101
UNIVERSITY OF VIRGINIA SCHOOL OF LAW
2003 SPRING PUBLIC LAW AND LEGAL THEORY RESEARCH PAPERS SERIES
WHEN CODE ISN'T LAW
TIM WU
WORKING PAPER NO. 03-10
This Paper Can Be Downloaded Without Charge From The
Social Science Research Network Electronic Paper Collection
Http://Ssrn.Com/Abstract_Id=413201
102 Virginia Law Review Vol. 89:nnn
2003] Compliance & Code 103
VIRGINIA LAW REVIEW
VOLUME 89 JUNE 2003 NUMBER 4
ARTICLE
WHEN CODE ISN'T LAW
Tim Wu
INTRODUCTION...................................................................................104
I. A THEORY OF CODE, CHANGE, AND AVOIDANCE...........................110
A. Reactions to Law in General Theories of Regulation .............110
B. When Groups Get Sick of Complying.....................................112
1. Avoidance Mechanisms ...................................................112
2. Change Mechanisms........................................................116
3. Summary: The Change/Avoidance Choice .......................118
C. Group Dynamics, Collective Action .......................................120
E. Avoidance, Change, and Regulatory Competition ..................126
F. How Code is Used to Avoid Law............................................128
II. COPYRIGHŤS LOOPHOLES ............................................................130
A. Copyright and Its Gatekeepers ..............................................132
B. The Erosion of the Gatekeeper System ...................................137
C. Elements of Peer Design........................................................138
D. Purity in Peer Design ............................................................140
E. Copyright and Social Norms..................................................143
III. THE EVOLUTION OF P2P DESIGN AND REGULATORY
COMPETITION...............................................................................147
A. Napster and its Predecessors.................................................147

Tim Wu, Associate Professor of Law, University of Virginia School of Law. I am grateful
to Barb Armacost, Jack Goldsmith, Scott Hemphill, Mitchell Kane, Jody Kraus, Lawrence
Lessig, Clarisa Long, Tom Nachbar, Irene Oh, Eric Posner, and Matthew Schruers for
discussion and comments on earlier and much earlier drafts, to workshop participants at the
2002 Canadian Law & Economics Conference and the Virginia Law School Faculty Workshop;
to the Stanford Law School Internet and Society Speaker Series; and to Miriam Cho
and Kelly DeMarchis for research assistance.
104 Virginia Law Review Vol. 89:nnn
B. Early Gnutella: 2000-2001....................................................151
C. The KaZaA Era: 2001 - Present.............................................154
1. FastTrack & KaZaA........................................................154
2. Next-Generation Gnutella ...............................................156
D. FastTrack and Gnutella Go to Court .....................................157
E. The Reaction to the Reaction .................................................160
1. Investments in Change.....................................................160
2. Extralegal Investments ....................................................162
IV. THE SOCIAL DYNAMICS OF P2P FILESHARING ...............................164
A. Copyrighťs Divided Subjects ................................................165
B. Disorganized Political Action ................................................166
CONCLUSION......................................................................................170
INTRODUCTION
hen the Supreme Court upheld extended copyright terms in Eldred
v. Ascroft,1
many Internet activists called for renewed political
action in the form of appeals to Congress or even a campaign
to amend the Constitution. But others suggested a very different
course: They argued that it would be wiser to forgo institutions controlled
by the powers of the past, and to return instead to the keyboard to
write the next generation of "law-busting" code. In the words of one observer,
"tech people are probably better off spending their energy writing
code than being part of the political process" because "[t]haťs where
their competitive advantage lies."2
The idea that computer code may be emerging as a meaningful instrument
of political will remains one of the most evocative and poorly
understood propositions in the study of law and technology. The prominent
effects of computer code have made it difficult to ignore the fact
that code can be used to produce regulatory effects similar to laws.
Hence, the popularity of Professor Lawrence Lessig's idea that (for
computer users at least) "code is law."3
But what this really means remains extremely vague. The subject remains
the focus of grand speculation, ranging from claims that computer
1
123 S. Ct. 769 (2003).
2
Declan McCullagh, Geeks in government: A good idea?, at http://news.com.com/
2010-1071-949275.html (Aug. 12, 2002) (on file with the Virginia Law Review Association)
(quoting Sonia Arrison of the Pacific Research Institute).
3
See Lawrence Lessig, Code and Other Laws of Cyberspace 89 (1999).
W
2003] Compliance & Code 105
code will arise as a kind of utopian sovereign to improve on perceived
failures of state regulation,4
to concerns that code may be used to negate
basic freedoms,5
and, of course, the claim that nothing of legal novelty
has happened, or perhaps ever will happen.6
Most problematically, none of these understandings of code and law
explains a central issue: compliance. Specifically, they do not explain
the shifting patterns of legal compliance in the 2000s. Explosions of
non-compliance in areas such as copyright, pornography, financial
fraud, and prescription drugs fuel the sense of a legal breakdown, yet the
vast majority of laws remains unaffected. The mixed compliance pattern
finds little explanation in the concept that "code is law" or in notions
that technological self-help can offer a substitute for legal systems.7
This Article proposes a new and concrete way to understand the relationship
between code and compliance with law. I propose to study the
design of code as an aspect of interest group behavior: as simply one of
several mechanisms that groups use to minimize legal costs. Code design,
in other words, can be usefully studied as an alternative to lobbying
campaigns, tax avoidance, or any other approach that a group might use
4
These claims are described and discussed in Timothy Wu, When Law & the Internet First
Met, 3 Green Bag 2d 171, 172­73 (2000). See also Tom W. Bell, Escape from Copyright:
Market Success vs. Statutory Failure in the Protection of Expressive Works, 69 U. Cin. L.
Rev. 741 (2001) (arguing that the efficacy of technological self-help should allow voluntary
exit from the copyright regime); Kenneth W. Dam, Self-Help in the Digital Jungle, 28 J. Legal
Stud. 393 (1999) (arguing that technological self-help will play a positive role in the
growth of electronic commerce).
5
See Lessig, supra note 3, at 233.
6
See, e.g., Bruce P. Keller, The Game's the Same: Why Gambling in Cyberspace Violates
Federal Law, 108 Yale L.J. 1569 (1999) (arguing that Internet gambling should be regulated
as usual); Edward A. Morse, State Taxation of Internet Commerce: Something New Under
the Sun?, 30 Creighton L. Rev. 1113 (1997) (arguing that issues of state taxation of Internetbased
commerce are familiar); James B. Speta, Internet Theology, 2 Green Bag 2d 227
(1999) (arguing that Internet publication does not justify major changes to the First Amendment
regime). One could attribute this view to Jack Goldsmith, though in his view, unfairly.
See Jack L. Goldsmith, Against Cyberanarchy, 65 U. Chi. L. Rev. 1199, 1201 (1998).
7
E.g., Lessig, supra note 3; Bell, supra note 4. Nor does the scholarship examining the
metaphors used for Internet conduct explain compliance patterns. See, e.g., Dan Hunter, Cyberspace
as Place, and the Tragedy of the Digital Anticommons, 90 Cal. L. Rev. (forthcoming
2003) (noting the persistence of the space metaphor); Orin S. Kerr, The Problem of Perspective
in Internet Law, 91 Geo. L.J. (forthcoming 2003) (arguing that technological
perspectives decide Internet cases); Timothy Wu, Application-Centered Internet Analysis, 85
Va. L. Rev. 1163 (1999) (arguing that analysis should focus on application development);
Alfred C. Yen, Western Frontier or Feudal Society?: Metaphors and Perceptions of Cyberspace,
17 Berkeley Tech. L.J. 1207 (2002) (comparing metaphors of the frontier with cyber-
space).
106 Virginia Law Review Vol. 89:nnn
to seek legal advantage. The approach aims to separate two different aspects
of code's relatio nship with law. The first is Lessig's concept of a
regulatory mechanism: that computer code can substitute for law or
other forms of regulation. The second aspect is as an anti-regulatory
mechanism: a tool to minimize the costs of law that certain groups will
use to their advantage.
The code designer, I suggest, redesigns behavior for legal advantage.
The programmer is not unlike the tax lawyer, exploiting differences between
stated goals of the law, and its legal or practical limits. He targets
specific weaknesses in legal regimes, and has no means to rewrite laws
in general. Therefore, I argue that the long-term significance of the programmer's
methods for the legal system can be described in a fairly
straightforward fashion. At its greatest extent, the design of code may
provide a new option for influencing specific laws. It will be of the
greatest importance to individuals or large, disorganized groups poorly
equipped to take advantage of existing means of political influence. And
as such, the code option may mean some change in the relative power of
interest groups, as it makes organization slightly less important.
The gains to diffuse groups may seem to be a positive development.
But there is a darker side. Code design, as a means of avoiding laws,
serves as a particularly useful device for exploiting the internal dynamics
of regulated groups. It is, as this Article shows, a useful way for the
computer-savvy to avoid legal burdens while continuing to enjoy the
benefits of an ordered society, thanks to the continued compliance of the
technophobic.
In short, increasing use of code to minimize the burden of laws has
interesting and complicated effects for both the legal system and political
system that have been misunderstood. The effects are categorically
different from the fundamental challenge to the legal system that some
had imagined, and analytically distinct from the concept that code is a
form of regulation.
The important case of peer-to-peer ("P2P") filesharing, explored in
depth in this Article, illustrates the possibility of using code design as an
alternative mechanism of interest group behavior. These ingenious programs,
bearing names like "KaZaA" and "BearShare," make it free and
easy to trade digital content (usually copyrighted songs) with millions of
new-found friends.
P2P filesharing represents the most amibtious effort to undermine an
existing legal system using computer code. The significance of P2P for
2003] Compliance & Code 107
copyright is substantial. The efforts of P2P programmers have provided
computer-savvy music listeners with a continuing reduction in the costs
of the copyright system, comparable to a temporary repeal of copyright
laws for computer geeks. P2P underlines the reality of code design as an
alternative mechanism of interest group behavior.
But P2P filesharing also makes the limits of this alternative mechanism
clear. The efficacy of P2P filesharing depends on two powerful and
often unrecognized weaknesses of the copyright regime: the law's dependence
on a gatekeeper enforcement mechanism and the severe lack
of normative support among the regulated. Successful P2P networks
relegate the law to an exercise in primary enforcement against a multitude
of end-users.
P2P's success may depend on a unique collective action dynamic
among music consumers that stems from the nature of copyrighted
works. The works available on peer networks are generally nonrivalrous
goods.8
As a result, the sub-group of P2P users, young and
computer-savvy,9
can take advantage of the continued compliance of
regular consumers. The mass of regular users pay for the works, thereby
maintaining incentives for artists to create them, while the P2P subgroup
defects en masse, occupying the game-theorisťs version of utopia.
These weaknesses, however, represent unique problems for copyright
law and are not more generalized weaknesses of the legal system. For
that reason, the utility of P2P as a means of avoiding law appears limited.
Only a few regimes may contain other particularized enforcement
weaknesses comparable to those of the copyright regime.
This Article's claims will rely on a model of compliance and interest
group behavior with certain novel features. Namely, the focus is on the
mechanisms through which groups influence law. It is the goal of Part I
of the Article to make the underlying model clear.
Laws impose costs upon regulated groups. Those groups that seek to
minimize the costs of law face a fundamental choice between mecha-
8
That is to say, one individuaľs consumption does not diminish another user's value of
the product. I emphasize this characteristic because some scholars have suggested that songs
on peer networks display rivalrous features. See Ramayya Krishnan et al., The Economics of
Peer-To-Peer Networks at 5 (August 2002 draft), available at
http://www.heinz.cmu.edu/~mds/ (on file with the Virginia Law Review Association).
9
According to an Ipsos-Reid study, those who use peer filesharing networks are predominantly
between the ages of twelve and twenty-four. See Robyn Greenspan, Making Money
on Free Music, at http://www.internetnews.com/stats/article.php/
1365161 (last visited May 12, 2003) (on file with the Virginia Law Review Association).
108 Virginia Law Review Vol. 89:nnn
nisms of change and avoidance. Both mechanisms have the effect of
lowering the expected costs of law, but the similarities end there.
Mechanisms of change (principally lobbying) decrease the sanction attached
to certain conduct and tend to require collective action. Mechanisms
of avoidance, on the other hand, decrease the probability of detection
and typically do not require that groups act collectively, but depend
on specific vulnerabilities in the law.
This understanding, while not exhaustive, is descriptively useful even
in the simple form presented. It clarifies the link between problems of
compliance and group dynamics--the extent of organization of the regulated.
It shows that changes in the costs of mechanisms of legal influence
can dramatically affect the function of a given law.
Part I will conclude by explaining how the design of code can be
viewed as a mechanism of legal influence. It will argue that code is used
to reshape behavior to take advantage of loopholes and ambiguities in
legal systems. As such, code is a mechanism of avoidance, displaying
the properties of avoidance described in the model.
Part II will demonstrate the unique vulnerabilities of copyright laws
and code's ability to exploit those weaknesses. Copyright enforcement
has long relied on what Professor Reinier Kraakman first called a gatekeeper
regime.10
In other words, the copyright regime has achieved its
goals through enforcement against specialized intermediaries--those capable
of distributing creative works on a mass scale. Peer networks exploit
that enforcement structure by creating a distribution network that
eliminates intermediaries. While eliminating intermediaries presents a
serious technical challenge, the goal is clear--to remove the enforcement
efficiency of a gatekeeper system, leaving primary enforcement
against end-users as the only option.
P2P networks also exploit an important ambiguity regarding the ethics
of home copying. Compliance with laws pertaining to the theft of
real property is facilitated in part by the status of clearly established
norms. These norms help prevent certain forms of economic injury to
copyright owners, like the stealing of books or CDs from stores. Studies
show that people are generally untroubled by the non-commercial home
copying of copyrighted content.11
P2P applications are designed to look
10
Reinier H. Kraakman, Gatekeepers: The Anatomy of a Third-party Enforcement Strategy,
2 J. L. Econ. & Org. 53, 53­54 (1986).
11
See Amanda Lenhart et al., The Pew Internet & Am. Life Project, Downloading Free
Music: Internet music lovers don't think iťs stealing 5 (Sept. 28, 2000), available at
2003] Compliance & Code 109
and feel more like non-commercial home copying than like breaking
into a record store. The design therefore successfully exploits the normative
distinction between illegal "stealing" and innocuous "copying."12
Part III will demonstrate how P2P protocols have grown, through
several iterations, to specialize in exploiting copyrighťs gatekeeper system.
By its end, the P2P story suggests real limits on network design's
ability to influence law. Influencing the law in such a manner requires,
particular vulnerabilities in the law and a group that lacks better options.
The limits in generalizing the P2P model to other areas of law demonstrate
why the compliance challenge is specific to certain classes of vulnerable
laws, not a general challenge to the legal system.
Part IV will conclude by studying the fit between P2P applic ations
and music consumers as an interest group. A fascinating aspect of the
peer filesharing story is the lack of coordination and organization that
characterized its development. Developers bicker and work independently,
and etiquette among users must be engineered or, as Professor
Lior Strahilevitz argues, induced with "charismatic code."13
Despite the
chaos, peer networks have managed to provide a subset of music listeners
with a continuing reduction in the costs of copyright laws.
Such results from disorganized efforts are consistent with the distinction
between mechanisms of avoidance and of change. The disorganization
supports the claim that matters as an option for groups whose inability
to act collectively precludes better options.
Finally, the results may also reflect the current ability of P2P users to
take advantage of the continued compliance of the majority of the population.
The copyright regime's subjects are divided by a technological
line between the computer-savvy and regular users. Because consumption
of copyrighted works is non-rivalrous, P2P users may rely on regular
users to pay for music and to provide incentives for its creation, freeriding
on the results.
http://www.pewinternet.org/reports/toc.asp?Report=23 (on file with the Virginia Law Review
Association); Office of Tech. Assessment, U.S. Cong., Copyright and Home Copying:
Technology Challenges the Law 163 (Oct. 1989), at
http://www.wws.princeton.edu/~ota/disk1/1989/8910_n.html (on file with the Virginia Law
Review Association).
12
Cf. Lior Jacob Strahilevitz, Charismatic Code, Social Norms and the Emergence of Cooperation
on the File-Swapping Networks, 89 Va. L. Rev. 505 (2003) (arguing that charismatic
code creates an illusion of reciprocity that accounts for why people contribute to a
filesharing network).
13
See generally, id.
110 Virginia Law Review Vol. 89:nnn
Analyzing code design as a mechanism of interest group behavior
yields a nuanced picture. It departs from the grandiose predictions that
dominate discussion in this area. As with the onset of lobbying, impact
litigation, or sophisticated tax evasion, the rise of copyright evasion is
best understood as a change in power dynamics among and within regulated
groups.
I. A THEORY OF CODE, CHANGE, AND AVOIDANCE
The design of anti-regulatory code is best analyzed as one of many
mechanisms that interest groups might use to influence the effects of
law. Implicit in this argument is a set of assumptions and arguments that
Part I seeks to clarify.
A. Reactions to Law in General Theories of Regulation
John Austin, lecturing on jurisprudence in the early 1800s, sought to
separate law, the "appropriate matter of jurisprudence," from morals, religious
scruples, and other distractions.14
Two hundred years later, positive
legal scholarship has come full circle. Rather than focusing on separating
law from norms or ethics, it has pushed instead toward
understanding law as part of more general theories of regulation.15
Led
by the law and society movement and Robert Ellickson's book, Order
Without Law, theorists routinely study the regulatory effects of law,
group rules, social norms, and even the regulatory potential of code.16
Such scholarship reflects an effort to understand all the "forces" of regulation
that might be acting on an individual, reasoning that understanding
the study of law alone gives an incomplete picture. Robert Ellickson
even gave the study of law in isolation a pejorative label: "legal cen-
tralis[m]."17
14
See John Austin, The Province of Jurisprudence Determined 26 (1832).
15
See, e.g., Robert C. Ellickson, Order Without Law: How Neighbors Settle Disputes
126­32 (1991) (describing five different sources of regulation); Lessig, supra note 3, at 86­
90 (describing four modalities of regulation: law, markets, norms, and architecture (code)).
The antecedents for such general theories are in related sociological efforts. See, e.g., Donald
Black, Toward a General Theory of Social Control xi (Donald Black ed., 1984) (collecting
articles). A survey of legal scholarship in this vein can be found in Lawrence Lessig, The
New Chicago School, 27 J. Legal Stud. 661 (1998).
16
Lessig, supra note 3, at 86­90.
17
Ellickson, supra note 15, at 4, 137­47. Oliver Williamson coined the phrase, see Oliver
E. Williamson, Credible Commitments: Using Hostages to Support Exchange, 83 Am. Econ.
Rev. 519, 520 (1983), although Ellickson popularized and expanded on the criticism.
2003] Compliance & Code 111
Based on this work, the Internet law writers of the 1990s added the
idea that the design of computer code could be understood as an alternative
means of regulation, leading to the catchphrase "Code is Law."18
The idea is that programmers make choices that constrain online capability,
and that such choices are regulatory in their effects. Lawrence
Lessig, for example, argued that the size and weight of office buildings
can be understood as a mechanism for preventing their theft, just like a
law against larceny.19
Similarly, he reasoned, code-based copyright protection
for programs that make software difficult to steal are a form of
regulation.20
The same goes for code-based content-filters that might it
make it easier, or harder, to reach an intended audience. The design of
filters is simply the code-based regulation of speech.21
But as the scope of regulatory scholarship increases, it becomes more
apparent that there is something lopsided to the effort. Current scholarship
pays great attention to the range of options available to regulators.
But how much attention is paid to the reactions of the regulated? The
spirit of positive scholarship is to leave no stone unturned in the assessment
of regulatory effect. Fidelity to that approach necessitates understanding
not only regulation options, but also how the regulated might
undermine or compromise a regulatory scheme. If the goal of positive
scholarship is to understand the net effect of the regulatory forces acting
on a body, the model is incomplete without incorporating the reaction to
those forces. But what form will such reactions take? And how effective
will they be?
Today, such questions are answered in different ways by different
bodies of scholarship. In general, one answer comes from the compliance
literature: Groups will avoid laws they find burdensome. Another
answer comes from writings in political choice: groups will act to
change disagreeable laws. This Part proposes to reconcile and unite
these divergent accounts of the behavior of the regulated by analyzing
the choice between avoidance and change.
18
See generally, Lessig, supra note 3.
19
Lessig, supra note 3, at 86­90.
20
Id.
21
See generally, Lawrence Lessig, What Things Regulate Speech: CDA 2.0 vs. Filtering,
38 Jurimetrics J. 629 (1998).
112 Virginia Law Review Vol. 89:nnn
B. When Groups Get Sick of Complying
What choices face an individual or group that decides to quit complying
with the law and to invest in some mechanism to change its effects?
This Section outlines the fundamental choice between efforts to change
and efforts to avoid laws.
First, a few assumptions should be made clear. Laws and other regulations
prevent groups from doing what they would otherwise want to
do. As Professor Tom Tyler puts it, "Laws are passed and enforced to
mandate behavior that people would prefer to avoid . . . . It is a basic
tenet of political theory that any society . . . fails to provide its citizens
with some thing they want and feel they deserve."22
A related assumption
is that the initial content of laws are exogenous, the result of an unspecified
political process.23
As a result, groups often face laws with
which they disagree and would prefer to not follow, either in individual
cases or as a general matter. In this model, compliance is driven by expected
costs (punishments) deriving from legal sanctions (other sources
are possible, but omitted for the present).24
Finally, a mechanism of legal
influence is anything that, for a given price, buys a decrease in the expected
punishment associated with violating a given law.
1. Avoidance Mechanisms
When and why do groups obey the law? Basic economic models of
compliance give a very simple answer: Laws are followed when the expected
costs of legal punishment exceed the expected benefits of the
banned behavior.25
The result is commendably simple, but, as theorists
point out, only because it does not accurately describe when subjects
obey the law. Two important sets of contributing factors are neglected.
The first is extra-legal forces, such as social norms, that might contribute
22
Tom R. Tyler, Why People Obey the Law 19­20 (1990).
23
The assumption that the content of laws is exogenous becomes difficult to maintain
when we consider changing laws as a mechanism of response. In a subsequent section, I
consider what happens when the assumption that laws are exogenous is relaxed. See infra
text accompanying notes 72­78.
24
Cf. Leo Katz, Ill-Gotten Gains: Evasion, Blackmail, Fraud and Kindred Puzzles of The
Law 17­30 (1996) (describing avoision of moral and ethical rules as comparable to avoision
of law). The concept of avoision is described more fully infra note 32.
25
Albeit with much built into each side of the equation. See, e.g., Richard Posner, Economic
Analysis of Law 242 (5th ed. 1998) ("The model can be very simple: A person commits
a crime because the expected benefits of the crime to him exceed the expected costs.").
2003] Compliance & Code 113
to compliance. The second is investments in mechanisms of avoidance,
or efforts that would lower the expected costs of the law, which might
lead to greater non-compliance.
Efforts to broaden the basic model have focused on the first point, focusing
on the role that social norms and other factors play in creating
compliance. Both theory and some empirical studies suggest that the
threat of legal punishments alone cannot and does not fully explain why
people obey or do not obey the law.26
Supplemental explanations tend to
rely either on normative theories or more advanced models of selfinterested
behavior. Some, like Professor Tyler, argue that normative
considerations are central to understanding the public's decision whether
to comply or not.27
Others, like Professor Eric Posner, model extra-legal
compliance as self-interested signaling.28
Still others have modeled it as
a part of self-interested models of group interaction following gametheoretic
models.29
This Section, however, focuses on a different criticism of the basic
economic model of compliance--that it fails to take into account investments
in efforts to avoid the law. As much as the regulative effect of
26
See, e.g., Tyler, supra note 22, at 22 ("[T]he legal system cannot function if it can influence
people only by manipulating rewards and costs."); Ellickson, supra note 15, at 137­47
(arguing that law's punishments only explain some of the social order we see); Eric A. Posner,
Law and Social Norms: The Case Of Tax Compliance, 86 Va. L. Rev. 1781, 1782
(2000) (observing that state punishment cannot explain tax compliance); Paul G. Mahoney &
Chris William Sanchirico, Norms, Repeated Games, and the Role of Law 41­48 (2002) (unpublished
manuscript, on file with the Virginia Law Review Association), available at
http://papers.ssrn.com/sol3/papers.cfm?
abstract_id=311879 (suggesting that state punishment of deviants supports social orders otherwise
maintained by group sanctions). With mixed answers, some of the empirical studies
of the relationship between legal threats and compliance include Isaac Ehrlich, Crime, Punishment,
and the Market for Offenses, 10 J. Econ. Persp. 43 (1996) (surveying empirical
work in this area); Daniel S. Nagin & Raymond Paternoster, The Preventive Effects of the
Perceived Risk of Arrest: Testing and Expanded Conception of Deterrence, 29 Criminology
561, 580­81 (1991) (arguing that certainty of punishment plays a clear but minor role in determing
compliance); Raymond Paternoster, The Deterrent Effect of Perceived Certainty and
Severity of Punishment: A Review of the Evidence and Issues, 4 Just. Q. 173 (1987) (suggesting
weak correlation between perceived certainty of detection and drug use).
27
See E. Allan Lind & Tom R. Tyler, The Social Psychology of Procedural Justice 230­31
(1988) (developing a group value model to explain compliance); see generally, Tyler, supra
note 22 (arguing that perception of legitimacy affects the decision to comply).
28
Eric A. Posner, Law and Social Norms 88­111 (2000). Posner points out that the normative
and self-interested models of compliance can be unified by recognizing that effective
signaling depends on laws being considered legitimate. See id. at 111.
29
See, e.g., Ellickson, supra note 15, at 137­47 (presenting a model premised on iterated
prisoner's dilemma); Mahoney & Sanchirico, supra note 26 (same).
114 Virginia Law Review Vol. 89:nnn
social norms may create more compliance than the basic model predicts,
investments in efforts to decrease or eliminate punishments may result in
less compliance than predicted. In their classic article, Law Enforcement,
Malfeasance and Compensation of Enforcers, Professors Gary Becker
and George Stigler first argued that investments in avoidance should be
considerations of compliance.30
They added investments in bribery or
intimidation to a model of criminal behavior, pointing out that if a person
had already violated the law, she would be willing to invest up to the
costs of the sanction to avoid punishment.31
This insight suggests a very
basic point: compliance is not simply a function of punishments, but
also of the cost of mechanisms to avoid punis hment.32
The compliance literature surrounding particular statutory regimes
gives more particularized insight into how groups avoid laws. Avoidance
of laws is a particular focus of writings on tax compliance,33
and is
also the subject of study in labor law,34
criminal law,35
environmental
law,36
and international law.37
From these areas, a pattern emerges, indi-
30
See Gary S. Becker & George J. Stigler, Law Enforcement, Malfeasance and Compensation
of Enforcers, 3 J. Legal Stud. 1 (1974).
31
Id. at 2­6. The observation was a short stop enroute to their proposal for private enforcement
of criminal law, and the debate over their paper has focused on the merits of private
and public law enforcement. See, e.g., Mark A. Cohen & Paul H. Rubin, Private Enforcement
of Public Policy, 3 Yale J. on Reg. 167 (1985) (arguing for shifting responsibility
for implementing and enforcing public policy to private enforcement agents); William M.
Landes & Richard A. Posner, The Private Enforcement of Law, 4 J. Legal Stud. 1 (1975)
(responding to Becker and Stigler's proposal to privatize criminal law enforcement).
32
This insight is described in greater depth at infra text accompanying notes 66­70.
33
See, e.g., Marsha Blumenthal et al., Do Normative Appeals Affect Tax Compliance?
Evidence from a Controlled Experiment in Minnesota, 54 Naťl Tax J. 125 (2001) (concluding
from a tax compliance study that normatively appealing to a taxpayer's conscience via a
letter had an insignificant overall impact on tax compliance); Michael J. Graetz et al., The
Tax Compliance Game: Toward an Interactive Theory of Law Enforcement, 2 J.L. Econ. &
Org. 1 (1986) (modeling tax compliance as a game); Posner, supra note 26, at 1782 (proposing
a signaling model rather than the standard state sanctioning model to explain tax compliance);
David A. Weisbach, Formalism in the Tax Law, 66 U. Chi. L. Rev. 860, 884 (1999)
(arguing that anti-abuse standards would be more efficient than rules aimed at curbing tax
avoidance).
34
See Ronald Turner, Reactions of the Regulated: A Federal Labor Law Example, 17 Lab.
Law. 479 (2002) (detailing ways in which groups practice "avoision" of labor laws).
35
See, e.g., Neal Kumar Katyal, Deterrence's Difficulty, 95 Mich. L. Rev. 2385, 2414­15
(1997) (noting that a deterrence model in criminal law should focus on the role of substitute
products and complements to banned products and behavior).
36
See, e.g., Keith N. Hylton, When Should We Prefer Tort Law to Environmental Regulation?,
41 Washburn L.J. 515 (2002) (comparing the benefits of using tort law as a system of
2003] Compliance & Code 115
cating that there are two fundamentally different ways to avoid a law's
sanctions. The first can be termed evasion. Evasion can be defined as an
investment in decreasing the odds of being punished for violating a law.
Wearing a mask to rob a bank, buying a radar detector, hiring expensive
defense lawyers, and bribing police officers are all examples.38
Each, for
a certain price, decreases the odds of being punished after the law is dis-
obeyed.
There exists a second, less obvious way to avoid legal punis hment.
This is what Professor Leo Katz calls "avoision," which can be defined
as efforts to exploit the differences between a law's goals and its self defined
limits. As Professor Ronald Turner describes it, avoision represents
"efforts to change legal mandates or the avoidance of laws in ways
that evade the law's intent or purpose but do not actually constitute
unlawful behavior."39
Consider the example of the pornographer who,
worried about running afoul of decency laws, puts his photos in a book
along with incisive essays on "sex in marriage." Or consider the taxpayer
who, blocked from deducting a transfer of money to her son, devises
a complicated loan scheme to achieve the same effect. Professor
Leo Katz's book on avoision is full of such examples from law and other
aspects of life.40
One may identify a similar dynamic in Professor Neal
Katyaľs study of the role of substitute products in criminal deterrence.41
If, for example, the goal of the drug laws is to prevent addiction and
abuse, a person who opts to become an alcoholic (legal) instead of a
crack addict (illegal) is practicing avoision.
These writings paint the following picture: Groups, to minimize the
burdens of laws, will sometimes invest in avoidance. If the price is right
privately enforced environmental protection to traditional public statute-based regulatory
schemes).
37
Compliance in international law is studied in the absence of a centralized enforcement
system, creating concerns more akin to the study of compliance with social norms. See, e.g.,
Abram Chayes & Antonia Chayes, The New Sovereignty: Compliance with International
Regulatory Agreements (1995) (studying compliance with treaties); Jack L. Goldsmith &
Eric A. Posner, A Theory of Customary International Law, 66 U. Chi. L. Rev. 1113 (1999)
(studying compliance with customary international law).
38
Some of these are ex post examples, others are ex ante. For present purposes they are
considered together.
39
See Turner, supra note 34, at 479.
40
See generally, Katz, supra note 24 (presenting examples of avoision).
41
See generally, Katyal, supra note 35 (studying role of substitute products in models of
criminal deterrence).
116 Virginia Law Review Vol. 89:nnn
(more on this later),42
they will invest in mechanisms to lower or eliminate
the probability of being punished for disregarding a law. Groups
may either decrease the probability of detection (as in Becker's example
of a bribe) or adopt other forms of conduct with the same effects (as in
Katyaľs substitution effect, or Katz's avoision).
This might seem to deliver a full picture of how groups react to laws.
But even at this level of generality, writings on compliance still deliver a
limited picture of how individuals or groups might try to defeat a regulatory
scheme. For, as the political choice literature teaches, groups also
react to burdensome laws with efforts to change the law. The next section
considers change mechanisms as an alternative.
2. Change Mechanisms
In the early 1990s, the dietary supplement industry faced a serious legal
threat. Following several well-publicized deaths, the Food and Drug
Administration ("FDA") proposed to regulate popular dietary supplements
like other drugs, requiring proof of therapeutic value and carefully
determined dosages.43
The reaction of the supplement industry was to
invest in an expensive but successful lobbying campaign to change the
law. Within a short time, Congress had passed legislation limiting the
FDA's authority to regulate these products.44
It is by now a familiar insight
from public choice theory that groups that find a law disagreeable
may try to change it.45
In the 1970s, a series of articles written by
economists George J. Stigler and Sam Peltzman,46
followed by Robert E.
McCormick and Robert D. Tollison's book, Politicians, Legislation, and
the Economy,47
first modeled legislation as wealth transfers that interest
groups purchased with money and votes. As Peltzman put the basic
42
The effect of prices of mechanisms is discussed in infra text accompanying notes 66­70.
43
See Regulation of Dietary Supplements, 58 Fed. Reg. 33,690 (proposed June 18, 1993).
44
See Dietary Supplement Health and Education Act of 1994, Pub. L. No. 103-417, 108
Stat. 4325, 4326 (1994). This Act amended the Federal Food, Drug and Cosmetic Act
("FDCA") classifying dietary supplements as a new category of food, thereby preventing the
FDA from regulating supplements as drugs or food additives.
45
For a summary of work in this area, see Dennis C. Mueller, Public Choice II (1989),
particularly chapters 13 and 16.
46
See Sam Peltzman, Toward a More General Theory of Regulation, 19 J.L. & Econ. 211
(1976); George J. Stigler, The Size of Legislatures, 5 J. Legal Stud. 17 (1976); George J.
Stigler, The Theory of Economic Regulation, 2 Bell J. of Econ. & Mgmt. Sci. 3 (1971) (presenting
a general interest group theory of politics).
47
Robert E. McCormick & Robert D. Tollison, Politicians, Legislation and the Economy
(1981).
2003] Compliance & Code 117
premise: "I begin with the presumption that what is basically at stake in
regulatory processes is a transfer of wealth . . . . [B]eneficiaries [of
wealth transfers] pay with both votes and dollars."48
Or, as Professors
Richard Posner and William Landes described the legislative process,
laws are sold for "campaign contributions, votes, implicit promises of
future favors, and sometimes outright bribes."49
The basic model treats legislative change as a commodity available
for purchase. Since the introduction of the model, the literature studying
the specific mechanics of interest groups and lawmaking has become
more sophisticated. Professor Fred McChesney, for example, proposes
that law-makers are more extortionists than bribees.50
He highlights lobbying's
defensive aspects (Congress threatening legislation that groups
pay to avoid), and concludes that much of the political process can be
better described as rent-extraction instead of rent-creation.51
A series of
papers in the economics literature, meanwhile, tries to improve on the
simple bribery model with information theory, asserting that lobbying
works through the selective presentation of information.52
Despite these
refinements, however, lobbying continues to be studied as a change
mechanism--a tool that delivers or prevents legal change for a price.
The process of achieving legal change through litigation has also,
though less often, been studied as an investment model. In Professors
Landes's and Posner's first analysis of the independent judiciary, litigation
served as a means of extending the value of the legislative bargains
made between interest groups and the legislators.53
Professor Jeremy
Rabkin, in a 1989 work, broadly argued that, through their litigation
48
Peltzman, supra note 46, at 213­14.
49
William M. Landes & Richard A. Posner, The Independent Judiciary in an InterestGroup
Perspective, 18 J.L. & Econ. 875, 877 (1975).
50
See Fred S. McChesney, Rent Extraction and Rent Creation in the Economic Theory of
Regulation, 16 J. Legal Stud. 101 (1987) [hereinafter McChesney, Rent Extraction]; see also
Fred S. McChesney, Money for Nothing (1997) (developing and broadening the rent extraction
model) [hereinafter McChesney, Money for Nothing].
51
McChesney, Rent Extraction, supra note 50, at 109­12.
52
See, e.g., David Austen-Smith & John R. Wright, Competitive lobbying for a legislators'
vote, 9 Soc. Choice & Welfare 229 (1992) (developing a model of interest group behavior
based on the notion that such lobbying is the exercise of strategic information transmission);
Johan Lagerlof, Lobbying, information and private and social welfare, 13 Eur. J.
Pol. Econ. 615 (1997) (same); Susanne Lohmann, Information, access, and contributions: A
signaling model of lobbying, 85 Pub. Choice 267 (1995) (same).
53
See Landes & Posner, supra note 49; Posner, supra note 25, at 587­90.
118 Virginia Law Review Vol. 89:nnn
strategies, interest groups determine or radically influence the regulatory
agendas of agencies.54
In a 1991 essay Einer Elhauge argued that the litigation process was
equally, if not more, susceptible to interest group influence.55
He argued
that, generally speaking, "the same interest groups that have an organizational
advantage in collecting resources to influence legislators and
agencies also have an organizational advantage in collecting resources to
influence the courts."56
Therefore, "[I]ncreasing the lawmaking power of
the courts may only exacerbate the influence of interest groups."57
Whether Elhauge's specific conclusion is right or wrong, he demonstrates
that litigation campaigns can also be interpreted as investments in
legal change.
This literature shows that mechanisms of change can be viewed as an
alternative to mechanisms of evasion for lowering the costs of law.
3. Summary: The Change/Avoidance Choice
This Part has suggested that groups and individuals face a choice between
avoidance and change mechanisms when deciding how to react to
burdensome laws. Very simply, if a law is a cost on its subject, then
avoidance and change mechanisms, the subjects of the compliance and
political choice literatures respectively, can be pictured as different directions
of reaction, as follows.
54
Jeremy Rabkin, Judicial Compulsions: How Public Law Distorts Public Policy (1989).
55
See Einer R. Elhauge, Does Interest Group Theory Justify More Instrusive Judicial Review?,
101 Yale L.J. 31 (1991).
56
Id. at 67­68.
57
Id. at 68.
Law
Subject
Law
Change Avoid
2003] Compliance & Code 119
As identified in the discussion above, within each broader category of
mechanism, are specific subcategories, such as lobbying or litigation in
the case of change mechanisms, and evasion and avoision in the case of
avoidance mechanisms. Finally, while the model here focuses on law as
the regulatory modality, the basic framework of analysis is meant for
any source of regulation.
The following table summarizes the signal features of avoidance and
change mechanisms.58
Table 1:Change & Avoidance
Change Avoidance
Types Litigation, Lobbying Evasion, Avoision
Literature Public Choice Compliance
58
Some people may feel discomfort at comparing change and avoidance in this fashion,
but this discomfort may be useful and instructive. One can draw a parallel to Albert Hirschman's
work on institutional feedback. See Exit, Voice, and Loyalty, Albert Hirschman
(1970). Hirschman emphasized that members of declining institutions faced a fundamental
choice between "voice" and "exit" as forms of feedback. Despite the different "feel" of voice
and exit--study by different fields of scholarship, and the sense of disloyalty evident in the
latter--Hirschman maintained that a useful picture of organizational feedback required understanding
the choice. This Part suggests that focusing on the choice between avoidance
and change for groups faced with burdensome laws will yield similar dividends. As with
voice and exit, we want to know the conditions under which each option will prevail, and
each strategy's comparative efficiency. And if tools of avoidance are growing in sophistication,
as the example of code design here studied suggests, it makes sense to understand what
the consequences will be.
120 Virginia Law Review Vol. 89:nnn
Nature of Good Collective Good Excludable Good
C. Group Dynamics, Collective Action
The distinction between a group's choice of a change or avoidance
strategy is fundamental to understanding how groups deal with laws they
do not like. This Section links that choice to questions of group dynamics
and problems of collective action.
In 1964, Professor Mancur Olson made a well-known contribution to
the study of interest group behavior.59
Using the logic of collective action,
he divided those affected by regulation into two main groups--
those capable of effective political action, and the "forgotten groups"
who, he argued, "suffer in silence."60
The dividing line lay in the ability
to overcome collective action problems. Olson asserted that effective political
action would generally represent a problem of collective action,
making small groups and those organized for some other purpose (like
unions) effective political actors and rendering large and disorganized
groups essentially victims of the legislative process.61
His model predicted
that lobbies representing business, labor, agriculture, and professionals
would enjoy a perpetual advantage, leaving consumers and other
latent groups forgotten and even oppressed.62
The change/avoidance dichotomy suggests a different conclusion.
Forgotten groups do not necessarily suffer in silence; instead, they avoid
laws with which they disagree, so long as doing so is convenient. In the
terms used here, the groups Olson identified as incapable of collective
action will generally lack the capacity to invest in change mechanisms.
But that does not necessarily make them inert when faced with burdensome
laws. Rather, their recourse is limited to investing in avoidance
mechanisms to decrease the costs of laws. One may better understand
59
See Mancur Olson Jr., The Logic of Collective Action (1965). The logic of collective
action and the problem of free-riding now underlie most present-day studies of lobbying and
interest group behavior. See, e.g., Daniel A. Farber & Philip P. Frickey, Law and Public
Choice: A Critical Introduction 17­21 (1991); McCormick & Tollison, supra note 47, at 17­
18 (discussing organizing costs); Glynn S. Lunney, Jr., A Critical Reexamination of the Takings
Jurisprudence, 90 Mich. L. Rev. 1892, 1949­52 (1992) (summarizing the lobbying advantages
available to a small interest group).
60
Olson, supra note 59, at 165.
61
Id. at 53­57, 132­34, 165­67.
62
Id. at 133­67.
2003] Compliance & Code 121
Olson's dichotomy between groups as an indication of who can take advantage
of change mechanisms.
This follows because change presents a collective action problem,
while avoidance does not. Changes in laws display the classic attributes
of public goods. The repeal of the prohibition on alcohol in the Eighteenth
Amendment,63
for example, benefited all drinkers, not just those
who contributed to the effort to repeal it.64
Nor was there any possibility
that the repeal would be consumed or dissipated by overuse. As a result,
economic theory predicts a free-riding or collective action problem: The
beneficiaries of the change will wait for others to invest in it and will
subsequently free-ride on those efforts.
None of this is true of avoidance mechanisms. When a thief wears a
mask to rob a bank, he is the sole and direct beneficiary of his investment.
The driver using a radar detector keeps the benefits for herself.
When a firm invests in a complicated tax avoidance scheme, its competitors
do not benefit. In other words, investments in avoidance mechanisms
create excludable, rivalrous goods. In general, avoidance mechanisms
will side-step the problems of collective action inherent in change
mechanisms.65
63
See U.S. Const. Amend. XXI, § 1.
64
The repeal also cannot be "used up" by overconsumption. Legal change is an example
of what economists call a public or collective good: It is both non-rival and non-excludable.
See Olson, supra note 57, at 14 (defining public good).
65
A clever observer might object that this collective/private good distinction seems to blur
on further inspection. Yes, a tax avoidance scheme delivers benefits for the schemer, but if
successful, it may serve as a useful model for others. A police officer, once bribed, might be
easier to bribe in the future. The cars behind the driver with the radar detector might guess
why she brakes suddenly. So does avoidance really present a different kind of collective action
problem than investments in change?
It does, I suggest, because all of the examples posited simply represent the consumption of
a private good that creates a positive externality. This distinction can be illustrated by the
"popcorn/incense" example. Consider that cooking and eating popcorn will create a delicious
fragrance from which others cannot be excluded. That fact does not make the popcorn itself
a public good. The buyer of the popcorn reaps the reward of her investment, while also conferring
a benefit on her peers. Hence, diffuse unorganized groups should be expected to eat
popcorn despite the collective benefit conferred.
Replace popcorn with bribing a police officer and the same results are obtained. The briber
personally reaps the benefits of the bribe in an fashion excludable and rivalrous, but she also
confers a benefit on all future bribers.
Conversely, in the domain of public fragrance, the appropriate parallel to a change mechanism
like lobbying is the burning of incense. It costs money to burn incense so as to produce
a pleasant fragrance for the benefit of all. Hence, unlike popcorn, only organized groups will
burn incense, just as only organized groups will invest in lobbying campaigns.
122 Virginia Law Review Vol. 89:nnn
As a consequence it behooves third parties to sell avoidance to diffuse
groups. A third party can invent a mechanism for reducing the costs of a
given law, and then sell it to members of a diffuse group for profit or
fame. This is what happens when drivers buy radar detectors, companies
hire tax lawyers, or when music listeners download file-sharing software.
A legal entrepreneur invests in creating a means of lowering the
costs of law, and then sells it to groups that would otherwise comply.
A final complication with respect to avoidance and internal group dynamics
is worth stating. This Part, for simplicity's sake, has modeled all
laws simply as a cost to a regulated group, from which they derive no
benefit. But many laws provide both benefits and costs, and this fact
makes a difference for understanding the attraction of avoidance mechanisms.
Avoidance mechanisms can be used to lower the cost of a given
legal regime, while continuing to enjoy the benefits, through the rational
exploitation of the compliance of the rest of the regulated group. The
successful bank robber wants to steal money, but also wants to benefit
from a healthy financial system. Tax dodgers want to avoid paying taxes
while ideally enjoying public services paid for by everyone else. And, as
the P2P filesharing case study explores in greater depth, getting music
for free probably works best when most of the population continues to
pay retail.
D. Deciding to Quit
Groups do not spend all their time avoiding laws or trying to change
them; most people comply with most laws most of the time. When do
individuals or groups decide to quit obeying the law and instead invest
in some way to way to avoid or change it? The basic deterrence model
discussed above suggests that this happens when the cost of compliance
exceeds the expected cost of punishment. Theorists supplement that
model by accounting for compliance stemming from costs associated
with social norms and other sources. One may derive a more complete
answer by introducing the option of investing in mechanisms to decrease
legal66
or other costs. The following discussion will show two things.
First, compliance can be understood to depend less on punishment than
on the cost of mechanisms of change or avoidance. Second, this discus-
66
A caveat is necessary. At this stage, the model that follows is admittedly legally-centrist.
For simplicity's sake, it does not include the compliance produced by norms or other modalities
of regulation.
2003] Compliance & Code 123
sion will demonstrate the effect of a group's ability to act collectively,
pooling resources to invest in legal change.
First, examine a basic case where groups obey the law when expected
costs of disobedience exceed expected benefits and where there are no
mechanisms to influence the law. If a traffic law mandates a fifty-five
mile-per-hour speed limit, the expected benefit of ignoring the law and
driving eighty miles-per-hour might be $50, while the expected cost will
be the price of the speeding ticket multiplied by the chance of getting
caught (say, 20% x $500 = $100). With these parameters the driver will
not speed. The result is compliance and the law is a "success."
Now, allow for the option of investing in a mechanism that influences
the expected costs of the law. As discussed above, Becker and Stigler's
original example was the bribe; for a certain fee, a bribe reduces the expected
costs of a law to zero (by eliminating any chance of detection).67
There are, however, a wide variety of mechanisms beyond bribes that
will accomplish the same effect. For the driver, there exists a strategy of
avoidance and one of change: investing in a radar detector and lobbying
to repeal the speeding law, respectively.
Individuals and groups will invest in a mechanism of legal influence
when it becomes cheaper to do so than to simply comply with the law.
Entities will invest in such mechanisms when the expected benefits exceed
the sum of the response strategy cost and the expected costs of noncompliance
(as reduced through the mechanism).
One may describe this dynamic with a very simple equation. Groups
that have the option of purchasing mechanisms of legal influence will do
so when:
Expected Benefits > (Expected Costs ­ Mechanism Effect) +
67
Becker & Stigler, supra note 30, at 5­6.
Compliance With No Investments in Response
Assume:
(1) Speed limit = 55 mph
(2) Benefit of driving 80 mph = $50
(3) Expected Costs = (Sanction)*(Probability of Detection) = $500*0.2 =
$100
Result: Driver complies with law, because expected costs > expected
benefits
124 Virginia Law Review Vol. 89:nnn
Cost of Mechanism
Apply this framework to two of the preceding examples: radar detectors
and lobbying. First, consider a $40 radar detector that eliminates
any chance of being caught speeding. For the driver discussed above,
this is a worthwhile investment. For the price of the radar detector ($40),
he gets to drive at eighty miles-per-hour (benefit $50) and is therefore
$10 ahead. The driver is pleased, but the regulator is not; the law that
was once a "success" is now a "failure."68
This example demonstrates that if the mechanism of legal influence is
100% effective, like our radar detector, the expected cost of legal sanctions
is reduced to zero, and thus can be eliminated from the basic investment
equation. Therefore, given perfectly effective mechanisms, the
only relevant inputs are the expected benefits and the cost of the response
strategy, and the equation can be simplified as follows:
Investment when:
Expected Benefit > Cost of Mechanism
In other words, in a world where avoidance or change is entirely effective,
compliance with current law has little to do with punis hment,
but is instead a direct function of how much it costs to buy a way out.
Consider a few implications of this analysis. The first example involves
an avoidance strategy. If the speed limit were one hundred milesper-hour,
and hence not much of a burden, few individuals would buy
the perfect $40 radar detector. Conversely, if the speed limit were lowered
to ten miles-per-hour, an onerous burden, everyone would want a
68
Notice that for simplicity's sake, this hypothetical has neglected the governmenťs response:
government can, as some states do, ban the radar detector (but more on this later).
Assume:
(4) Probability of detection = 0%
(5) Expected benefit of speeding = $50
(6) Expected cost of radar detector = $40
(7) Expected cost of legal punishment = $500 x 0 = 0
$50 > 0 + $40
Result: Driver disregards law
2003] Compliance & Code 125
perfect radar detector, even if it cost $500. Finally, notice that if the
price of the perfect radar detector suddenly falls to $1, it may become a
worthwhile investment, even for the nearly costless one hundred mileper-hour
speed limit.69
The second example involves a lobbying campaign. Assume it would
cost $100,000 to organize a campaign to repeal the speeding laws. For
the individual driver, the lobbying campaign is not a worthwhile purchase.
The benefit of driving at eighty miles-per-hour is only $50. The
cost of the campaign would leave the driver $99,950 in the red, unless
he were somehow able to charge his fellow drivers for the successful repeal,
an unlikely prospect.
Would it make sense for the affected group (all drivers) to invest in a
campaign to repeal the speeding laws? Assume that there are 100,000
drivers in the lawmaking jurisdiction (a state). If the drivers organize
themselves so as to divide the costs of the repeal campaign, they pay $1
each, and such an arrangement is clearly a good deal for all involved.
Stated otherwise, the cost of compliance for the group is $50 times
69
The benefits of not having to comply with the law must be greater than the cost of a response
strategy for any investment to happen at all. This necessity is a consequence of perspective;
the assumption is that the individual is complying with the law, and deciding
whether to invest in some way to make it worthwhile not to comply. In contrast, Becker and
Stigler's original model posited a criminal already in violation of the law, and suggested that
"[t]he violator would be willing to bribe as much as [the fine] to ignore the evidence."
Becker & Stigler, supra note 30, at 5. While the behavior of violators is of interest, it generally
seems more interesting to understand what individuals already regulated by the law will
do, instead of assuming that they will break it.
Compliance Given a $100,000 Lobbying Campaign
Assume:
(8) Repeal makes expected costs = $0
(9) Expected benefit of speeding = $50
(10) Expected cost of campaign = $100,000
$50 < $100,000
Result: Driver complies with law.
Compliance Given a $100,000 Lobbying Campaign (2)
Assume:
(11) Repeal makes expected costs = $0
(12) Group expected benefit of speeding = $5 million
(13) Expected cost of campaign = $100,000
$5,000,000 > $100,000
Result: Group repeals speeding laws.
126 Virginia Law Review Vol. 89:nnn
100,000 drivers, or $5 million. The lobbying campaign is, therefore, a
bargain.
If these numbers are even close to realistic, then why are there traffic
laws or any other laws that large groups find disagreeable? As already
demonstrated and as basic political choice theory teaches, the answer is
that groups such as drivers are not organized and have no effective
mechanism to divide the costs of a campaign to change the law.70
This
demonstrates the conclusion urged above: Groups incapable of collective
action tend toward avoidance mechanisms, while the organized invest
in mechanisms of change.
E. Avoidance, Change, and Regulatory Competition
This Article has until now focused on first-generation reactions--
those of an interest group to a disagreeable law. This Section adds "reactions
to the reaction" to show how regulatory competition between two
opposing groups develops, with each investing in efforts to influence the
law in its favor. For this analysis, the model of rent-seeking competitions
is a useful descriptive, though not necessarily normative, guide.
The model has thus far treated laws exclusively as exogenously imposed
costs on regulated groups. A more realistic model recognizes that
the content of laws is a function of group interests, so that for every
regulated group there exists a beneficiary group.71
For example, if a law
bans noisy sound trucks then the law regulates advertisers in the interest
of town residents.72
Successful efforts to avoid or change the law may,
therefore, inspire the beneficiary group to invest in its own mechanism
of legal influence in an effort to restore the lost benefit. This investment,
in turn, may inspire the regulated group to reinvest in mechanisms of influence,
leading to a full-fledged cycle of regulatory competition. The
cycle continues as long as each group values sufficiently the prize of a
law tailored in its favor.
Just as group identity and dynamics influenced the actions of the
regulated group, we should expect them to do the same for the benefic iary
group. An organized, politically effective beneficiary group faced
70
See discussion of group dynamics, supra text accompanying notes 59­64.
71
Cf. McCormick & Tollison, supra note 47 (modeling groups in competition for legislative
wealth transfers).
72
See Kovacs v. Cooper, 336 U.S. 77 (1949).
2003] Compliance & Code 127
with evasion may turn to the legislature with a request to "restore the
balance." On the other hand, diffuse benefic iaries may do little to react
effectively.
Consider the following contrast. The P2P story features a subset of
music consumers, in ferocious competition with the music industry, trying
to avoid copyright laws. Faced with a threat to their copyright rents,
the industry reacted with litigation, lobbying, and even technological
countermeasures (detailed in Part III). In contrast, avoidance of state
taxation through online and mail-order catalogues is now a regular phenomenon.
Yet the diffuse benefic iaries of state taxation have done little
to resist the eroding collection of state value-added taxes.73
Unsurprisingly,
the organization of the beneficiaries matters as much as the organization
of the regulated.
The notions of regulatory competitions are a favorite subject of the
rent-seeking literature, and it is tempting to cast matters in such terms.
Professor Anne Krueger's original description of rent-seeking suggested
that laws create rents and that people will compete for them in various
ways: "[s]ometimes, such competition is perfectly legal. In other instances,
rent seeking takes other forms, such as bribery, corruption,
smuggling, and black markets."74
Arguably, any group interested in
changing a law to minimize its regulatory costs is engaged in a form of
rent-seeking. The battle between P2P programmers and the recording
industry, described in Part III, can be described as a gigantic dissipation
of rents created by the monopolistic copyright system.
Groups reacting to law are acting in a self-interested fashion, and this
may lead to a competition to influence the law's effects. For several reasons,
however, I am hesitant to cast the questions studied in this Part
within the normative framework of rent-seeking. Rent-seeking is a useful
tool when it suggests that certain models of regulation will encourage
wasteful behavior and should therefore be avoided. In other words, the
study of rent-seeking is the study of waste management. The goals of
this Part, however, are different. They are to develop a positive model
describing the choices that groups face under burdensome regulation.
Determining whether the reduction in rents is "worth" any particular legal
regime is beyond this Parťs scope.
73
On the contrary, Congress passed the Internet Tax Freedom Act, 47 U.S.C. 1151 (1998),
restricting the power of states to tax Internet-based commerce .
74
Anne O. Krueger, The Political Economy of the Rent-Seeking Society, 64 Am. Econ.
Rev. 291, 291 (1974).
128 Virginia Law Review Vol. 89:nnn
In addition, the interests of the rent-seeking literature are different
than those of this Article. What makes a tool interesting to the rentseeking
literature is its potential for generating waste and the existence
or absence of any socially valuable byproduct. Hence, what scholars
study for rent-dissipating effects can range from research and development
(rent dissipation in pursuit of patent follow-ons)75
to follow-on
creation in copyright76
to efforts to monopolize.77
It is nonetheless extremely
difficult to evaluate whether alternative mechanisms of undermining
legal systems have less or more valuable byproducts.78
Is investing
in a tax shelter more or less socially wasteful than lobbying? Such
questions seem nearly impossible to answer. What this Part examines is
not the relative wastefulness of mechanisms used to influence law, but
their relative cost and relationship to group dynamics.
F. How Code is Used to Avoid Law
The premise of this Article is that "law-busting" code should be studied
as a mechanism of legal influence. That is to say, it can usefully be
studied alongside litigation, lobbying, tax avoision, and other ways
groups seek to influence the law in their favor. This final Section asks:
how exactly does code influence the effects of law? And how does it fit
within the avoidance/change dichotomy just described?
The hint of an answer comes from existing work that tries to understand
the role code plays in the legal environment.79
In Code and Other
75
See generally, Edmund W. Kitch, The Nature and Function of the Patent System, 20 J.L.
& Econ. 265 (1977) (describing patents as prospects that prevent waste in follow-on devel-
opment).
76
See generally, Michael Abramowicz, Copyright Redundancy (Geo. Mason L. & Econ.
Research Paper No. 03-03, 2003), available at http://papers.ssrn.com/sol3/
papers.cfm?abstract_id=374580 (on file with the Virginia Law Review Association) (arguing
that copyright laws prevent wasteful redundancy) .
77
See e.g., Richard Posner, The Social Costs of Monopoly, 83 J. Pol. Econ. 807 (1975)
(modeling and estimating social costs of monopoly and monopoly-inducing regulation in the
U.S.).
78
See id. at 811 (analyzing assumption that expenditures on monopolizing have no socially
beneficial byproduct).
79
A common question is this: Is there any particular significance to code in this regard, as
opposed to just advances in technology and their effects on compliance? The argument for a
special relevance for code relies on the idea that computer code has achieved a greater
granularity than the technologies that preceded it: Programmers can very precisely shape behavior
using code to match the particularized loopholes in laws. At previous levels of technology,
conversely, such questions would arise less frequently.
2003] Compliance & Code 129
Laws of Cyberspace, Professor Lawrence Lessig writes that "[i]n cyberspace
we must understand how code regulates . . . . Code is law."80
Similarly, writers like Professors Tom Bell or Kenneth Dam, interested
in "technological self-help," are primarily concerned with the use of
code as a substitute for contract, copyright, or other legal systems.81
Even though this work is interested in code "as law," its depiction of
how code achieves regulatory effects if useful. The idea is that code
regulates by directly constraining behavior. Lessig argues that code
"constitute[s] a set of constraints on how you behave;"82
it "constrains
some behavior by making other behavior possible, or impossible."83
Just
as a brick wall built in the middle of the road modifies behavior, code
regulates by specifiying, in advance, what behavior is and is not possible.
Similarly, I propose that code can influence the effects of law by redesigning
behavior for legal advantage. That is to say, the reason that
code matters for law at all is its capability to define behavior on a mass
scale. This capability can mean constraints on behavior, in which case
code regulates, but it can also mean shaping behavior into legally advantageous
forms.
In this view, the code designer acts like a tax lawyer. He looks for
loopholes or ambiguities in the operation of law (or, sometimes, ethics).
More precisely, he looks for places where the stated goals of the law are
different than its self-defined or practical limits. The designer then redesigns
behavior to exploit the legal weakness.
Code design, as we have seen it, is a mechanism of avoidance rather
than a mechanism of change. Nothing the code designer does rewrites
laws. Instead, code design defines behavior to avoid legal sanctions.
This description of how code "works" to influence law's effects, I suggest,
fits most of the major efforts to use code for legal advantage. Consider
four examples:
Virtual Child Pornography. Congress passes a law banning child pornography,
citing a compelling interest in preventing harm to children.
Programmers create child pornography that involve no children in its
80
Lessig, supra note 3, at 89.
81
See Bell, supra note 4 (model of technological self-help); Dam, supra note 4 (same).
82
Lessig, supra note 3, at 6, 89.
83
Id.
130 Virginia Law Review Vol. 89:nnn
production. The behavior has been reshaped to adapt to the limit on governmenťs
power in the First Amendment.84
Overseas Gambling. Laws banning gambling are territorial in jurisdiction.
Casinos place their servers overseas. The conduct of gambling
has been reshaped to avoid the law's self-defined jurisdictional limits.
Junk Email. Unsolicited advertising by mail and fax are regulated by
laws specific to the mail system and fax machine, respectively. Advertisers
design programs to transmit electronic mail and pop-up advertisements.
The use of junk email gives advertisers an unregulated partial
substitute for the mail or fax machines.
P2P Filesharing. The legality and ethics of "home copying" are
somewhat ambiguous, and copyright has no record of enforcement
against end-users. Designers build software that shapes the mass distribution
of copyrighted works into a form resembling home copying.
None of this, of course, is a comment on whether these strategies will
be successful in the long term. Each, as previous sections suggesst, may
incur a reaction to the reaction--an effort to change the law to "restore
the balance." But it is clear from these examples how code design
achieves its effects.
This basic theory of mechanisms underlies the claims in the rest of the
Article. Part II examines the important example of P2P filesharing to
show how, in practice, the design of code influences the effects of law.
II. COPYRIGHŤS LOOPHOLES
On December 8, 1999, a group of eighteen record companies announced
that they had sued a small startup company for copyright in-
fringement.85
The Recording Industry Association of America ("RIAA")
forecast that it could do 100 million dollars in damage to sales,86
yet the
company was virtually unknown. In the mainstream press the company
had previously drawn only a blurb, described by Fortune magazine as "a
unique online MP3 trading community . . . that enables users to trade
songs directly."87
84
Cf. Ashcroft v. Free Speech Coalition, 122 S. Ct. 1389 (2002) (striking down ban on
computer-generated child pornography).
85
See Don Clark, Recording Industry Group Sues Napster, Alleging Copyright Infringement
on Net, Wall St. J., Dec. 9, 1999, at B18.
86
Id.
87
Lauren Goldstein, Tune In: MP3 goes mainstream, but Internet music has yet to find its
perfect form, Tune In, Fortune, Dec. 1, 1999, at 268.
2003] Compliance & Code 131
This unknown company was Napster. Its product was an application
that facilitated the trading of music files. Napster functioned like a "bazaar,"
alleged the plaintiffs,88
but one where the goods were all free. Users
logged in, searched a central database of songs that other users had
made available, and then took the files they wanted directly from other
users.89
Lawyers for the recording industry accused the little company of
operating a "haven for music piracy on an unprecedented scale" and an
"online bazaar" for illegal trading.90
Napster responded that it simply
provided a "listing service."91
If not as ruinous as the recording industry suggested it would be,92
Napster emerged as a powerful force in the distribution of music. At its
height, Napster claimed sixty million registered users and as many as
twenty-six million active ones.93
By February of 2001, analysts estimated
that Napster users were trading nearly three billion songs, or the
equivalent of two hundred million CDs, in a single month.94
The economic
effects of Napster on the music industry were, naturally, disputed
in litigation.95
According to some figures, global music sales tumbled
88
Complaint at 2, A & M Records v. Napster, Inc., 114 F. Supp. 2d 896 (N.D. Cal. 2000)
(No. C99-5183-MHP), available at http://news.findlaw.com/cnn/docs/napster/
riaa/napster_complaint.pdf (on file with the Virginia Law Review Association) [hereinafter
Complaint].
89
See Damien A. Riehl, Peer-to-Peer Distribution Systems: Will Napster, Gnutella and
Freenet Create a Copyright Nirvana or Gehenna?, 27 Wm. Mitchell L. Rev. 1761, 1768
(2001).
90
Complaint, supra note 88, at 2.
91
Clark, supra note 85.
92
See Complaint, supra note 88, at 3 (alleging that "Napster's conduct has caused and continues
to cause plaintiffs grave and irreparable harm").
93
The estimates of Napster's use vary. See, e.g., Jon Healey, Napster CEO Pitching a New
Tune to Labels, L.A. Times, Nov. 25, 2001, at C5 (reporting sixty million active users at
Napster's peak); Napster Use Slumps 65%, BBC News, at
http://news.bbc.co.uk/2/hi/business/1449127.stm (July 20, 2001) (on file with the Virginia
Law Review Association) (reporting statistics from Jupiter Media Metrix stating that Napster
had 26.4 million active users in February 2001 before the numbers began to decline).
94
See Geoff Nicholson, Will the RIAA pass up Napster's $1 billion offer?, at
http://www.hitsquad.com/smm/news/708/ (Feb. 21, 2001) (on file with the Virginia Law Review
Association).
95
See A & M Records v. Napster, Inc., 114 F. Supp. 2d 896, 909­11 (N.D. Cal. 2000)
(summarizing the findings of several studies of Napster's economic impact). A later study by
economist Stan Leibowitz concludes that Napster's effects were not proven in the Napster
litigation, but that peer filesharing should be expected to hurt the music industry in the long
term. Stan Liebowitz, Policing Pirates in the Networked Age 14­15 (Cato Policy Analysis
No. 438 May 15, 2002), available at http://www.cato.org/pubs/pas/pa438.pdf (on file with
the Virginia Law Review Association).
132 Virginia Law Review Vol. 89:nnn
nearly half a billion dollars in 2000.96
Sales of CD singles (the clearest
Napster competitor) declined nearly forty percent that year.97
In contrast,
other studies suggested that Napster actually led its users to buy more
CDs.98
How did any of this happen? How did a simple program have such a
powerful effect on levels of compliance with copyright law? Everyone
knows the basic story, but students of enforcement and compliance lack
an explanation for why the copyright regime, relative to other sets of
laws, proved so vulnerable to code-based attack. What is it about the enforcement
structure of the copyright system that made it so easy to defeat?
And does it share characteristics with other legal enforcement sys-
tems?
This Part argues that the success of P2P depends on two powerful and
often unrecognized weaknesses of the copyright regime. The first is the
law's dependence on a gatekeeper enforcement regime. The second is a
severe and unusual lack of normative support among the regulated.
These weaknesses suggest several conclusions about the nature of
P2P and code design as mechanisms of avoidance. P2P, in particular,
probably implicates the specific weaknesses of the copyright system
more than it implicates vulnerabilities in other sets of legal rules. As a
general rule, code design will depend on identifiable weaknesses in legal
enforcement.
A. Copyright and Its Gatekeepers
Common intuition dictates that laws can be vulnerable to mass disobedience,
whether at rock concerts or during tax time. These problems
stem from the limits and costs of "primary" enforcement (enforcement
against individual violators). The costs of raising punishments increase
while the benefits exhibit diminishing returns. Theorists explain these
limits as stemming from administrative and third party costs, the limited
96
Patrick Brethour, Music sales tumble 1.3% worldwide, The Globe and Mail (Boston),
Apr. 20, 2001, at B1.
97
Jeff Leeds, Record Industry Says Napster Hurt Sales, L.A. Times, Feb. 24, 2001, at C1.
98
See Napster, 114 F. Supp. 2d at 910 (citing several studies but refusing to rely on them);
Kim Chipman, Napster More Likely to Help, Not Hurt, Music Sales, Bloomberg News, July
21, 2000 (noting that "most attrition [cited by the RIAA] took place before Napster's
launch"); Liam Lahey, Angus Reid Study: Napster is improving CD sales, ComputerWorld
Canada, Sept. 22, 2000, at 1, available at
http://www.itworldcanada.com/portals/portalDisplay.cfm?oid=E19EF5FC-8783-45AEAB14E3C8BA85856F
(on file with the Virginia Law Review Association).
2003] Compliance & Code 133
net worth of defendants, the lack of any punishment beyond the death
penalty, and even the constitutional prohibition on cruel and unusual
punishments.99
Due to the limitations of primary enforcement, many legal regimes
charged with mass regulation come to depend on supplemental enforcement
measures. A chief example is what Professor Kraakman termed a
"gatekeeper" regime.100
To supplement direct enforcement of a law, the
state attaches liability to the provision of specialized goods or services,
disrupting misconduct in advance.101
Doctors, for example, are gatekeepers
for prescription drugs. By withholding their provision of drugs
to would-be abusers, doctors aid in the enforcement of the laws regulating
controlled substances.
Copyright law's long dependence on a gatekeeping regime is under-
recognized.102
The copyright law regulates a large and disparate group of
content consumers, such as music listeners and book readers. The solu-
99
These reasons are summarized in Posner, supra note 25, at 243­50; Kraakman, supra
note 10, at 56­57. See also Katyal, supra note 35, at 2414­15 ("But the range of sanction
levels may be subject to a maximum sanction constraint--either because there is no room for
increased penalty (beyond death) or because such equality in punishment would contravene
other, moral, theories of punishment.").
100
Kraakman, supra note 10, at 53.
101
In his influential 1986 article, Kraakman demonstrated that "gatekeeper liability" could
create additional deterrence relative to primary enforcement. See id. at 87­93. That article
has inspired a gatekeeper literature, primarily focused on gatekeepers in the financial services
industries. See, e.g., Stephen Choi, Market Lessons for Gatekeepers, 92 Nw. U. L.
Rev. 916, 918 (1998) (arguing that analysis of reputational intermediaries remains incomplete
without consideration of a variety of additional factors); Luigi Alberto Franzoni, Independent
Auditors as Fiscal Gatekeepers, 18 Inťl Rev. L. & Econ. 365, 365 (1998) (analyzing
gatekeeper regimes in tax enforcement); Ronald J. Gilson, The Devolution of the Legal Profession:
A Demand Side Perspective, 49 Md. L. Rev. 869, 883­84 (1990) (analyzing the
gatekeeper role lawyers play in avoiding strategic litigation); Frank Partnoy, Barbarians At
The Gatekeepers?: A Proposal For A Modified Strict Liability Regime, 79 Wash. U. L.Q.
491, 491­93 (2001) (arguing for a strict liability gatekeeper regime for securities fraud).
None, however, considers a statute's dependence on gatekeeper liability to be a potential
weakness.
102
One notable exception to this generalization is Randal C. Picker, Copyright as Entry
Policy: The Case of Digital Distribution, 47 Antitrust Bull. 423, 432 (2002). A similar notion
is reflected in the distinction between "broad-based" and "targeted" enforcement in Rick
Harbaugh & Rahul Khemka, Does copyright enforcement encourage piracy? (Claremont
Colleges working paper in economics, Aug. 2001), available at
http://econ.mckenna.edu/papers/2000-14.pdf (on file with the Virginia Law Review Association).
One reason copyrighťs dependence on gatekeepers may be under-recognized is possibly
because most of copyright law is found under the civil, as opposed to the criminal titles
of the law. Yet there is no reason to suppose from first principles that a civil regime cannot
also harness the power of a gatekeeper liability regime.
134 Virginia Law Review Vol. 89:nnn
tion to mass disobedience in this area has involved one such gatekeeper
regime. That is, copyright law achieved compliance through the imposition
of liability on a limited number of intermediaries--those capable of
copying and distributing works on a mass scale. The gatekeepers were
book publishers at first; later gatekeepers included record manufacturers,
film studios, and others who produced works on a mass scale. Their role
resembled that of doctors with respect to prescription drugs--they prevented
evasion of the law by blocking the opportunity to buy an infringing
product in the first place.
That intermediaries play some role in copyright enforcement is
widely recognized103
--it could not be otherwise after the United States
Supreme Courťs decision in Sony Corp. of America v. Universal City
Studios.104
Writers have hinted at the potential dependence of copyright
on a gatekeeper system. As Professor Jane Ginsberg noted in 1995:
Copyright owners have traditionally avoided targeting end users of
copyrighted works. This is in part because pursuing the ultimate consumer
is costly and unpopular. But the primary reason has been because
end users did not copy works of authorship--or if they did copy,
the reproduction was insignificant and rarely the subject of widespread
further dissemination.105
There is evidence to suggest that copyright was in fact entirely dependent
on gatekeeper enforcement until quite recently. Unfortunately, academic
study of copyright enforcement is sparse.106
What we can learn
about enforcement patterns comes largely from the few hearings and
103
See, e.g., Jessica Litman, Digital Copyright 111 (2001) ("Our copyright laws have, until
now, focused primarily on the relationships among those who write works of authorship
and disseminate those works to the public."); Jane C. Ginsburg, Putting Cars On The "Information
Superhighway": Authors, Exploiters, and Copyright in Cyberspace, 95 Colum. L.
Rev. 1466, 1488 (1995) (discussing the role of intermediaries).
104
464 U.S. 417 (1984). In the Sony litigation, the broadcasting industry targeted Sony
and its new Betamax videotape recorder, as opposed to end-users, when it unsuccessfully
tried to have Sony held contributorily liable for any illegal taping of television shows. Id. at
456.
105
Ginsburg, supra note 103, at 1488.
106
While many authors discuss the challenge of new technology for intellectual property
laws, it is difficult to find academic work on actual patterns of enforcement. One student
note has tackled the problem, relying principally on congressional sources. See Jayashri Srikantiah,
The Response of Copyright to the Enforcement Strain of Inexpensive Copying
Technology, 71 N.Y.U. L. Rev. 1634, 1643­45 (1996).
2003] Compliance & Code 135
congressional studies on copyright enforcement and the case record it-
self.
Reflecting an interest in bigger targets, copyright laws reflected an
indifference to private, home copying in the 1960s and early 1970s. In
1971, Congress commented that copyright was never meant to "restrain
the home recording, from broadcasts or from tapes or records, of recorded
performances."107
Congress described the practice of noncommercial
home recordings as "common and unrestrained."108
In the
1973 photocopying case Williams & Wilkins Co. v. United States, the
United States Court of Claims similarly stated, "[I]t is almost unanimously
accepted that a scholar can make a handwritten copy of an entire
copyrighted article for his own use . . . . These customary facts of copyright-life
are among our givens."109
Even in the 1976 Copyright Act, Congress made the decision to limit
the exclusive right of performance of audiovisual works to public performances,
thereby excluding private or home performances.110
In recommending
this limit, the Copyright Office explained that "[n]ew technical
devices will probably make it practical in the future to reproduce
televised motion pictures in the home. We do not believe the private use
of such a reproduction can or should be precluded by copyright."111
The
law's indifference toward home copying was evident in the obvious lack
of enforcement. The case record is perhaps the strongest evidence of the
operation of the old regime. One is pressed to find any example of copyright
law being enforced against individuals for home copying (as opposed
to commercial activity) prior to 1990. In the 1979 Sony Betamax
case, copyright owners added a representative individual to the complaint,
but they did not seek relief against him.112
Beyond this limited
example, individualized infringement actions were absent until the
1990s.113
107
H.R. Rep. No. 92-487, at 7 (1971).
108
Id.
109
487 F.2d 1345, 1350 (Ct. Cl. 1973), aff'd by an equally divided Court, 420 U.S. 376
(1975).
110
17 U.S.C. § 106(4) (2000).
111
Register of Copyrights, 87th Cong., Report of the Register of Copyrights on the General
Revision of the U.S. Copyright Law 30 (Comm. Print 1961).
112
Universal City Studios v. Sony Corp. of Am., 480 F. Supp. 429, 432 (C.D. Cal. 1979);
see also Sony, 464 U.S. at 434 ("The two respondents in this case do not seek relief against
the Betamax users who have allegedly infringed their copyrights.").
113
The 1990s saw an effort by software copyright owners to enforce copyrights against
end-users, who tend to be fairly large entities. See, e.g., Elizabeth Hurt, Software Pirates
136 Virginia Law Review Vol. 89:nnn
The Supreme Courťs decision in Dowling v. United States and others
like it come closest to primary enforcement against individuals.114
Dowling featured two Elvis enthusiasts who pressed unreleased recordings
without permission--so-called "bootleggers."115
But these
bootleggers actually created sizable distribution channels. The two hobbyists
grew to do "substantial business,"116
eventually functioning just
like regular record-sellers themselves. They printed catalogs and advertisements,
and they sold and distributed thousands of albums.117
Were
these Elvis bootleggers gatekeepers in the enforcement sense? They
were, in the sense that the end-users of the Elvis bootlegs would be unable
to obtain their product without the cooperation of Dowling and
company.
Mass home copying became an issue in the late 1980s and prompted
some examination of how copyright enforcement worked.118
As the
Congressional Office of Technology Assessment stated in its 1989 report,
"All U.S. copyright law, including the Copyright Act of 1976, proceeds
on the assumption that effective and efficient copying is a largescale,
publicly visible, commercial activity, and therefore, that legal
prohibitions against unauthorized copying are enforceable."119
This report,
echoed by hearings on copyright enforcement in the 1980s, confirmed
that the existing pattern of enforcement by the RIAA and the motion
picture industry targetted large-scale commercial pirates.120
After
clarifying copyrighťs long reliance on a gatekeeper system, one may
Sued: Alleged culprits targeted online auction bidders, Business 2.0, (Jan. 26, 2001), at
http://www.business2.com/articles/web/print/0,1650,16147,00.html (on file with the Virginia
Law Review Association). For an argument that such enforcement actually creates more piracy,
see Harbaugh & Khemka, supra note 102, at 2.
114
473 U.S. 207 (1985). Other examples of enforcement against small intermediaries include
United States v. Drum, 733 F.2d 1503 (11th Cir. 1984) (enforcing against a bootle gging
enterprise), and Paramount Pictures Corp. v. Labus, No. 89-C-797-C, 1990 U.S. Dist.
LEXIS 11754 (W.D. Wis. Mar. 23, 1990) (involving the operator of a small resort sued for
renting pirated movies to his customers).
115
473 U.S. at 210­11.
116
Id. at 212.
117
Id. at 211­12.
118
The Office of Technology Assessment noted that the proportion of people who made
home audiotapes doubled in the 1980s. Office of Tech. Assessment, supra note 11, at iii.
119
Id. at 7.
120
See Civil and Criminal Enforcement of the Copyright Laws: Hearing Before the Subcomm.
on Patents, Copyrights and Trademarks of the Senate Comm. on the Judiciary, 99th
Cong. 41 (1985) (statement of Donald C. Curran, Acting Register of Copyrights) ("RIAA is
selective in what they refer to Justice, turning over only the most egregious cases.").
2003] Compliance & Code 137
specify more precisely why the changes of the 1980s and 1990s altered
the face of copyright enforcement.
B. The Erosion of the Gatekeeper System
Gatekeeper regimes have an obvious weakness: They depend on a
specialized good or service remaining specialized. For the 270 years following
copyrighťs 1710 debut, this remained the case for copyrighted
works--copies could not be produced by just anyone. As demonstrated
by Dowling, there could and did arise corruptible publishers who would
produce illicit copies (just as corruptible doctors hand out illicit drugs),
but so long as the costs of finding such corrupted intermediaries remained
reasonable, gatekeeper liability continued to prevent copyright
infringement.
The erosion of copyrighťs gatekeeper system is an ongoing and incomplete
process. The erosion proceeded in several steps, culminating in
the advanced versions of P2P filesharing networks evident today.
Digitalization--the ability to make perfect digital copies of content--
was the beginning of a real problem for the gatekeeper regime. It made
copying certain forms of content possible for anyone with a computer.
As the Office of Technology Assessment documented in 1989, the extent
of an individuaľs copying power was mainly limited to computer
software and analog taping of television programs and music.121
By the
1990s, an individuaľs ability to copy spread to music (with the advent
of powerful compression algorithms) and, to some extent, books and
film.
It is important to understand that digitalization itself did not mean the
end of the gatekeeper system: It simply put home copying within easy
reach. Mass distribution, however, remained (and still remains, for most
works) a gate kept by a few. So long as mass distributors of content remained
identifiable and easy to sue--retail outlets, publishers, and so
on--the gatekeeper regime could remain effective.
Hence, the mass popularity of the Internet in the mid-1990s was another
step toward the erosion of the gatekeeper system. But it is also a
mistake to confuse the potential of the Internet as a mass dissemination
system with the development of an application for such purposes.122
121
See Office of Tech. Assessment, supra note 11, at iii.
122
See generally, Wu, supra note 7 (pointing out that the Internet and its applications
should be understood separately for legal analysis).
138 Virginia Law Review Vol. 89:nnn
Web-based Internet outlets--say, online retailers like Amazon.com--
were and are no less amenable to being copyright gatekeepers. However,
it took the design of P2P filesharing systems to realize the full extent of
the network's structural challenge to a gatekeeper liability system.
A pure P2P design is the logical corollary to a gatekeeper enforcement
system. The design goal of a pure P2P network is the complete
elimination of intermediaries. Such a pure P2P network is a network of
perfect equals, each of which is both a consumer and a distributor of
copyrighted materials. Such a network would force those who enforce
copyrights to rely exclusively on primary enforcement, with its attendant
difficulties.
Today's successful P2P filesharing applications approach, but do not
achieve, a pure P2P model. The following Section explains why.
C. Elements of Peer Design
The design of P2P applications to avoid copyright presents a technical
challenge with implications not fully appreciated by legal scholarship.123
The technical study of P2P design shows that designing a P2P filesharing
network to avoid copyright requires important deviations from the
optimal design for speed, control, and usability. The programmers of a
copyright-resistant P2P network must balance an interest in avoiding legal
liability against the competing interests of ensuring performance on a
mass scale, maintaining system stability, and fostering network trust.
These matters all require control over the network, while a pure peer design
eliminates control as much as possible.
The goals of peer filesharing applications are a good place to begin
the discussion. Two people can trade files easily, using e-mail or a
floppy disk, but what about one million people? The general goal of a
P2P filesharing network is to enable millions of home users to trade files
amongst themselves, quickly and easily. Such a program generally requires
three elements. First, it requires a program that regular home users
can download--a program that, running on their computers, can locate
other users, creating a network of peers. Second, it requires a way
for each user to search the network (or parts of it) to determine what
123
For a good summary of some of these challenges, see Theodore Hong, Performance, in
Peer-to-Peer: Harnessing the Benefits of a Disruptive Technology 203, 205­06 (Andy Oram
ed., 2001).
2003] Compliance & Code 139
content others are making available. Third, it requires a way for users to
send files to each other once they have found something desirable.
Designers accomplish these filesharing goals using a P2P design.
Formally, a P2P network is an application architecture where each
"node," or computer, has equivalent rights and responsibilities.124
Figure 1: Design of a Peer-to-Peer versus a Client-Server Net-
work
This design, as the name suggests, makes a P2P network one of
equals, or peers. This network architecture should, usually, be distinguished
from a "client-server" network in which one computer (the
server) specializes in serving the needs of others (the clients).
Real-world metaphors help capture this important distinction. Consider
the difference between a study group comprised entirely of students
and a lecture led by a teacher. On the one hand, the study group is
a peer network. Each member has both the responsibility to share materials
and the right to take materials from others. On the other hand, the
classroom is a "client-server" network. The teacher specializes in teaching
the students. The students do not teach the teacher or each other. The
network is centralized, and each node is specialized.
124
See Michael A. Gallo & William A. Hancock, Networking Explained 11 (1999).
Peer A
Peer B
Peer C
Peer D
Server
Client A Client B Client C
"Study Group" "Lecture"
Peer Network Client-Server Network
PeerA
PeerB
PeerC
PeerD
"StudyGroup"
PeerNetwork
140 Virginia Law Review Vol. 89:nnn
A pure peer design is "flat," with equal, non-specialized members.
Client-server designs are hierarchical, with a specialized server. Each
design has it own uses, but only peer networks threaten the gatekeeper
structure of copyright enforcement.
D. Purity in Peer Design
The distinction between peer and client-server designs is critical to
understanding the challenge of building a network that resists copyright
enforcement. The closer a network comes to a pure P2P design, the more
disparate the targets for copyright infringement and the greater the threat
to a gatekeeper system.
Why not always build the most decentralized design possible? The
general answer is that it is difficult. Indeed, within the technical community,
variations from "purity" are so commonplace that there are
healthy debates over what should even be considered a peer network.125
Pure peer networks are a design challenge because eliminating intermediaries
decreases control over the network. The loss of control makes
it difficult to ensure performance on a mass scale, to establish network
trust, and even to perform simple tasks like keeping statistics. As networks
grow, these problems become more pronounced. It is simple, in
other words, to build a pure P2P network for six friends interested in
trading, just as it is simple to maintain a study group with six members.
It is difficult, however, to make the same design work for ten million
people.
In practice, there are four recognized classes of application design.
They are pictured in Figures 1 and 2 and are summarized in Table 2.
Figure 1 depicts the two extremes. The Interneťs most important application,
the World Wide Web, represents an archetypal client-server
model. "Pure" peer design, meanwhile, is what the early version of the
Gnutella peer filesharing programs adopted to avoid infringement liabil-
ity.
125
See, e.g., Gene Kan, Gnutella, in Peer-to-Peer: Harnessing the Benefits of a Disruptive
Technology, supra note 123, at 94, 117 ("the debate . . . burning in the technology community
. . . [is] what is truly peer-to-peer."); Clay Shirky, What is P2P . . . And What Isn't?, The
O'Reilly Networks, at http://www.oreillynet.com/pub/a/p2p/
2000/11/24/shirky1-whatisp2p.html (Nov. 24, 2000) (on file with the Virginia Law Review
Association) (arguing that the label describing what is happening to networks, "peer-topeer,"
does not clarify much).
2003] Compliance & Code 141
It is often useful in a peer design to have at least one central server in
which to store user information, search databases, and create a system of
trust. Such a design forms the "centrally coordinated" peer network, pictured
on the right of Figure 2. Napster used this architecture, as do popular
chat programs like AOL Instant Messenger.126
Figure 2: Hybrid Designs
To complete the classification, many of the most well-known networks
are hybrids that balance control and decentralization. They appear
to be P2P networks to the end-users but they are actually only P2P between
specialized servers. This "hierarchical peer-to-peer" network, pictured
on the left of Figure 2, supports regular Internet Protocol ("IP") email,
the Domain Name System (DNS), and the classic newsreader
"USENET."127
With e-mail, no central authority controls delivery of
126
See Nelson Minar et al., A Network of Peers: Peer-to-Peer Models Through the History
of the Internet, in Peer-to-Peer: Harnessing the benefits of a Disruptive Technology, supra
note 123, at 3, 17.
127
See M. Horton & R. Adams, The Internet Eng'g Task Force, Standard for Interchange
of USENET Messages, at http://www.ietf.org/rfc/rfc1036.txt (1987) (on file with the Virginia
Law Review Association) (detailing structure of USENET system).
Hierarchical Peer Network
"Email"
Peer Servers
Peer A
Peer C
Peer D
"Napster"
Server
Server
Server
Server
Clients
Clients
Clients
Server
Peer B
Centrally Coordinated
Peer Network
142 Virginia Law Review Vol. 89:nnn
messages. Rather, a particular university's or company's servers communicate
with other institutional servers in a P2P fashion.128
Table 2: Types of Network and Examples
Network Type Example
Client-Server World Wide Web
Centrally Coordinated Peer
Network
Napster, Instant Messager
Hierarchical Peer Network Email, Usenet, DNS
Pure Peer Network Gnutella
As this discussion shows, what is called a peer network may be decentralized
in only certain respects. Examining the life cycle of a node in
a peer network shows how often intermediaries are needed to smooth the
functioning of even the most basic network. Of course, every intermediary
becomes a potential legal target.
To begin life as a peer node, a user needs to install the appropriate
software. This usually means downloading it from an intermediary
(typically a web site). The node must find at least one peer to join the
peer network, but how can this location be done without knowing a peer
in advance? Again, the solution is usually reliance on some intermediary,
such as a "host cache," that grants the peer node one peer address so
that the user may begin to use the network.
To be useful, the peer node must have some ability to discover what
content is available on the network. For example, in a network meant to
128
See P. Mockapetris, The Internet Eng'g Task Force, Domain Names--Implementation
and Specification, at http://www.ietf.org/rfc/rfc1035.txt (1987) (on file with the Virginia
Law Review Association) (detailing design of domain name system).
2003] Compliance & Code 143
share music, a user needs to know what songs are actually available,
preferably by searching by artist, song title, etc. The very volume of
search traffic thus generated, however, can strain a network design to the
point of collapse.129
Designers may minimize this effect if they design
the network to access a finite amount of content (for example, hit songs).
Nonetheless, designing a network remains a fundamental challenge. It is
easiest to store search information in one place, but if search information
is centralized, as it is in the Napster design, it creates yet another specialized
intermediary.
Finally, peer networks need to provide for connections among peers.
Here, the greatest problems for non-centralized peer models come from
user abuse of anonymity. In a music network, copyright owners could
potentially send around fake files. In network terms, this is the problem
of "trust." Trust systems are difficult, if not impossible, to create without
some centralized system of verification.
The preceding description is a summary of the challenges facing P2P
technology. The point is that P2P design represents a serious challenge
for designers because it requires compromise. Fewer intermediaries
means fewer targets for an infringement lawsuit. The existence of fewer
intermediaries, however, makes it harder for users to use the system,
creates a greater risk of system crashes, and increases the risk of anonymous
attacks. There is a tension between an optimal P2P filesharing network
and the goal of avoiding copyright liability. This condition will
bear strongly in the examination of P2P programming incentives. The
next Section considers how P2P designs have sidestepped social norms
that might have prevented copyright infringement.
E. Copyright and Social Norms
According to a 2000 Pew Internet Project study, seventy-eight percent
of those who download music do not consider it to be stealing and sixtyone
percent do not care if the music they download is copyrighted.130
A
survey reported by two economists showed that only fourteen percent of
respondents considered illegal copying of software to be a serious crime,
compared to thirty percent who felt that way about driving forty miles
per hour in a twenty-five miles-per-hour zone.131
These statistics suggest
129
See Kan, supra note 125, at 112­14.
130
See Lenhart et al., supra note 11, at 5.
131
See Harbaugh & Khemka, supra note 102, at 6.
144 Virginia Law Review Vol. 89:nnn
that P2P applications have not only undermined copyrighťs gatekeeper
regime, but have also successfully sidestepped social norms that might
otherwise bolster compliance with the copyright regime. This Section
describes how code designers structured their applicatio ns to avoid social
norms.
As discussed above, theorists have suggested that the possibility of
state punishment provides an incomplete explanation for observed compliance
with societal rules.132
Rather, they suggest that other systems of
social control, including social norms, account for compliance.133
While
accounts differ, the arguments contend that some mix of the threat of external
social sanctions,134
the fear of sending the wrong signals to oth-
ers,135
and the internalization of ethics136
creates compliance that exceeds
what would be observed as a simple reaction to the threat of punishment.
Those who benefit from copyright laws benefit from the norm that
physically stealing a CD or DVD is socially unacceptable. They are hurt,
however, by the norm that makes copying the same CD at home acceptable.
Despite their cosmetic differences, economically speaking, each
instance of copying represents approximately the same economic loss in
the form of a lost potential sale.137
Therefore, the system of social norms, like the gatekeeper regime, is
an alternative mechanism for creating compliance with a given legal
rule. If norms track the substance of legal rules, it stands to reason that a
rational, widespread effort to reduce the costs of regulation may sidestep
the enforcement of legal rules by manipulating social norms. If it were
considered disgraceful to download music on the Internet, copyright
compliance could be achieved without active, primary enforcement. The
design of P2P networks, however, successfully exploits the status of
132
See supra note 26 and accompanying text.
133
Understanding exactly how norms operate to ensure compliance with legal rules is beyond
the scope of this paper. For a new account of this issue see Mahoney & Sanchirico, supra
note 26, at 41­48 (suggesting that the state's punishments play a role in sustaining
strategies of cooperation with legal rules).
134
A classic external sanction model is provided in Ellickson, supra note 15, at 124­26.
135
The signaling theory is presented in Eric A. Posner, Symbols, Signals, and Social
Norms in Politics and the Law, 27 J. Legal Stud. 765, 766­67 (1998).
136
See Robert D. Cooter, Decentralized Law for a Complex Economy: The Structural Approach
to Adjudicating the New Law Merchant, 144 U. Pa. L. Rev. 1643, 1661­66 (1996)
(characterizing internalization as the precondition of a norm's existence).
137
Assuming a similar likelihood that the thief or friend would have otherwise paid the full
price for the music.
2003] Compliance & Code 145
copyright norms, taking full advantage of an existing ambiguity as to
whether home, non-commercial copying is "wrong."
In one of the few disinterested studies of its time, the Congressional
Office of Technology Assessment conducted a 1989 survey regarding
attitudes toward home copying.138
The study found a simple norm: people
think copying for friends is okay, but copying for money is wrong.139
More precisely, it found that large majorities (sixty-three percent) of
consumers considered making a taped copy of audio materials for a
friend to be "acceptable" or "perfectly acceptable."140
On the other hand,
the greater majority (seventy-six percent) found selling copied materials
unacceptable.141
The survey mirrors widespread anecdotal evidence142
suggesting a normative difference between commercial and noncommercial
copying.
P2P filesharing exploits this distinction brilliantly. P2P clients create
no sensation or impression of stealing (the absence of this quality typifies
what Lior Strahilevitz would call "charismatic code" design).143
Instead,
the user is invited to a "community" of peers who exchange song
files. A user, importantly, has no sense that she is "selling" copyrighted
materials. The design therefore exploits the distinction between the acceptance
of non-commercial copying and the non-acceptance of commercial
copying. While the economic consequences of peer filesharing
could be large, the superficial absence of commercial exchange makes
filesharing more acceptable under the norms of home copying.
Figure 3: The Friendly Face of the Bear-Share Community
138
See Office of Tech. Assessment, supra note 11, at 139­65.
139
Id. at 163.
140
Id.
141
Id.
142
Jessica Litman, for example, argues that in general people "do not observe copyight
rules in their daily behavior," because "people don't believe the copyright law says what it
does say." Litman, supra note 103, at 111­12.
143
See Strahilevitz, supra note 12, at 549.
146 Virginia Law Review Vol. 89:nnn
As an illustrative example, consider the BearShare client pictured
above. There is little on the screen to suggest that a user is engaging in a
morally ambiguous operation or is committing an act of theft. The
friendly bear in BearShare is an icon of charismatic code.
The exploitation of social norms seems to have succeeded in facilitating
a robust filesharing community. The 2000 Pew Internet Project Survey
overwhelmingly supports the view that those who use filesharing
networks do not think they are stealing.144
That same study also suggests
that fifty-three percent of all Internet users, and forty percent of all
Americans, think that by sharing music through the Internet they are not
doing anything wrong.145
Along these same lines, a 2002 survey by
Business Software Alliance found that only thirty-eight percent of Internet
users claimed they would never download a potentially pirated program
to save money.146
In the end, P2P networks not only exploit the
limits of legal enforcement, but also dodge the system of social norms
that fortifies the relevant legal rules.
This Part has demonstrated that the success of P2P depends on the
presence of certain vulnerabilities peculiar to copyright law. Part III
considers the reaction of the beneficiaries of copyright law and the regulatory
competition that followed it.
144
Lenhart et al., supra note 11, at 5.
145
Id. at 6.
146
See Business Software Alliance, Survey Spotlights Growing Problem of Online Software
Piracy, at www.bsa.org/resources/2002-05-29.117.pdf (May 29, 2002) (on file with the
Virginia Law Review Association).
2003] Compliance & Code 147
III. THE EVOLUTION OF P2P DESIGN AND REGULATORY COMPETITION
The years 1999 to 2003 represented a period of regulatory competition
between P2P users and the incumbent industry. At stake were substantial
rents--the monopoly rents obtainable when the copyright law is
enforced. As the succeeding narrative shows, the two groups had different
comparative advantages: one had code, the other had litigation and
legal change. In other words, the competition pitted methods of avoidance
against methods of change.
There are two outstanding aspects to this story. The first is the degree
to which code design evolved to better target the weaknesses of the
copyright regime. Part II demonstrated that P2P networks were generally
designed to target copyrighťs dependence on a gatekeeper system and to
exploit the lack of clear normative support for the copyright system.
This Part will show that the design evolved to take advantage of a specific
legal doctrine--copyrighťs contributory liability doctrine--
embodied in the decision in Sony Corp. of America v. Universal City
Studios,147
and elaborated in A & M Records v. Napster.148
The second is the nature of the reaction to the P2P network. The recording
industry is obviously the beneficiary of the existing copyright
laws, and the erosion of the gatekeeper regime provoked an investment
in various mechanisms of legal change (investments in efforts to change
copyright law). These patterns follow the model of regulatory competition
described in Part I.
A. Napster and its Predecessors
While Napster was the first laboratory for a peer response, it was itself
a reaction to an earlier model. The very first efforts at mass distribution
of copyrighted materials employed a purely client-server model--
essentially, web sites with songs available for download. The company
"MP3.com," which debuted in 1996, is one well-known example. Its
"My.MP3" service allowed users to download, among other things,
copyrighted MP3 files, provided they owned the CD that corresponded
to the file in question.149
This service effectively gave users remote access
to music that they already owned.
147
464 U.S. 417 (1984).
148
114 F. Supp. 2d 896 (N.D. Cal. 2000).
149
See Lawrence Lessig, The Future of Ideas: The Fate of the Commons in a Connected
World 192 (2001).
148 Virginia Law Review Vol. 89:nnn
The architecture of My.MP3 and other web-based services, not the
fair-use issue, is of particular interest here.150
My.MP3 relied on a pure
client-server model. It placed a huge amount of copyrighted material in a
single space. When the recording industry sued, the company's activities
were deemed clearly illegal under the traditional model of copyright enforcement.
151
The recording industry's case was not much different,
enforcement-wise, from the Elvis bootleggers in Dowling v. United
States152
--both were large, centralized copiers of copyrighted materials.
Other sources of copyrighted sounds in the early 1990s were the
primitive, anonymous websites that simply made MP3s available for
download.153
But these sites faced two serious technological problems.
First, if a site became popular it would quickly become overburdened
with user traffic. Second, there were few reliable and straightforward
means for finding such sites.154
Then came Napster. The beta version of Napster debuted on June 1,
1999. Napster's revolutionary design was a response to the legal and
technical problems of the web-based companies. As one commentator
noted, "[Napster] was written to solve a problem--[legal] limitations on
file copying."155
Napster eliminated the intermediary that had doomed My.MP3 and
others. It designed a network that decentralized the infringing content,
leaving the songs on the hard drives of individual home users. Napster
differentiated itself from the traditional commercial copyright pirate by
styling itself as a place to trade music rather than as a place to sell or distribute
it.
Napster, however, was not completely decentralized. Napster's programmers,
Shawn Fanning and Jordan Ritter, were also aware of the
150
Because MP3.com required users to own the CD for the MP3s they were given the right
to download, there was a good argument that MP3.com's copying of the files to facilitate
"space-shifting" was fair use. See id. at 193­94.
151
See UMG Recordings v. MP3.com, Inc., 92 F. Supp. 2d 349, 350 (S.D.N.Y. 2000).
152
473 U.S. 207 (1985).
153
See Bruce Haring, You can't stop the music on the Net: Recording industry debates
MP3 piracy issue, USA Today, Nov. 4, 1998, at 5D, available at 1998 WL 5740934 (noting
the "abundance of sites both legal and illegal").
154
For example, the website MP3Board offered a search engine for such sites and was
quickly sued. See MP3Board v. Recording Indus. Ass'n of Am., No. C-00-20606RMW,
2001 WL 804502, at *3 (N.D. Cal. Feb. 27, 2001) (staying a California countersuit).
155
Shirky, supra note 125, at 28.
2003] Compliance & Code 149
challenge to P2P networks of operating on a mass scale.156
Napster
mixed client-server and peer elements in order to make the search for
songs a fast and scalable solution. Hence, the Napster server facilitated
both database searching and brokering of individual connections.
The design scaled impeccably. While estimates vary, at its height,
Napster had tens of millions of active users, an astonishing technological
accomplishment.157
But the failure to remove itself as an intermediary
with control over parts of the process made Napster, the company, a target
for an infringement lawsuit. That lawsuit came on December 6,
1999.158
The infringement case against Napster boiled down to a question of
control, intimately connected to the network design questions studied
here. The situation would have been different if Napster had been a form
of multi-purpose copying technology over which Napster itself had no
specific power. This relationship between technology and ownership
would have put Napster in the same position as cameras, VCRs, and
other forms of "copying equipment" described in Sony Corp. of America
v. Universal City Studios.159
The makers of VCRs and photocopiers obviously
know that their products are often used to infringe copyright, but
since they are powerless to do anything about these violations and because
the equipment has substantial non-infringing uses, they are not
made liable.160
Napster's argument--that it was a mere instrument of both legal and
illegal uses--was belied by its design. One overriding factual finding
doomed the company: The court found that "[Napster] could block access
to the system by suppliers of the infringing material."161
This fact made Napster the sponsor, rather than just the instrument, of
infringing conduct. Instead of a VCR, Napster's design put it in the classic
position of the dance hall that chooses to allow an infringing artist to
156
See Jordan Ritter, Why Gnutella Can't Scale. No, Really., at
http://www.darkridge.com/~jpr5/doc/gnutella.html (Feb. 2001) (on file with the Virginia
Law Review Association) (discussing scaling problems in P2P networks).
157
See Healey, supra note 93.
158
See A & M Records v. Napster, Inc., 114 F. Supp. 2d 896, 900 (N.D. Cal. 2000).
159
464 U.S. 417, 442 (1984) ("[T]he sale of copying equipment, like the sale of other articles
of commerce, does not constitute contributory infringement if the product is widely used
for legitimate, unobjectionable purposes.").
160
The Sony Court described this as the "staple article of commerce doctrine." Id.
161
A & M Records v. Napster, Inc., 239 F.3d 1004, 1022 (9th Cir. 2001). See also id. at
1023 ("The district court correctly determined that Napster had the right and ability to police
its system.").
150 Virginia Law Review Vol. 89:nnn
play despite having the power to stop the performance.162
The Sony
Court itself declared that when a defendant is "in a position to control
the use of copyrighted works by others," the "imposition of vicarious liability
is manifestly just."163
After the court found that Napster exercised control, holding it to be
both a contributory and a vicarious infringer was easy. Napster's design
allowed the record industry to use the "notice and failure to remove"
formula to prove knowledge (an element of contributory copyright in-
fringement).164
The record industry sent Napster notice of thousands of
infringing files available through the system and then proved that these
files remained available for download.165
On the issue of vicarious liability, the decisive legal question also involved
Napster's degree of control. As the appeals court framed it, the
question was whether Napster had "the right and ability to supervise the
infringing activity and also ha[d] a direct financial interest in such ac-
tivities."166
Napster's architecture again provided an answer. As the
court noted, "Napster retains the right to control access to its system."167
This ruling led Napster to bankruptcy 168
and also taught several legal
lessons to P2P code designers. As the late Gene Kan, a post-Napster developer,
wrote, "[T]he recording industry . . . is sensitizing software developers
and technologists to the legal ramifications of their inventions.
Napster looked like a pretty good idea a year ago, but today Gnutella
and Freenet look like much better ideas."169
Napster taught peer network
designers that both lack of control and general functionality had to be
162
The classic dance hall case is Dreamland Ball Room v. Shapiro, Bernstein & Co., 36
F.2d 354, 355 (7th Cir. 1929) (holding a dance hall liable for copyright infringement because
they hired an infringing orchestra to supply music to paying customers). In contrast, landlords
have traditionally not been held liable for the infringements of their tenants. See, e.g.,
Deutsch v. Arnold, 98 F.2d 686, 688 (2d Cir. 1938) (refusing to hold a landlord liable for the
copyright infringement committed by a tenant on the premises).
163
Sony, 464 U.S. at 437.
164
Napster, 239 F.3d at 1020. This formula has become the favored technique for proving
knowledge in service provider cases. See Religious Tech. Ctr. v. Netcom On-Line Communication
Servs., 907 F. Supp. 1361, 1373­75 (N.D. Cal. 1995) (allowing actual knowledge to
be demonstrated in this manner).
165
Napster, 239 F.3d at 1022.
166
Id. at 1022 n.6.
167
Id. at 1023.
168
The Ninth Circuiťs ruling on the preliminary injunction was the effective, but not the
formal, end of the litigation over Napster. See A & M Records v. Napster, Inc., 284 F.3d
1091, 1099 (9th Cir. 2002) (affirming the district courťs preliminary injunction).
169
Kan, supra note 125, at 121.
2003] Compliance & Code 151
comprehensive and credible to avoid contributory liability. The relationship
between developers and peer networks needed to be more like that
between Xerox and its photocopiers. The response, Napster suggested,
should take the form of a protocol rather than an application. Email and
Usenet had never been sued for copyright infringement, despite their
widespread use for illegal purposes. The lesson was simple--Napster
had not gone far enough.
There was a flurry of attempts to succeed Napster; many so technologically
unsuccessful (Napigator) or so clearly liable under Napster
(Scour) as to be unworthy of discussion. Over the years 1999-2002 there
were approximately fifty-eight different filesharing clients released to
the market.170
Of those, only four or five have enjoyed lasting signifi-
cance.171
One successor was different. It was founded on concepts of
radical decentralization and was clearly designed to avoid the copyright
lawsuit that had befallen Napster. That successor was the protocol
named Gnutella.
B. Early Gnutella: 2000-2001
"Before [Gnutella], systems were centralized and boring."172
Gnutella was a child of the open-source movement. Its unusual name,
non-linear development origins, and relative difficulty of use are all
hallmarks of an open-source work product.173
Gnutella delivered a radi-
170
The compiled list of filesharing clients from 1999­2002 includes: Abe's MP3 finder,
Aimster (now named Madster), Ares, Audio-Galaxy, AudioGnome, BadBlue, Bearshare,
Blubster, CuteMX.Com, DirectConnect, eDonkey, FileAngel, Filetopia, File Navigator, File
Rogue, FileSpree, Free Haven, Freenet, Frost, Gnotella, Gnucleus, Gnutella 0.56, Gnutmeg,
Grokster, Groove Network, Hotline Communications, iMesh, iSwipe, Junge Monkey, KaZaA,
KonSpire, Limewire, Mactella, Mojo Nation, Morpheus, MyNapster, Myster, NapMX,
Napster, Nutella, Ohaha, OnSystem, OpenNap, Phex, Phosphor, Pointera, Publius, Qtella,
Qube, Scour.com, Shareaza, Spinfrenzy, SongSpy, Taxee, Voodoo Vision, WinMX, Xolox.
Of course, many of these are clients for the same networks, as in the multiple GnutellaNet
and FastTrack clients.
171
In focusing on the major developments, some might argue that I have shortchanged
programs like Scour.com and Aimster in the process.
172
Kan, supra note 125.
173
Gnutella was released in March of 2000. It was invented by Justin Frankel and Tom
Pepper, two programmers working for AOĽs Nullsoft. Id. at 95. AOL quashed the effort,
but Gnutella's cause was picked up by the open-source movement. See Andy Oram,
Gnutella and Freenet Represent True Technological Innovation, at
http://www.oreillynet.com/lpt/a/208 (May 5, 2000) (on file with the Virginia Law Review
152 Virginia Law Review Vol. 89:nnn
cally decentralized design that made it a darling of academic study. The
design was an intentional effort to create a filesharing protocol that
could avoid a lawsuit. Although it succeeded, it did so at the expense of
social and scalability problems.
Gnutella's decentralization was nearly absolute. No single node on
the early Gnutella network was different than any other. Searching, file
transferring, and peer finding were all accomplished without the creation
of specialized intermediaries. The only identifiable intermediaries were
those relatively limited sites that made the early Gnutella client (version
0.56) available for download.
Gnutella developers compare the network they designed to a cocktail
party where users trade files with whomever happens to be nearby.174
The design implements the idea that "Gnutella is not a program, it is a
protocol."175
In other words, Gnutella's designers created a filesharing
network--GnutellaNet--that was unowned and uncontrolled and to
which various Gnutella programs could provide access. The relationship
between the application and the network was similar to that between
various email programs (Eudora, Microsoft Outlook, Hotmail) and the
one-serves-all email network that cannot be said to be owned by anyone.
GnutellaNet was designed as a general filesharing network capable of
sharing any computer file.176
Gnutella was a success on the legal front. Gnutella's radical decentralization
avoided the legal liability that had plagued Napster. To date,
neither GnutellaNet nor its main application designers have been sued,177
despite the substantial volume of infringement they facilitate.
Association). Gnutella's full development followed (and still follows) the non-linear path
characteristic of open-source code. Id.
174
See Kan, supra note 125, at 97­98.
175
From Strategic Vision to a 10-Point Tactical Plan: A followup to The Revolution Will
Be Webcast (working paper), at http://economicdemocracy.org/counterspinner.
html (last visited Mar. 20, 2003) (on file with the Virginia Law Review Association).
176
This characteristic was even more evident in another network, FreeNet, aimed at
achieving the goals of the World Wide Web (storage of information) in a decentralized,
purely P2P fashion. A discussion of the methods used by FreeNet can be found in Ian Clarke
et al., Freenet: A Distributed Anonymous Information Storage and Retrieval System, in Designing
Privacy Enhancing Technologies: International Workshop on Design Issues in Anonymity
and Unobservability 46 (Hannes Federrath ed., 2001); see also Adam Langley,
Freenet, in Peer-to-Peer: Harnessing the Benefits of a Disruptive Technology, supra note
123, at 123 (describing the development and structure of Freenet).
177
It is true that Morpheus switched to Gnutella after it was sued as one of the three FastTrack
companies, but no Gnutella developer qua Gnutella developer has been sued.
2003] Compliance & Code 153
The early GnutellaNet, however, was plagued by stability and performance
problems attributable to its decentralized design. In late July
of 2000, the Gnutella network underwent its first major crash, leaving
the network unusable for more than a month.178
The 2000 crash was the
first sign that the early Gnutella client design had traded resistance to
litigation for system instability.
Commentators quickly diagnosed the problem. 179
Early GnutellaNeťs
stability relied on users' willingness to donate both bandwidth and music
files to a common cause and to limit judiciously their own use of the
network. Once a certain number of users joined the network, stark differences
in user bandwidth and the lack of a central mechanism for allocating
traffic to more capable users made a crash inevitable. While some
touted the theoretical scaling capabilities of Gnutella,180
the instability of
early GnutellaNet was undeniable. In addition to the scaling problem,
there were also "social" problems. There was no incentive (not even social
incentives, given the anonymous nature of the network) to act selflessly.
A 2000 Xerox/PARC study established that almost seventy percent
of Gnutella users shared no files and that nearly fifty percent of all
responses were returned by the top one percent of sharing hosts.181
While this did not necessarily matter if the goal was trading the 100
most popular songs, Napster's deeper appeal had been the range of content
it made available. The lack of any mechanism to police selfishness
in Gnutella compromised the potential of the common solution.
178
See Steve McCannell, The Second Coming of Gnutella, WebReview (Mar. 2, 2000), at
http://www.webreview.com/mmedia/2001/03_02_01.shtml (on file with the Virginia Law
Review Association) (detailing reasons for the crash). Interestingly, the crash came directly
in the wake of the Napster injunction as thousands of Napster users attempted to migrate to
Gnutella. Id. The crash provided a dramatic demonstration of the difference in scaling capabilities
between the two approaches.
179
See id; Matei Ripeanu, Peer-to-Peer Architecture Case Study: Gnutella Network, at
http://www.cs.uchicago.edu/files/tr_authentic/TR-2001-26.pdf (last visited Mar. 20, 2003)
(on file with the Virginia Law Review Association) (describing Gnutella's scaling problems);
Ritter, supra note 156; See also Hong, supra note 123, at 206­07 (summarizing a now
unavailable Clip2 study of the crash). A network engineer would diagnose the problem as
follows: Gnutella's layer 7 topology did not map carefully to the physical network, meaning
the network failed to make use of available resources.
180
Gene Kan, for example, argued that Gnutella would scale perfectly well, and that the
2000 crash was caused by an inappropriate add-on technology. Kan, supra note 125, at 109­
17.
181
Eytan Adar & Bernardo A. Huberman, Free Riding on Gnutella, 5 First Monday 10
(Oct. 2000), at http://www.firstmonday.dk/issues/issue5_10/adar/index.html (on file with the
Virginia Law Review Association).
154 Virginia Law Review Vol. 89:nnn
The problems of Gnutella 2000 were generally recognized,182
yet
Gnutella's failures were not the end of the peer filesharing response. The
crashes and instability led to a new generation of peer filesharing software.
These new-generation programs, bearing names such as KaZaA,
Grokster, Morpheus, and BearShare, are, for now, the latest chapter of
the peer response story.
C. The KaZaA Era: 2001 - Present
The legal vulnerabilities of Napster and the stability and social problems
of Gnutella inspired a new approach. Led by the enigmatic KaZaA,
and its FastTrack engine, a new generation of peer-sharing applications
tried to strike a balance between suability and scalability. Unlike the
original Gnutella, they allowed some hierarchy and made some effort to
engineer polite behavior. At the same time, they tried to avoid the centralized
control that doomed Napster. The results are programs of great
sophistication, attuned carefully to the doctrines of copyright.
The new generation reintroduced hierarchy among users. They created
a distinction between "regular peers" and "superpeers" based on detected
resources--in particular, bandwidth.183
In this hierarchy, college
students are on top: high bandwidth users (college students on university
networks, home DSL, and cable users) are superpeers, while dial-up users
(home modem users) are regular peers.
Dozens of programs grew into the technological gap between Napster
and Gnutella. Only a few, however, reached mass scale for any length of
time.184
1. FastTrack & KaZaA
FastTrack returned filesharing to an enterprise of substantial scale. In
late 2001, the FastTrack network grew to be the largest filesharing network
since Napster, with an average of two to four million users online
182
See Hong, supra note 123, at 206­07; McCannell, supra note 178; Ripeanu, supra note
179.
183
Names vary: BearShare groups users into "ultrapeers" and "leafs." See BearShare,
Gnutella Good Citizen Tips, at http://www.bearshare.com/help/citizen.htm (last visited Feb.
9, 2003) (on file with the Virginia Law Review Association).
184
Some of the more major programs from this period not discussed here include AudioGalaxy,
Aimster (now named Madster), WinMX, iMesh, and OpenNAP.
2003] Compliance & Code 155
at any given time.185
Dutch programmers Niklas Zennstrom and Janus
Friis created the FastTrack protocol late in 2000 and wrote a client application,
KaZaA, to access the FastTrack network.186
Unlike Gnutella,
the protocol was never released as an open-source standard.187
Instead,
KaZaA insisted that companies pay to access the FastTrack network.
The companies Grokster and Morpheus188
did so, creating several client
alternatives.
The FastTrack companies fit somewhere between early Gnutella and
Napster in their elimination of intermediaries. The protocol borrows
heavily from Gnutella. It also maintains the distinction between the protocol
and the clients; the company KaZaA, for instance, maintains no
power to "shut down" the network.189
FastTrack deviates from the pure design of early Gnutella in several
significant ways. First, it implements a very sophisticated system of superpeering
designed to avoid scaling problems. This system has been a
success. The KaZaA superpeer system, from user accounts, produces
much better performance than even the next-generation Gnutella cli-
ents.190
Such tiering, however, means that not all users are equal; a finite
number of superpeers do the bulk of the work.
Second, the FastTrack companies have, like Napster, centralized several
functions. A central server is still responsible for maintaining user
registrations, logging users in to the system (in order to maintain statistics),
and helping the process of finding peers in the first place. As previously
discussed, efficient operation is difficult to maintain in the face
185
This number is based on Active Users statistics, at www.sylk.com (July 22­Aug. 8,
2002) (on file with the Virginia Law Review Association).
186
Kevin Maney, Fight over digital music filesharing keeps getting weirder, USA Today,
Sept. 25, 2002, at B3.
187
It has been reverse-engineered by several groups, who create clients that access the
FastTrack network without permission. A prominent example is giFT (giFT isn't FastTrack).
See generally What is the giFT project?, at
http://gift.sourceforge.net/docs.php?document=whatis.html (Sept. 14, 2002) (on file with the
Virginia Law Review Association) (describing giFT and OpenFT).
188
See Benny Evangelista, Morpheus software morphing: Maker of filesharing program to
put limits on MP3 swapping, S.F. Chron., Mar. 14, 2002, at B1. Morpheus later reverted to
Gnutella, after licensing disagreements with KaZaA. See id.
189
See Amy Harmon, Music Industry in Global Fight on Web Copies, N.Y. Times, Oct. 7,
2002, at A1.
190
See, e.g., Morpheus 2.0--Revisited, Slyck, at http://www.slyck.com/newsaug2002/
081902b.html (Aug. 19, 2002) (on file with the Virginia Law Review Association) (discussing
loss of performance when Morpheus switched from FastTrack to Gnutella network).
156 Virginia Law Review Vol. 89:nnn
of radical decentralization.191
Third, at least one of the FastTrack companies
(KaZaA) promotes selfless behavior by sharing user files without
telling the user. A 2002 Hewlett Packard study demonstrated that the
KaZaA client made it difficult to know what files users were sharing.192
The study demonstrated that many users were sharing all of the files on
their computers, but were totally unaware of that fact.193
Increasing the
number of shared files, of course, improves the performance of the net-
work.
Finally, the FastTrack companies also adopted another avoidance
strategy--jurisdictional exit. 194
KaZaA's parent is incorporated in
Vanuatu, a group of islands in the South Pacific noted for its lack of a
copyright law. Grokster maintains its servers in Nevis, a thirty-six
square mile nation-state in the West Indies. Only Morpheus resides in
the United States.195
2. Next-Generation Gnutella
GnutellaNet, meanwhile, continued to operate on a smaller scale. Recall
that neither GnutellaNet nor any Gnutella client has ever been
sued--their problems are instead self-generated. Gnutella responded to
its scaling and social problems by adopting a superpeer design similar to
that of FastTrack. The best known of the new GnutellaNet developers
are Bearshare and Limewire. Both compromise a purely decentralized
design of equal users by distinguishing between high- and lowbandwidth
users and by giving the former more duties.
The continued growth of Gnutella was marked by a lack of coordination
among developers. As Kelly Truelove writes, "Unfortunately,
Gnutella has a history of aborted, failed or poorly supported attempts to
unite developers; the analogy of herding cats has rarely been so apt."196
191
See supra text accompanying notes 172­82.
192
Nathaniel S. Good & Aaron Krekelberg, Usability and privacy: a study of Kazaa P2P
file-sharing, HP Laboratories, at http://www.hpl.hp.com/techreports/2002/HPL-2002163.pdf
(June 5, 2002) (on file with the Virginia Law Review Association).
193
Id.
194
See Harmon, supra note 189, at A1.
195
See id.
196
Kelly Truelove, Gnutella: Alive, Well, and Changing Fast, OpenP2P.com, at
http://www.openp2p.com/pub/a/p2p/2001/01/25/truelove0101.html?page=2 (Jan. 25, 2001)
(on file with the Virginia Law Review Association).
2003] Compliance & Code 157
Major Gnutella clients have also taken measures to "engineer good
behavior." For example, Bearshare and Limewire block requests from
clients who do not contribute files to GnutellaNet.197
These efforts, as
was the case with the FastTrack companies, may make these clients easier
to sue because they suggest an increased quantum of "control" over
the Gnutella network.
Finally, despite the change, Gnutella still appears to have scaling
problems. Statistics kept by Limewire show that, during the first half of
2002, the network size rarely reached more than 500,000. 198
By July
2002, GnutellaNet had declined to an average of 160,000 nodes.199
Gnutella experts point to the same general problem: no control over selfish
behavior. An anonymous source at Limewire explained the problem:
"Client `A' may excessively query (hammer) three or more UltraPeers.
While this may produce plentiful results, the overall affect [sic] on the
network is negative as it slows queries from more reasonable clients."200
These concerns show the continuing difficulty in balancing decentralization
and selfless behavior. Yet the fact that GnutellaNet remains unsued
endows it with an aura of continued importance in the filesharing story.
D. FastTrack and Gnutella Go to Court
On October 2, 2001, the music industry sued Grokster, Morpheus, and
KaZaA: the three principal FastTrack companies.201
This ongoing lawsuit
is a signal test of the viability of designing code to avoid legal liability.
Programmers wrote FastTrack and Gnutella to exploit loopholes left
by the Napster decision. The case, styled Metro-Goldwyn-Mayer Studios
v. Grokster, asks whether the P2P programmers have succeeded. The
initial answer is yes.
The music industry's complaint made every effort to stress the similarity
between Napster and KaZaA and the other FastTrack companies.
197
Namely, the clients Gnute and Gnutella.it allowed users simply to use GnutellaNet to
download files. See id.
198
Gnutella's Decline, Slyck, at http://www.slyck.com/newsjuly2002/071702a.html (July
17, 2002) (on file with the Virginia Law Review Association).
199
Id.
200
Id.
201
Complaint for Damages and Injunctive Relief for Copyright Infringement at 2, 8,
Metro-Goldwyn-Mayer Studios v. Grokster, 2003 WL 186657 (C.D. Cal. Jan. 9, 2003) (No.
Civ.01-08541) (seeking damages and injunctive relief for copyright infringement), available
at http://www.mpaa.org/Press/KaZaA_Complaint.htm (last visited Feb. 9, 2003) (on file with
the Virginia Law Review Association).
158 Virginia Law Review Vol. 89:nnn
Once again it accused the companies of creating "a 21st century piratical
bazaar." It noted that the defendants grant access to "a closed computer
network, controlled by Defendants." 202
It also put emphasis of the fact
that communications are centrally encrypted.203
The complaint highlights
these facts to support the argument that the FastTrack companies,
like Napster, "are capable of controlling the activities of their users."204
But the facts did not support this contention (by design). District
Judge Stephen Wilson, granting summary judgment in favor of Grokster
and Morpheus,205
refused to buy the comparison to Napster (the program).
The opinion suggests that the changes in design "worked," at
least with respect to negating the element of control that sealed Napster's
fate.
Just as in Napster, the court took the issue of control as the sine qua
non of contributory liability. As the court put it: "[T]he critical question
is whether [defendants] do anything, aside from distributing software, to
actively facilitate--or whether they could do anything to stop--their users'
infringing activity."206
It was here that the changes in design made a
difference. Judge Wilson pointed out, in a crucial factual holding:
Neither StreamCast nor Grokster facilitates the exchange of files between
users in the way Napster did. Users connect to the respective
networks, select which files to share, send and receive searches, and
download files, all with no material involvement of Defendants. If either
Defendant closed their doors and deactivated all computers within
their control, users of their products could continue sharing files with
little or no interruption.207
As noted above, Gnutella and FastTrack embody a self-conscious effort
to make P2P filesharing more like the VCR or photocopier. And the
court accepted just that rationale, concluding, "Grokster and StreamCast
are not significantly different from companies that sell home video
recorders or copy machines, both of which can be and are used to
202
Id. (seeking damages and injunctive relief for copyright infringement).
203
Id. at 2­3.
204
Id. at 10.
205
See Metro-Goldwyn-Mayer Studios, Inc. v. Grokster, Ltd., __F. Supp. 2d__, 2003 WL
1989129 (C.D. Cal. Apr. 25, 2003). Default judgment was entered against a shell company
named KaZaA BV, holding no relevant assets, who declined to defend the lawsuit. See id at
4 n.2. Sharman Networks, the current operator of the KaZaA system, was not a party to the
motions. See id. at 6.
206
Id. at 19.
207
Id. at 23­24.
2003] Compliance & Code 159
corders or copy machines, both of which can be and are used to infringe
copyrights."208
Hence, "absent evidence of active and substantial contribution
to the infringement itself, Defendants cannot be liable."209
Will the judgment survive appeal? There are reasons to suspect it will
not. The district courťs decision ultimately depends on Sony as interpreted
by Napster. Sony in turn, can be read as a policy judgment aimed
principally at correcting a perceived market failure.210
While clearer in
hindsight, it is apparent that the Sony decision correctly addressed a
market failure.211
The VCR broadened the addressable market for television
shows (via time-shifting) and for movies (via rentals). Though there
is an argument that filesharing helps the music industry, the desirability
of the "help" is much less apparent.212
Filesharing looks more like a replacement
for legitimate music sales; such reasoning may compel a
court to find some way to assess liability on P2P developers regardless
of the Napster precedent. In addition, the ratio of infringing to noninfringing
use must be at the forefront of the ultimate policy judgment in
this area. If the alleged non-infringing uses retain their de minimis character
the Court of Appeals will presumably feel that there is little consequence
in ruling against the P2P filesharing companies.
Perhaps the only policy reason to think otherwise is an institutional
argument. Ultimate settlement of the filesharing dispute, the argument
goes, is a task for Congress; and a decision against the copyright owners
will force such settlement. Such "settlement-forcing" decisions in copyright
have a long pedigree, from the 1908 piano-roll case White-Smith
Music Publishing Co. v. Apollo Co.,213
to the cable-broadcast decisions
of the 1960s and 1970s.214
Both prompted Congressional action to settle
a dispute between an incumbent and challenger technology.215
Judge
208
Id. at 27.
209
Id.
210
See Wendy Gordon, Fair Use as Market Failure: A Structural and Economic Analysis
of the Betamax Case and its Predecessors, 82 Colum. L. Rev. 1600 (1982) (describing fair
use as a mechanism for correcting market failure).
211
See id; see also Alfred C. Yen, A Preliminary Economic Analysis of Napster: Internet
Technology, Copyright Liability, and the Possibility Of Coasean Bargaining, 26 U. Dayton
L. Rev. 247, 260­63 (2001) (giving the basic economic argument that Napster be held liable
and suggesting Coasean complications).
212
For studies supporting this position, see Chipman, supra note 98; Lahey, supra note 98.
213
209 U.S. 1 (1908).
214
See Fortnightly Corp. v. United Artists, 392 U.S. 390 (1968); Teleprompter Corp. v.
Columbia Broad. Sys., Inc., 415 U.S. 394 (1974).
215
See 17 U.S.C. §§ 111, 115 (cable and mechanical compulsory licenses).
160 Virginia Law Review Vol. 89:nnn
Wilson clearly had this in mind, writing "[w]hile the Court need not decide
whether steps could be taken to reduce the susceptibility of such
software to unlawful use . . . additional legislative guidance may be
well-counseled."216
It remains to be seen whether the Courts of Appeals
will think that trying to force settlement of copyright disputes remains
an appropriate court function.
E. The Reaction to the Reaction
The recording industry's reactions to P2P filesharing is a paradigm
for understanding the modern face of regulatory competition. As preceding
Sections have shown, the filesharer's comparative advantage lay in
designing code to avoid copyright law. The recording industry, meanwhile,
has invested in a broader range of mechanisms to influence law
and its effects. The content industry invested in changing the law (by
controlling access, increasing intermediary liability, and increasing
criminal liability), in changing social norms, and in changing code to attack
P2P networks. The extent of increased investment can be seen from
the annual increases in the budget of the RIAA itself. In the early 1990s,
the RIAA's budget was estimated at $10 million.217
By 1995, the RIAA
was spending $14.7 million. By 2000, the budget had tripled to $39 million
and in 2001 stood at $44 million.218
1. Investments in Change
The recording industry's investments in legal change are most prominent.
In the 1990s, the content industry invested considerable time and
energy to ensure the passage of three critical laws to buttress copyright
enforcement: the anti-circumvention provisions of the DMCA, 219
the
ISP-liability sections of that same bill,220
and the enhancement of copyrighťs
criminal penalties in the No Electronic Theft ("NET") Act.221
The anti-circumventions provisions of the DMCA reinforce various
technological techniques of preventing copying by criminalizing cir-
216
Grokster, __F. Supp. 2d__ at 33.
217
See Bill Holland, Tougher RIAA gives pirates chase but still running uphill, BillBoard,
Mar. 14, 1992, available at 1992 WL 11645300.
218
See Bill Holland, Performers Give Testimony Before Judges And Lawmakers, BillBoard,
Sept. 22, 2001, available at 2001 WL 24692410.
219
17 U.S.C. § 512(1998).
220
17 U.S.C. § 512(1998)
221
17 U.S.C. §§ 101, 506 (2000).
2003] Compliance & Code 161
cumvention of copy protection systems.222
The provisions can be understood
as an effort to restore an eroding gatekeeper system. Technological
copy protection "respecializes" the creation and mass distribution of
copyrighted works, while the DMCA's anti-circumvention law makes it
a crime to undo the respecialization.223
The law can be understood as an
effort to return content owners to the 1970s, when they were free to sit
back and police the few intermediaries licensed to access the copyprotected
content.224
The anti-circumvention provisions are usually discussed
in conjunction with the much-discussed possibility of effective
digital rights management.225
Some have speculated that there may be a
future where content owners manage to encrypt content so carefully and
comprehensively from the outset, and maintain their control continuously,
that the code prevents infringement ex ante. 226
Such efforts remain
in their early stages and are highly speculative, but they would
eventually transform the economics of copyright enforcement.227
The ISP-specific sections of the DMCA, Section 512 et seq., represent
the culmination of an effort to replace the lost intermediaries of
times past with ISPs. There have been a few attempts to use these sections
to target filesharing. In the summer of 2002, the RIAA filed a lawsuit
against various telephone companies who operate the backbone of
the Internet, based on their failure to contain overseas copyright infringement,
but it dropped the suit a week later.228
More recently, using a
222
See 17 U.S.C. § 1201 (1998).
223
A more recent example in the same vein is the well-known "Hollings Bill." Consumer
Broadband and Digital Television Promotion Act, S. 2048, 107th Cong. (2002). The bill
would require all "digital media devices" to include copy protection technology in their designs.
Id. § 5(a). It can be otherwise described as an effort to place the burden of preventing
copyright infringement on electronics manufacturers.
224
It is worth briefly noting, however, that there is a problem with the DMCA strategy as a
response to filesharing. Unless completely successful in blocking access, digital protection
schemes can simply make legal, protected products even less attractive than the competitors
available through filesharing networks.
225
See, e.g., Lessig, supra note 149, at 177­99; see also Raymond Ku, The Creative Destruction
Of Copyright: Napster And The New Economics Of Digital Technology, 69 U.
Chi. L. Rev. 263, 275­76 (2002) (describing digital rights management systems).
226
These ideas are explored more fully in Lessig, supra note 149, at 177­99.
227
Cf. Ku, supra note 225, at 275­76 (arguing against copyright protection for digital
works because the economics of digital technology undercuts prior assumptions about the
efficacy of a private property regime for information).
228
Alex Pham, Tactics Toughen on Music Piracy, L.A. Times, Aug. 21, 2002, at C1; Alex
Pham, Technology RIAA Drops Suit Targeting Piracy Site, L.A. Times, Aug. 22, 2002, at
C5.
162 Virginia Law Review Vol. 89:nnn
different section of the DMCA, the RIAA successfully convinced a federal
judge to require Verizon to identify a subscriber accused of
downloading hundreds of copyrighted files in a single day.229
Finally, the efforts to pass the NET Act of 1997 and subsequent lobbying
represent an effort to turn to the criminal side of copyright to enhance
primary enforcement. Under the little-noticed NET Act, the federal
government may criminally prosecute relatively minor copyright
infringements.230
While this criminal statute still requires "private financial
gain," the NET Act defines "financial gain" to include "receipt, or
expectation of receipt, of anything of value, including the receipt of
other copyrighted works."231
This definition makes quid pro quo filetrading
potentially criminal.
Copyright owners have mounted an effort to convince the Justice Department
to enforce the NET Act against individual peer filesharers.232
This amounts to an attempt to increase the sanction, if not the probability
of detection, for copyright infringement. The untested question remains
whether either criminal or civil primary enforcement will be effective
in deterring illegal P2P filesharing, given the limits of primary
enforcement in producing deterrence.233
2. Extralegal Investments
In addition to legal changes, the recording industry has also invested
in trying to change the social norms surrounding copyright infringement
and has made some efforts to combat P2P filesharing directly. As discussed
above,234
when it comes to copying files, people have proven to
be unaffected by the ethical tug of the copyright statute. The software
and recording industries have spent a decade trying to change that atti-
229
In re Verizon Internet Services, No. Civ.A.02-MS-0323, 2003 WL 141147 (D.D.C. Jan.
21, 2003).
230
17 U.S.C. §§ 101, 506 (2000).
231
Id § 101.
232
See Benny Evangelista, Casting a wider net/Recording industry may target individuals
in online piracy battles, S.F. Chron., Aug. 22, 2002, at E1, available at 2002 WL 4028698
(recounting efforts to have the Justice Department enforce the criminal side of copyright
law).
233
See Lisa M. Bowman, File-traders in the crosshairs, CNET News, at
http://news..com/2100-1023-943881.html (July 15, 2002) (on file with the Virginia Law Review
Association) (reporting that the recording industry is considering a program of lawsuits
against end-users).
234
See discussion supra Section II.E.
2003] Compliance & Code 163
tude. The RIAA's "Sound-Byting" campaign, for example, is an investment
to try to change the attitudes of college students toward copyright
infringement. The central message of this campaign is: "[U]ploading and
downloading somebody else's music without their permission isn't just
against the law. Iťs a rip-off. Simple as that." 235
Hilary Rosen, the
President of the RIAA, even participated in a well-publicized public debate
at Oxford, arguing that illegal filesharing is immoral. 236
As the statistics
cited here and the Pew Internet Project Study discussed above in-
dicate,237
it is unclear whether these efforts have had much success in
changing public attitudes toward filesharing.
Perhaps most interestingly, content owners may also be taking a page
from the book of P2P designers themselves, using code to influence the
enforcement of copyright law by attacking the P2P networks that undermine
copyright enforcement. Reports on the use of attacks on P2P
networks are hard to verify. However, there are several methods through
which content owners might try to disable P2P networks. One method
seeks to decrease the attractiveness of P2P networks, often by flooding
the network with dummy or broken music files. 238
Users then must
spend more time looking for good files, increasing the attractiveness of
conventional distribution channels. Another method would simply attack
important network nodes using techniques familiar to computer hackers.
239
More fanciful examples of this type of strategy include that of a virus
designed to detect illegally copied materials.240
The extent to which
these methods are used today is a carefully guarded secret.241
The continued
activity of peer filesharing networks, however, suggests either
limited success or limited usage of such techniques.
235
See Soundbyting Home Page, at http://www.soundbyting.com/html/who_we_are/
are_index.html (last visited July 24, 2002).
236
See Matt Bai, Hating Hilary, Wired 11.02 (Feb. 2003), at http://www.wired.com/
wired/archive/11.02/hating_pr.html (on file with the Virginia Law Review Association).
237
See supra text accompanying note 130.
238
For an entertaining account of how anonymity can be used against peer networks, see
Doug Lichtman & David Jacobson, Anonymity a double-edged sword for pirates on-line,
Chi. Trib., Apr. 13, 2000, at 25.
239
For example, a "Denial of Service" attack, which floods a given network node with re-
quests.
240
See Michael Adler, Note, Cyberspace, General Searches, and Digital Contraband: The
Fourth Amendment and the Net-Wide Search, 105 Yale L.J. 1093, 1098­1100 (1996) (presenting
the hypothetical of a program that roamed the net searching for contraband).
241
See generally Todd Woody, The Race to Kill Kazaa, Wired 11.02 (Feb. 2003), at
http://www.wired.com./wired/archive/11.02/kazaa-pr.html (on file with the Virginia Law
Review Association) (surveying methods of technological self-help).
164 Virginia Law Review Vol. 89:nnn
On June 25, 2002, Representative Howard Berman of North Hollywood,
California proposed that "[t]echnological self-help" should help
provide the solution to "unbridled" peer network piracy.242
He proposed
a bill that would give legal license for copyright owners to disrupt peer
networks.243
Representative Berman phrased his support of the bill, interestingly,
in terms of "freedom to respond":
[W]hile P2P technology is free to innovate new and more efficient
methods of distribution that further exacerbate the piracy problem,
copyright owners are not equally free to craft technological responses.
This is not fair and I believe Congress should free copyright creators
to develop and deploy technological tools to address P2P piracy.244
The Berman bill, while unlikely to pass, shows the dramatic extent and
even creativity of efforts to gain advantage in the regulatory competition
surrounding copyright.
The story of the competition between the RIAA and P2P users delivers
a snapshot of the future of understanding compliance and legal effects.
A law's meaning and effects, success or failure, seem ever less a
function of drafting or enforcement. Rather, the question is what
forces--social, economic, technological or otherwise--may be recruited
for or against the cause.
IV. THE SOCIAL DYNAMICS OF P2P FILESHARING
"As the largest grassroots effort in the history of the world, file trading
is essentially the average person's way of saying we don't agree with
the status quo."245
Over the last four years, P2P networks have provided a sub-group of
media consumers with the equivalent of a temporary repeal of copyright
laws for the technologically inclined. How can one explain the growth
242
Press Release, Representative Howard Berman, Berman Announces Legislation To Foil
Peer To Peer Piracy, at http://www.house.gov/berman/pr062502.htm (June 25, 2002) (on file
with the Virginia Law Review Association).
243
To amend Title 17, United States Code, to limit the liability of copyright owners for
protecting their works on peer-to-peer networks: Hearing on H.R. 5211 Before the House,
107th Cong. (2002) (statement of Rep. Howard Berman).
244
Press Release, supra note 242.
245
Richard Menta, RIAA and MPAA sue Morpheus, Grokster and KaZaa,
MP3newswire.net, at http://www.mp3newswire.net/stories/2001/sue_morpheus.html (Oct. 3,
2001) (on file with the Virginia Law Review Association).
2003] Compliance & Code 165
and popularity of the peer filesharing movement? This final Part analyzes
the particular fit between P2P filesharing and its beneficiaries. It
shows first, that P2P may represent the rational exploitation of the larger
group of music consumers by a subset of computer savvy P2P users, and
second, that peer filesharing uniquely suited the disorganized nature of
copyright consumers as a group.
A. Copyrighťs Divided Subjects
One reason P2P filesharing may have been successful is because users
rationally exploited "regular" consumers who lack the knowledge or resources
to use P2P networks. In the standard (if sometimes disputed) account,
copyright law is said to serve the interests of content consumers.
246
The law provides financial and, debatably, expressive incentives to
create materials that would otherwise not exist. If this is right, why
would consumers ever want to disobey copyright law?
The intuitive answer is that everyone likes getting things for free, but
the answer from economic theory is more enlightening. While complying
with some form of copyright law may serve the collective interest of
consumers, it is not in any given individuaľs interest to comply. More
generally, the logic of collective action suggests that the ideal strategy
for an individual or sub-group under copyright law is to create a system
that limits evasion of copyright to an "in-group," leaving everyone else
to pay for the incentives to create. To defect while others remain in
compliance is to live in the game theorisťs version of utopia.
In the mid-to-late 1990s, an important demographic trend favored the
development of just such a strategy. Social commentators began to use
the term "digital divide"247
to refer to the fact--confirmed by empirical
246
Whether copyright does indeed encourage creative expression is a question beyond the
scope of this study of response. I therefore do not address the position held by some that
copyright retards the creation of content. For examples of such positions see, e.g., Eben Moglen,
Liberation Musicology, The Nation, Mar. 12, 2001, at 5; Mark S. Nadel, Questioning
The Economic Justification For (and Thus Constitutionality of) Copyright Law's Prohibition
Against Unauthorized Copying: § 106 (unpublished manuscript), at
http://www.aei.brookings.org/admin/pdffiles/Nadel.pdf (Jan. 2003) (on file with the Virginia
Law Review Association); cf. Stephen Breyer, The Uneasy Case for Copyright: A Study of
Copyright in Books, Photocopies, and Computer Programs, 84 Harv. L. Rev. 281 (1970)
(questioning whether granting copyrights in books and computer programs is really necessary
to provide incentives to create and publish them).
247
The question of who coined the term "digital divide" remains something of a mystery.
See Sharon Foster & Adrianna Borkowski, Who Coined the Term?
166 Virginia Law Review Vol. 89:nnn
study--that there was a sharp division between a relatively small number
of computer literate, connected citizens and the rest of Americans. In
1998, for instance, the Clinton Administration found that collegeeducated
Americans were almost ten times more likely to own a computer
than those without any high school education (63.2% vs. 6.8%).248
The disparity in Internet access was even more prominent: 38.4% of college-educated
Americans had access, as compared to 9.6% of those with
a high school diploma, and just 1.8% of those without any high school
education.249
The existence of this division in content consumers provided ideal
conditions for the development of a copyright evasion system that could
be limited to a sub-group (the technologically savvy). Peer filesharing
networks made that system. By requiring at least a computer connection
and Internet access (and optimally broadband access and open-source
know-how), the networks guaranteed that only a certain percentage of
Americans would ever be able to take full advantage of the defection
from the copyright regime.
It is unlikely that the programmers of Napster and other applic ations
actively considered the dynamics of collective action before writing
code. But because filesharing remains confined to a limited group,
filesharers can see that their actions will not eliminate all incentives to
create music or seriously impoverish artists. Users of peer networks are
a select group that could and still do live by slightly different rules.
B. Disorganized Political Action
Even as a sub-group, however, P2P users remain disorganized. The
second reason that P2P was successful was that, as an avoidance mechanism,
it did not require collective action.
That content consumers have not had a strong influence on the shape
of copyright law is well-documented. The lobbying process that led to
the 1976 Copyright Act is a leading example. The Act was the workproduct
of a twenty-one-year-long negotiation between affected industry
Origin of `Digital Divide' Escapes Even the Experts, at http://
www1.soc.american.edu/students/ij/co_3/digitaldivide/history.htm (last visited Feb. 8, 2003)
(on file with the Virginia Law Review Association).
248
See Naťl Telecommunications and Info. Admin., Falling Through The Net II: New
Data On The Digital Divide 4 (July 1998).
249
Id.
2003] Compliance & Code 167
groups.250
Studies suggest, however, that groups representing consumer
interests had little or no influence on the shape of the 1976 Act. Professor
Jessica Litman concluded, "[T]he citizenry's interest in copyright
and copyrighted works was too varied and complex to be amenable to
interest-group championship."251
These studies show what is obvious: For an average consumer, lobbying
for copyright change is expensive, likely futile, and, even if successful,
an impossible change on which to capitalize. As a result, very few
consumers devote themselves to copyright lobbying.
Enter P2P. Individuals who participate in a peer filesharing network
immediately capture for themselves the benefits of their investment.
They save money on the music they download for free, with no need to
share those savings with others who did not participate. Moreover, the
programmers of peer filesharing programs do not, other than sometimes
adhering to a common protocol, even necessarily need to work together
or coordinate their efforts.252
It might be difficult to convince users to contribute, as opposed to
take, from the common pool of shared songs. The process of sharing,
however, has a relatively low cost. Moreover, as Professor Strahilevitz
demonstrates, the design of P2P clients can lead users to believe that
they are participating in a community, triggering norms of reciprocity.253
One salient question is whether the objective of collective action
moves to writing the peer application itself (the Napster program, etc.).
This shift does not seem to occur. First, provided that the program can
be sold, the programmer can appropriate some of the value produced by
the evasion of copyright law and can avoid the collective action problem.
Second, even if the collective action problem persists, the investment
needed to write a peer networking program may be small enough
250
See Litman, supra note 103, at 48­63 (discussing the negotiations behind the 1976
Act); see also Jessica Litman, Copyright Legislation and Technological Change, 68 Or. L.
Rev. 275, 279­82 (1989) (same).
251
Litman, supra note 103, at 52.
252
The creation of the protocols does represent a collective action problem if they are open
(free for anyone to develop around). Interestingly, the major open peer filesharing protocol,
Gnutella, was produced by an open-source programming effort. Open-source programming,
motivated by technological challenges, has proven its ability to create public goods. See Peter
Kollock, The economies of online cooperation: gifts and public goods in cyberspace, in
Communities in Cyberspace 220, 230­35 (Marc A. Smith & Peter Kollock eds., 1999) (examining
the creation of the Linux operating system as an example of a public good created
online despite potential collective action problems).
253
See Strahilevitz, supra note 12, at 547­71.
168 Virginia Law Review Vol. 89:nnn
that the programmer is motivated to write it if for no other reason than
just to serve his own needs. Third, the collaborative structure of opensource
software development may play a role in developing responses
that rely on non-monetary incentives. I will examine each explanation in
turn.
A program is a private good. If it is sold or otherwise used to generate
returns, its developer has the appropriate incentive to respond on behalf
of the group. This incentive is, apparently, what has driven much of the
peer filesharing response so far. For Shawn Fanning, the founder of
Napster, the returns were reputational. As Time remarked, he "reached a
level of fame unprecedented for a 19-year old who is neither a sports
hero nor a pop star."254
But the financial incentives for writing response
programs have not proved overwhelming.255
Most peer filesharing companies
today depend on the dot-com model of deriving revenue from
user traffic. Some developers claim that advertising revenue is enough to
stay in business. For example, the developer of WinMX (yet another
peer filesharing application) stated, "We stay in operation by keeping
our costs low . . . . [W]e think iťs smarter to skip the spyware, generate
revenue from quality ad exposures on www.winmx.com, and spend the
money on important things such as a small yet well rewarded development
team, legal contingency funds, etc."256
More seasoned companies, however, question the advertising model.
KaZaA, for example, depends on selling pop-up ads257
and plans to harness
and sell the unused computing resources of its millions of peered
users (derisively referred to as a "spyware" strategy).258
It has freely
admitted that the online advertising model does not deliver enough revenue
for it to support continued development.259
254
Karl Taro Greenfeld, Meet the Napster, Time, Oct. 2, 2000, at 60.
255
See John Borland, Rocky financial road awaits file swappers, CNET News.com, at
http:news.com.com/2102-1023-273245.html (Sept. 21, 2001) (on file with the Virginia Law
Review Association) (describing the failure of file swapping programs to make any money).
256
WinMX Interview with Kevin Hearn, President, Front Code Technologies, Slyck, at
http://www.slyck.com/newsjuly2002/071002c.html (July 10, 2002) (on file with the Virginia
Law Review Association).
257
See Erick Schonfeld, The True Cost of Free Music, Business 2.0 (May 24, 2002), at
http://www.business2.com/articles/web/print/0,1650,40816,00.html (on file with the Virginia
Law Review) (describing KaZaA's business model).
258
This strategy is referred to as "spyware" because it pretends to be performing one function
while actually performing another. Spyware also sometimes refers to programs that collect
and store information about the user.
259
Id.
2003] Compliance & Code 169
Ironically, this suggests that the continuing development of peer
filesharing may itself depend on copyright law's protection. That is to
say, if other revenue models prove unsuccessful, developers may have to
turn to selling programs or selling membership.260
Their ability to do so
will depend on copyright protection, either against unauthorized distribution
of the software client (perhaps using a peer network) or unauthorized
circumvention of a copy-protection scheme. Peer developers may
have to enlist copyright processes in their effort to evade copyright laws.
They may then, in a further twist, find their tools of copyright evasion
turned against them.
Alternatively, programming a peer response may be inexpensive
enough that some individuals will always be willing to undertake the
project for their own personal benefit. If a college student would otherwise
spend $500 a year on music, and if his time is not otherwise valuable,
he might consider it a worthwhile investment to program an improved
filesharing application. Similarly, it could be that the challenge
of peer networking development will continue to attract the collaborative
attention of open-source developers. How far the open-source
movement will take peer filesharing is an open question--it depends on
how interesting the problem remains to programmers.261
As suggested by the change/avoidance dichotomy in Part I, one of the
reasons for the success of P2P as a mechanism of legal influence is that
it avoids the collective action problem inherent in change mechanisms.
It has worked because certain members of the group have appropriate
incentives to write programs that then lower the cost of the copyright
system for all computer-savvy users. This fact explains the mass popularity
of P2P among disorganized consumers. As a result, Napster and
other programs have become an alternative to political lobbying less by
choice than by default.
260
For example, BearShare, at www.bearshare.com, already sells a "professional" version.
261
Opinions on what motivates open-source programmers vary. See, e.g., Kollock, supra
note 252, at 220­39 (describing a gift model); Eben Moglen, Anarchism Triumphant: Free
Software and the Death of Copyright, 4 First Monday 8, at
http://www.firstmonday.dk/issues/issue4_8/moglen/index.html (Aug. 2, 1999) (on file with
the Virginia Law Review Association) (arguing that economics cannot explain why people
write free software).
170 Virginia Law Review Vol. 89:nnn
CONCLUSION
Finding bold predictions for what political programming projects
means for the future of governance is not difficult to do. John Perry Barlow's
prophecies, for example, have not been understated:
Whaťs happening with global, peer-to-peer networking is not altogether
different from what happened when the American colonists realized
they were poorly served by the British Crown: The colonists
were obliged to cast off that power and develop an economy better
suited to their new environment . . . . No law can be successfully imposed
on a huge population that does not morally support it and possesses
easy means for its invisible evasion.262
My own prophecies are somewhat more modest. The ways groups influence
their government and the effects of its laws are changing. But
the effects are ambiguous. At best, the story suggests that groups that
have never fared well in the political process, due to disorganization or
unpopularity, will gain the most. At worst, already-privileged computer
users will simply find new ways to free-ride.
262
See John Perry Barlow, The Next Economy of Ideas, Wired 8.10 (Oct. 2000).