrieve infringing objects does not suffice to establish contributory infringement by the operator of the server. On the other hand, if the server has the predominant purpose of facilitating access to infringing objects, the operator of the server should be liable for contributory infringement.

  The White Paper apparently envisions a more extensive role for contributory liability. In regard to intermediary liability, the White Paper states:

  There is a view that on-line service providers, such as bulletin board operators, should be exempt from liability or given a higher standard for liability, such as imposing liability only in those cases where infringement was willful and repeated or where it was proven that the service provider had both "actual knowledge" of the infringing activity and the "ability and authority" to terminate such activity. The latter proposed standard would combine the contributory infringement standard with the requirements for vicarious liability and apply it to all infringements (including direct infringements) of the service provider. Altering the standards of liability for infringement would be a significant departure from current copyright principles and law and would result in a substantial derogation of the rights of copyright owners. It is a difficult issue, with colorable arguments on each side. [FN142]

  While acknowledging that it is virtually impossible for intermediaries to review messages and files for infringement possibilities contemporaneously, the White Paper also identifies other actors in the larger information infrastructure confronted with the same impossibility of screening. These actors include photo finishers, booksellers, record stores, newsstands, and computer-software retailers, all of whom nevertheless are subject to strict liability for infringement. [FN143] The White Paper also states that "on- line service providers can certainly investigate and take appropriate action when notified of the existence of infringing material on their systems and thus limit their liability for damages to those for infringement." [FN144] This has been part of the business costs for many other information distributors, and the authors of the White Paper saw no reason that NII intermediaries should be placed in a more favorable position. [FN145]

  The White Paper then focuses on a subclass of intermediaries. "On-line service providers have a business relationship with their subscribers. They-- and, perhaps, only they--are in the position to know the identity and activities of their subscribers and to stop unlawful activities." [FN146] This, of course, is not true of many Web servers. The White Paper acknowledges that "(n)o one rule may be appropriate." [FN147] For example, an entity providing:

  Only the wires and conduits--such as the telephone company . . . would have a good argument for an exemption if it was truly in the same position as a common carrier and could not control who or what was on its system. The same could be true for an on-line service provider who unknowingly transmitted encrypted infringing material. [FN148]

  The White Paper states that Congress believes that the limitation on damages for innocent infringers is adequate to protect intermediaries while sufficient to create incentives for intermediaries to work with copyright owners and their subscribers to protect against infringement. [FN149]

  The White Paper reviews various forms of vicarious liability as a basis for holding intermediaries liable. The dance-hall cases justify holding liable any intermediary who authorizes infringing uses of his facility, such as the sponsor of an anonymous ftp server or electronic bulletin-board design for uploading infringing material. [FN150] More broadly, the White Paper suggests that an intermediary who provides services or equipment relating to the direct infringement may be liable for contributory infringement. [FN151] It also suggests that "infringement liability may be based on the provision of equipment or other instrumentalities or goods used in or related to the infringement," [FN152] acknowledging Sony. [FN153] One could interpret the White Paper as arguing for a very broad regime of contributory infringement that might conceivably even sweep up Web and Gopher intermediaries who never engage in reproduction or direct distribution of copies.

  Even under a more traditional application of infringement concepts, Web intermediaries might be liable for making and distributing copies because they use caching. Caching refers to the automatic copying of material to improve performance of computer systems. It is used in wide-area networks like the Internet to reduce the time required for second and subsequent access to Web files. Because caching involves copying, it raises particular concerns about potential copyright violations.

  This Article presumes that the supplier of information content has made it available on the Web through a "content server" ("CS"), a Web server from which the content may be retrieved. The Article further assumes that this content supplier has imposed restrictions on the content thus published, permitting access for purposes of browsing and viewing, but not copying. [FN154]

Caching may constitute prima facie direct copyright infringement. [FN155] Suppose that there are three entities, besides the content server, involved in a Web transaction: an ultimate consumer, C; another Internet server that provides pointers on its Web pages to the content server, PS; and fire wall computers, F.

  First, suppose C initially establishes a session with PS to locate material of interest. C finds on PS a pointer to an item on CS. C clicks on this pointer. As a result, PS automatically uploads the selected pointer (in the form of a URL) to C's computer. [FN156] Now, the Web-browser software running on C's computer automatically establishes a connection with CS and retrieves the Web file pointed to by the pointer. C's Web browser caches the retrieved file so that if C requests it again, the cache copy can be loaded in the browser, rather than the browser having to fetch the file again over the Internet. In this transaction, PS never had the requested file, either before or after C requested it. Any downloading and caching occurred on the Web- browser application running on C's client computer.

  As a second example, suppose that C's computer connects to a Local Area Network that in turn connects to the Internet through a fire-wall computer. While there are a variety of fire-wall configurations, a common one would cause the fire-wall to serve as a "proxy" for C. The popular Web-browser software has a proxy option to accommodate such a fire wall role. The proxy function is best defined by describing how it works in a Web transaction. C seeks to establish a connection to PS. The connection request is forwarded to F, which duplicates the request. Any information returned by PS, including the pointer, is transferred to F. F then caches the information and sends a copy along to C. When C's computer activates the pointer to CS, CS returns the requested file to F, which caches it, and sends a copy along to C. C also caches it because it is running the same Web browser as in the first example.

  Some Web browsers, such as Netscape, keep the cached material even after the user exits the browser. The next time the browser runs, the cached files are available to it. An expiration date, included in the files sent by CS, determines how long a browser uses a cached version rather than reloading from the server from which the material originated. CS sets this expiration date, but the recipient may change it. [FN157] Thus, it is not factually correct to conceive of the browsers'--and the fire walls'--caching as ephemeral or transitory.

  As a third example, consider an information-services provider such as CompuServe, Prodigy, or America Online. Such a service provider might want to provide Web access to its customers so that information retrieved through the Web would be more or less indistinguishable from material retrieved from the provider's host computer. The provider would almost certainly also cache the Web material on the provider's host, so that if another subscriber wishes to retrieve the same material, the second subscriber could do so directly from the provider's host computer, without the host having to go out and fetch the material again across the Internet.

  Despite the engineering benefits of these three types of caching--faster response and less traffic on the network--all three types involve reproduction and thus potential copyright liability. The uncertainties confronting intermediaries can lead to a kind of de facto prior restraint on controversial content. The possibility of such prior restraint should stimulate more careful consideration of the tension between First Amendment and intellectual-property interests. [FN158]

  Even a fault-based standard forces an intermediary to be the judge of a copyright-infringement allegation. A mere allegation or rumor of infringing content may constitute "knowledge," thus exposing the intermediary to increasing damages the longer it permits the content to be disseminated through its service. Investigation and adjudication of copyright cases are expensive and require legal expertise. Because intermediaries face almost no liability if they remove a particular item or a pointer to that item, the economics of intermediary services dictate that in almost all cases, the intermediary simply will stop disseminating an item merely on accusation or rumor of infringement.

  Even in those instances in which a particular configuration of technology reduces the risk of no-fault direct infringement, as in the Web without server caching, intermediaries potentially are exposed to liability as contributory infringers. For example, a request to suppress certain material might give the intermediaries sufficient knowledge to make them liable for contributory infringement under the White Paper formulation.

  The White Paper acknowledges that the nature of increasingly complex intermediation technologies makes it infeasible for many intermediaries to check out the bona fides of the information they assist in disseminating. It may be feasible for a service provider that actually maintains copies of files and messages on its own computer to screen for offensive words or identifiable files. It certainly is not feasible, however, to automate screening for copyright or trademark infringement, especially for a provider who merely points to information resources created and maintained by others or who merely uploads textual files. Any attempt at such screening would clog up the dissemination technologies and dissipate their power. "Fair-use" arguments are subtle and cannot be decided by a computer algorithm, much less an automated filter that cannot possibly "know" enough about the surrounding circumstances to determine whether a particular file or message infringes someone else's copyright. Because of the volume of messages, inquiry into detailed factual circumstances would impose a major burden on providers.

  Thus, de facto prior restraint will occur because of the combination of the rules of intermediary liability for copyright infringement and the economics of intermediaries in the information infrastructure. The intermediary will likely impose prior restraint at the first hint of a controversy over intellectual property. The intermediary is subject to an injunction or criminal prosecution in advance of the dissemination, and the possibility of a damages judgment works to shut down the dissemination just as effectively.

  If courts give intermediaries more guidance as to what does and does not expose them to liability, and if the law insulates an intermediary from damages unless and until the intermediary receives a neutral determination that a particular item is infringing, the de facto prior-restraint threat would be reduced.

IV.  Encryption's Potential

  Encryption is attractive as a way of reducing free-riding potential without heavy reliance on intellectual-property law and without relying unduly on intermediary liability. However, there are important limitations on encryption as a solution.

  Content encryption extends technical protection beyond value-added processes to the content itself. A variety of proposals for digital libraries would rely on encrypted content in order to make the otherwise public good excludable. [FN159] They all rely on the basic concept that one can decrypt the desired information content only by using a key for which one would have to pay. Encryption applications deploy systems that would limit the number of times encrypted content could be accessed. The systems also would permit certain types of use and exclude other types, such as replication. Such encryption applications increase the pirate's ca, potentially to infinity. The same applications permit one to determine with high reliability whether content has been altered, such as by the removal of various notices and advertisements. [FN160] The applications increase the pirate's cll.

  Encryption has important limitations, however, that prevent its use as the main means of protecting intellectual property in the information infrastructure. Encryption works only when both the producer and the consumer of information adhere to the encryption standards. Nothing in the history of developing standards for computer formats suggests optimism about the feasibility of developing a truly universal standard for encryption. [FN161] Nor is it feasible or desirable for the government to mandate adherence to such a standard. Such a mandate would put the government in the position of prohibiting expression except through the approved encryption standard, and that would raise insuperable First Amendment problems. The experience with the proposed Clipper-chip initiative discourages enthusiasm about the potential for government-led standardization of encryption methods. The original proposal did not contemplate mandating the use of the Clipper standard. Nevertheless, the outcry against it was so great that it had to be withdrawn. [FN162]

  Moreover, any method of encryption has adverse implications for information-system performance or reliability. Copy protection of desktop-computer software revealed how strongly consumers resist these limitations. In the face of such consumer resistance, copy protection that was commonplace in 1985 had virtually disappeared from mass-marketed computer software by 1995. One can expect the same kind of resistance with respect to encryption arrangements that sometimes block access by authorized users and impede performance.

  For those markets in which consumer resistance impedes more profitable and larger scale electronic publishing, adding a technological basis for additional consumer resistance likely will not appeal to entrepreneurs who want market development. In other words, encryption may increase the originator's cm more than it increases the pirate's ca.

  Undue reliance on encryption solutions to intellectual challenges could produce a plethora of incompatible encryption systems. A collection of technically isolated archipelagos would exist, rather than a national information infrastructure, much as the online information services like CompuServe, America Online, Lexis/Nexis, Dialog, and Westlaw existed in isolation from each other before the NII encouraged interconnections and gateways.

  Encryption technology is best used not to protect content, but to facilitate payment systems so that unencrypted content may be sold cheaply in open network architectures.

V.  Solutions

  A regime that protects intellectual property and also shelters new creative and productive exploitation of information must target realistic transaction costs that impede bargaining. The attributes of public goods justify the artificial property constructs of copyright and patent law. The best policy is to leave copyright law intact, develop a richer array of pricing and business models for electronic publishing, develop private cooperative arrangements for licensing and policing intellectual property, and extend the infrastructure for electronic-payment systems.

  If the Copyright Act is amended, it should be amended not only to extend protection for content originators, as the White Paper suggests, but also to offer certain immunities for intermediaries in open network architectures, and to strengthen the role of the First Amendment.

A.  Proposals for Expanded Legal Protection Are Unsupported

  The Administration's White Paper mistakenly addresses new technological threats to intellectual property through the law and through expanding protection for content originators, and it minimizes the concerns of consumers and intermediaries. While claiming an intent only to "clarify," the White Paper recommends the establishment of a new transmission right in copyright law:

  It is not clear under the current law that a transmission can constitute a distribution of copies or phonorecords of a work. Yet, in the world of high speed communications systems, it is possible to transmit a copy of a work from one location to another. This may be the case, for instance, when a computer program is transmitted from one computer to ten other computers. When the transmission is complete, the original copy typically remains in the transmitting computer and a copy resides in the memory of, or in storage devices associated with, each of the other computers. The transmission results essentially in the distribution of ten copies of the work. However, the extent of the distribution right under the present law may be somewhat uncertain and subject to challenge. Therefore, the Working Group recommends that the Copyright Act be amended to expressly recognize that copies or phonorecords of works can be distributed to the public by transmission, and that such transmissions fall within the exclusive distribution right of the copyright owner. [FN163]

  According to the White Paper, the proposed amendment simply  "recognizes" the appropriate scope of the distribution right. [FN164] In the view of the Working Group, the amendment is necessary, among other things, to make it clear that distribution in the form of transmission is the exclusive prerogative of the copyright owner. The distribution right as currently interpreted covers only "publication." [FN165]

  The White Paper rejects any special safe harbors for intermediaries. While recognizing that intermediaries, such as online service providers and Internet servers, "play an integral role in the development of the NII and facilitate and promote the free exchange of ideas," [FN166] the White Paper suggests that this should not be grounds for removing or reducing liability for copyright infringement. "One can perform these functions without infringing or facilitating the infringement of the copyrighted expression of others."  [FN167] The White Paper further argues that intermediaries remain in the best position to detect copyright infringement, and that they could shift any undue risk of strict liability to their customers or suppliers through routine contracting arrangements. [FN168]

  The White Paper is troubling because is suggests that fair use has a narrow scope. It misleadingly de-emphasizes the role of exploitation of information by downstream authors and publishers as a core part of copyright analysis. In a similar vein, the White Paper overemphasizes the scope of vicarious liability under doctrines such as contributory infringement. The White Paper jumps to the conclusion that more legal protection is needed without carefully probing the technological and marketing underpinning of such a conclusion.

  A number of intellectual-property scholars proposed, in an extensive article published by the Columbia Law Review in late 1994, [FN169] a new system of protection for computer works. While this Manifesto is an incisive starting point for reconsidering the opposing positions in the controversy over protection of "look and feel" in computer program functioning, [FN170] it is of only limited use in shaping an appropriate form of property for information content. The Manifesto characterizes computer programs as processes. The distinction between computer programs and computer data is likely to become even more indistinct than at present, as html tags embody more procedural information and as object-oriented systems such as Java become more common. Information content, however, will remain more like data than programs. As such, a process-oriented intellectual-property regime is largely immaterial to protection of content contributors. Nevertheless, the Manifesto represents a more useful contribution to the policy dialogue than the White Paper. The Manifesto seeks to work out legal protections to fill actual gaps in technological protections of creative effort, rather than jumping to the conclusion that an expanded role for the law is desirable.

B.  Embrace New Pricing and Business Models

  Protecting intellectual property in a flourishing NII depends more on the imagination of entrepreneurs in designing and deploying new business and pricing models than on legal creativity. Radio and television patterns should be imitated; they were the last big revolution in information technology and necessitated changes in the way intellectual property was handled. To be sure, radio and television entrepreneurs did not forswear intellectual property, but traditional notions of copyright fit poorly because radio and television programming was not distributed in the form of tangible copies. To respond to these challenges, the entrepreneurs radically changed the location of the tollgate. [FN171]

  In cyberspace, new tollgates and new ways of pricing will enable content contributors to earn a fair return while undercutting efforts by pirates to free ride. For example, the costs of extracting large amounts of content from finely grained retrieval systems can be substantial. If persons cooperate in the marketing of information through such systems and set a price that is not much greater than a pirate's costs to extract the information unlawfully,  [FN172] then the economic incentives for piracy diminish. Alternatively, block pricing for access to licensed content on a nonexclusive basis reduces transaction costs [FN173] for content contributors and intermediaries. As their transaction costs diminish, [FN174] the margin for a pirate diminishes as well.

  The spread of nonexclusive publishing arrangements can assure content contributors a greater return because of larger markets. At the same time, these arrangements reduce the likelihood of closed systems swamping a surrounding open network architecture and reduce piracy incentives because content is available to a wider variety of intermediaries who competitively serve demand better. [FN175] Already, the popularity of the Internet has changed bargaining power and bargaining positions between content providers and intermediaries. [FN176] Whether or not authors agree to exclusive contracts with their publishers is purely a matter of bargaining power and practice.  [FN177]

C.  Methods of Protecting Content and their Classification

  World Wide Web technologies permit the originator of content to protect her investment without relying entirely on intellectual property. These new technology-enabled methods, summarized in Figure 4, fall into three overlapping categories: product design, business models, and technology.

  All three methods reduce the nonexcludability and nonrival characteristics of content sold by itself. [FN178] Product-design methods include planned obsolescence, finer granularity, addition of presentation markup, bundling vulnerable material to less vulnerable value-added elements, and concentrating investment on value-added features other than content. Planned obsolescence decreases the value of the content item that might be subject to piracy. By the time the pirate brings his copy to the market, consumers will prefer a more current version from the originator. Finer granularity and the addition of presentation markup increase ct because the pirate must assemble more grains and remove presentation markup. Finer granularity also may increase ca because more human intervention likely will be needed to download a marketable amount of material. [FN179]

  Bundling vulnerable material to less vulnerable value-added elements increases ca because it becomes more difficult for the pirate to get to the content he wishes to steal. [FN180] Shifting investment toward value-added elements and away from content might seem inefficient because it would result in the underproduction of content. But different ways to satisfy consumer needs exist, some exposing the originator to a lower risk of free riding. For example, one can author a work so that different structural components are identified by transitional language in the text. Alternatively, one can segment the work and make different parts retrievable separately. Writing the work so that its segments can be identified by text means that investment in structural differentiation is made in the form of compensation to the author. If one segments the work, one can invest separately in content authorship and the tags, pointers, and software that retrieve the individual segments and link them together. The value-added features are inherently more excludable, potentially confronting the pirate with very high ca and ct.

  A number of feasible business models exist in the Web and compensate for remaining nonrival and nonexcludability characteristics of content. The producers of downstream value may make subsidy payments to content producers in order to draw customers to their access and retrieval systems. They can reduce the necessary outlays if they cooperate with other producers of downstream value and make nonexclusive arrangements with content producers. A variety of institutional arrangements are also conceivable. Content producers could give global licenses in exchange for a share of aggregate revenues flowing to all producers of downstream value. Third parties could subsidize the production of content, as in the business model for television, radio broadcasting, and newspapers. Proprietors of electronic markets, such as automated help- wanted advertising sections, can buy content in order to attract customers to their markets.

  Technology also permits direct changes in the degree of excludability and nonrivalness. In Westlaw and CompuServe, information is excludable because the electronic gateway into the computers through which information is dispensed does not allow consumers and potential pirates into the system unless they have an account. Consumers must pay the bill to maintain their account. The same concept can be deployed in the Web. Particular items of information may be protected by passwords on certain Web pages by automating the login process for those particular pages.

  Commentators have proposed a variety of methods for content encryption and copyright management that use technology not only to make information excludable, but also to make it rival. One person's use thus defeats the use entitlement of the predecessor. [FN181]

  Technology also can reduce transaction costs for exchanges between buyers and sellers of information. For example, VISA andMastercard recently published standard documents for credit card transactions on the Internet. These documents likely will hasten the deployment and use of payment systems that permit low-cost exchanges of chunks of information so small that they are not worthwhile to steal. Thus, the payment-system technology and product design together reduce the relative transaction costs to payers and increase the acquisition and transformation costs for nonpayers.

D.  The Role of Private Cooperatives

  Historically, copyright-licensing collectives have played a useful role in reducing the transaction costs of disseminating intellectual property to authorized redistributors [FN182] and in reducing the costs of detecting violations. [FN183] The American Society of Composers, Authors, and Publishers ("ASCAP"), [FN184] Broadcast Music, Inc. ("BMI"),  [FN185] and the newer Copyright Clearance Center ("CCC") [FN186] have demonstrated their capability in this regard. Similar institutions can arise in the electronic environment, although they have not done so yet. [FN187] It is useful to identify the possible functions and configurations for copyright collectives in environments like the Web.

  1.  Detect violations.

  An intellectual-property cooperative can assist enforcement by detecting infringement. The cooperative could monitor advertisements and other promotional efforts by pirates, sample the virtual places in which infringement is likely, [FN188] and visit places in which jukeboxes and newer technological means are installed. [FN189]

  The cooperative could deploy "sniffers" that would detect bit patterns suggestive of infringing conduct. For example, if legitimate publishers use a common signature or trademark to signify authorization to publish copyrighted works, the sniffer could look for instances of that bit pattern transmitted by unauthorized servers on a network. Or, in the case of extremely popular works, the snif