Shrinking the Commons: Termination of Copyright Licenses and Transfers for the Benefit of the Public

Front matter

Timothy K. Armstrong

abstractFederal law limits the free alienability of copyright rights to prevent powerful transferees from forcing authors into unremunerative bargains. The limiting mechanism is a statutory provision that permits authors or their heirs, at their sole election, to terminate any transfer or license of any copyright interest during a defined period. Indeed, the applicable provisions of the Copyright Act go so far as to invalidate purported waivers by authors of their statutory termination powers.

These statutory provisions may constitute an impediment to the effective grant of rights for the benefit of the public under widely used “open content” licensing arrangements, such as the GNU General Public License (“GPL”) for software or the Creative Commons family of licenses for other sorts of expressive works. Although recent case law suggests that such open-source or open-content licensing arrangements should be analyzed under the same rules that govern other copyright licenses, doing so necessarily raises the possibility of termination of the license. If GPL or Creative Commons-type licenses are subject to later termination by authors (or their heirs), and this termination power cannot validly be waived, then users of such works must confront the possibility that the licenses may be revoked in the future and the works effectively with drawn from public use, with potentially chaotic results.

I. Introduction

Copyrights imperfectly resemble property. Like property, the various rights comprised in a copyright may be conveyed, separately or together, from one owner to another. Unlike transfers involving other forms of both tangible and intellectual property, however, all transfers or licenses of copyright interests by a work’s author are revocable. They may be terminated, during a defined period, at the sole election of the author or the author’s statutory heirs. Further strengthening authorial control, the Copyright Act expressly makes the author’s unilateral power to rescind the transfer irrevocable and nonwaiveable. Thus, copyright rights differ, in a fundamental way, from any other form of property: their initial ownership cannot voluntarily be permanently and unconditionally divested.

According to the legislative history, Congress intended the statutory provisions allowing termination of transfers to protect authors of expressive works from overreaching by powerful licensees, who may effectively pressure authors to make transfers on unremunerative terms. To be sure, examples of such overreaching are not difficult to locate in the cases construing the termination provisions. As between authors and publishers, the latter frequently enjoy superior bargaining power. Importantly, however, although redressing unremunerative bargains and preventing overreaching by licensees supply the underlying rationale for the termination rules, the statute expressly makes all copyright licenses or transfers by the author terminable without regard to the details of the parties’ bargain.

The statute’s termination provisions may pose an underappreciated risk to a wide variety of contemporary “open content” projects, which depend for their vitality upon specialized copyright licenses. From the Linux operating system to the Firefox web browser to the Wikipedia encyclopedia and far beyond, a host of well-known and widely used informational goods have been created by a global community of volunteers. Their efforts are coordinated—indeed, their self-definitions are partly governed—by a family of copyright licenses crafted to permit successive creators to build upon the efforts of their predecessors and to distribute their shared work product. Recent case law suggests that, consistent with their creators’ intent, these licenses are valid under the Copyright Act: they grant rights to users that would otherwise belong exclusively to the owner of the copyright in the

licensed works and impose enforceable conditions upon the exercise of those rights.

The ongoing vitality of “commons-based peer production” of informational goods depends critically upon the assumption that the governing licenses will make earlier users’ contributions freely available in perpetuity for later users to copy, modify, and redistribute. The termination provisions of the Copyright Act, however, may call that assumption into doubt. Nothing more definitive may yet be said, for the peer-production phenomenon is sufficiently recent that issues of termination have not yet arisen. The courts have not considered, for example, whether a computer programmer may release software code to the public under commonly used software licenses such as the GNU General Public License (“GPL”) or the BSD License, and then “terminate” that license many years later and recapture all of the exclusive rights that ordinarily accompany ownership of copyright in a work. Nor has precedent established whether the author of an expressive work published under a Creative Commons license may later terminate the license and sue anyone who distributes or remixes the work (in reliance on the stated terms of the license) for copyright infringement. Nor do we know, to take the most extreme example, whether an author may use the statute’s termination provisions to rescind her own express dedication of a work to the public domain, although countervailing policy considerations may weigh particularly heavily against allowing termination in this instance. The point is simply that the Copyright Act’s termination clauses may, by their terms, reach some or all of these situations, even though none of these scenarios presents the problem of unequal bargaining power that motivated Congress to enact the termination provisions. Particularly when one considers that the Copyright Act allows persons other than the original author to terminate a license in some circumstances, the possibilities for gamesman and abuse of the termination regime in the realm of open-content works appear substantial.

An author’s termination of an open-content license would present a host of practical difficulties. Rapid evolution characterizes many of the works created under such licenses, and untangling a terminating author’s long ago contributions from the contemporary version of a work may represent an exceedingly complex undertaking. In addition, open-content works often include contributions from many authors—thousands, or even millions, in the case of large-scale projects like Linux or Wikipedia—and the task of excising a terminating author’s contributions while simultaneously preserving later users’ contributions would prove particularly vexing.

Even aside from these practical considerations, termination would undoubtedly chill the vibrant creative environment that presently surrounds the development and use of open-content works. It would likely surprise later contributors to the project who had been led to believe that earlier users’ contributions would remain available for reuse in perpetuity (and whose own contributions may well have been made in reliance on that understanding). It would also present a clear affront to the community norms of nonproprietization and mutual sharing that characterize a number of the most vibrant open-content projects.

For all these reasons, the Copyright Act’s termination provisions make a poor fit with open content works. Because those provisions nevertheless appear on their face to permit termination of open-content licenses, clarification or revision of the law is necessary. Although some judge-made doctrines may be pressed into service towards this end, courts may be understandably reluctant to depart from what would appear to be plain statutory text. Therefore, protections against termination for open-content works may find firmer grounding in new legislation.

Part II of this Article will illustrate the variety of copyright licensing arrangements that underlie the peer-production phenomenon. The objective here is not to advance normative claims about the strengths or weaknesses of peer production as a mode of creative endeavor or to contrast it favorably or unfavorably with the more traditional proprietary model of production. This assessment occurs instead at an analytically “lower” level. It takes the peer-production phenomenon as a given and examines the particular licensing arrangements that bind open-content projects and their creators together. Given the dozens, or even hundreds, of model licensing instruments in existence, this inquiry can do no more than summarize a few of the most widely used alternatives, including the GPL and BSD Licenses for software and the GNU Free Documentation License (“GFDL”) and Creative Commons family of licenses for works of other types.

Part III then situates the licensing issue within federal copyright law. The termination clauses represent only one way in which the Copyright Act has evolved to make it more difficult for creators to make irrevocable grants of rights in their works to others. This portion of the Article considers, as illustrative of the complexities introduced by the evolving statute, whether authors may expressly abandon copyright in their works and dedicate them to the public domain—a question that can no longer be answered unequivocally in the affirmative. After surveying the changes Congress has made to vest rights inextricably in the hands of authors, the analysis turns to a consideration of the statute’s complex provisions governing the termination of licenses and transfers. The evolution of these provisions in the legislation that ultimately became the Copyright Act of 1976, and the pattern of case law construing those provisions, both show that Congress’s attention was focused on redressing unremunerative transfers made by authors in the face of superior bargaining power: a policy that, whatever its normative appeal, clearly has no application to the peer-production phenomenon. The need to address the termination issue draws added force from recent cases applying traditional copyright principles to open-content licenses.

With both the technical and legal aspects of the issue on the table, Part IV considers whether existing doctrine is sufficiently flexible to accommodate a judge-made exception to the statutory termination regime for open-content works. A few tools are available to courts inclined to protect open-content licensees from the prospect of termination by the licensor. The courts might, for example, modify the existing doctrine of copyright abandonment to permit partial, conditional, irrevocable abandonments of copyright rights, to the permanent detriment of authors and their heirs. Alternatively, the courts might extend their current practice of borrowing provisions from the Patent Act when construing the Copyright Act. The Patent Act presently includes several provisions that permit creators expressly to abandon rights in their creations and dedicate their works permanently to the public domain. Courts might interpret the Copyright Act harmoniously with these provisions of the Patent Act as a way of sidestepping the problems that termination of an open-content copyright license would pose. Neither of these solutions, however, is optimal; either would require a reviewing court to disregard seemingly forceful counterarguments, and a court’s receptivity to arguments grounded either in a reformulated abandonment doctrine or in the Patent Act cannot be known with certainty in advance. Such unpredictability, however, only multiplies the present uncertainty surrounding the durability and permanence of open-content licenses.

With no compelling judicial alternatives in view, Part V considers legislative solutions. Congress should recognize the peer production phenomenon as an essential change in the baseline assumptions that prevailed at the time of the Copyright Act of 1976—a change that demands its own recognition in positive law. The better approach, accordingly, would be for Con to amend the statute to place open-content licenses outside the reach of the Copyright Act’s termination clauses.

II. Open-Content Licensing

A. The Public Benefit of Open-Content Licenses

In recent years, a vast body of content has been created and distributed under licenses that grant unspecified members of the public rights that, by statute, ordinarily belong to the copyright holder alone. Before discussing the specifics of a few such licenses, it is useful to identify some important commonalities that most open-content licenses share.

First, and most fundamentally, all open-content licenses authorize otherwise unlawful conduct: that is, they expressly permit users of the licensed works to perform actions that would otherwise infringe the licensor’s copyright. Absent authorization (or other legal excuse such as fair or de minimis use), reproducing, distributing, or modifying a work infringes the author’s copyright. Open-content licenses make such otherwise infringing activities lawful and thereby facilitate uses of works that federal law would ordinarily prohibit.

Second, all the open-content licenses considered herein are universally available: the works are offered to all on equivalent terms without the need for individualized bargaining. This is not, of course, to deny the possibility of dual licensing: a prospective licensee dissatisfied with some of the conditions of an open-content license may negotiate with the work’s author for a license on different terms. Absent dual licensing, however, a licensee who is willing to observe the conditions stated in an open-content license is free to use the work without the particularized approval of (and indeed, without notice to) the licensor. An unavoidable consequence is that an open-content licensor ordinarily has no way of knowing the identities of all the licensees, for the parties to the license may never communicate.

Third, the universal availability of the content on the terms stated in the license minimizes transaction costs. Open-content licenses are available “off the shelf” to accommodate several types of concerns that may confront users of the licensed works. Use of the works in a fashion consistent with the license entails no greater burden than reading the license; there is no need to incur the expense of negotiating an individual usage agreement (again, assuming that the licensee is satisfied with the conditions stated in the license). Even the burden of reading the license, moreover, is minimized, because a relatively small number of standard-form licenses govern a wide variety of projects available from different open-content suppliers. There is, correspondingly, a diminished need to become familiar with particularized license terms that may vary from one vendor to the next outside the open-content domain.

Finally, combining the foregoing characteristics yields the defining feature of all open-content licenses: they promote freedom. Open-content licenses expressly authorize a large community of users, at their sole election and without further negotiation or expense, to use a variety of works in a manner that would otherwise infringe copyright. In doing so, the licenses facilitate at least some uses that would not otherwise occur. The licenses create, in economic terms, a commons: a pool of raw materials upon which members of the public are free to draw for their own consumptive or productive ends.

The public domain presents a possibly more familiar model of such a commons. Any creator is free to draw from public-domain materials and to incorporate them into her own creation. A growing literature recognizes the

beneficial effects on creative production from the existence and availability of freely usable public-domain works. By reducing the costs creators must pay to reuse others’ creations in their own, a rich commons pays public dividends in the form of greater production of expressive works. As discussed below, however, changes to U.S. copyright law over the past three decades have tended to restrict the entry of new works into the public domain. If a robust public domain fuels creativity, the strengthening of private property rights may have the opposite effect—they may increase authors’ costs and deter future creativity, creating what Michael Heller calls “the tragedy of the anticommons.” The parallel rise of the open-content movement, which leverages existing copyright and contract principles to create a new commons outside the public domain, might be seen as a reaction to the excessive expansion of property rights into fields of intellectual and creative

endeavor, and a conscious effort to provide an alternative supply of expressive resources for future authors.

The public derives at least two identifiable benefits from works produced under open-content licenses. First, perhaps most obviously, open-content works may be consumed without fear of liability. Individuals may freely use software, or listen to (or share) music issued under open-content licenses (assuming, as always, that any applicable conditions in the accompanying licenses are observed). This is a particularly valuable benefit insofar as it aligns with internet-user norms, which have favored the sharing and reuse of works even in the face of legal threats. Second, making works available for use under open-content licenses has a multiplier effect, permitting the creation of new works which may never have come into existence absent the raw materials that the licenses place into the commons. These benefits to the public, in turn, help to justify legal protections for the creators and users of open content works against the uncertainty of termination.

B. Open-Content Licensing: Software

The contemporary peer-production phenomenon began in the software world in the 1970s and 1980s, with later open-content projects modeling themselves on the paradigm established by free and open source software (“FOSS”). Software licensing, accordingly, supplies the logical starting point for an examination of the legal underpinnings of commons-based peer production. Although dozens of standard-form licenses exist for FOSS works, many of them share a common intellectual ancestor: the GNU Project’s celebrated GPL.

1. The GNU GPL and LGPL

In the FOSS world, the GNU GPL is the most widely used license. There have been three versions of the GPL: Version 1, which was promul in 1989 (“GPLv1”); Version 2, in 1991 (“GPLv2”); and Version 3, in 2007 (“GPLv3”). The various versions of the GPL do not displace or supersede one another, and authors of FOSS works may select earlier versions of the GPL if those versions better reflect their intent.

Although a FOSS author’s use of the GPL effectively places the software into the commons, thereby making it available for copying and modification, GPL-licensed software is not in the public domain. Rather, all GPL-licensed software is copyrighted. The Preamble of GPLv3 includes

the following clause (versions of which also appeared in GPLv1 and GPLv2) explaining the function of the license: “Developers that use the GNU GPL protect your rights with two steps: (1) assert copyright on the software, and (2) offer you this License giving you legal permission to copy, distribute and/or modify it.”

Every version of the GPL grants users of the licensed software the rights to engage in the otherwise-infringing acts of copying, modifying, and redistributing the licensed work. Like all open-content licenses, however, the GPL attaches conditions that limit the scope of the authorization the license confers. A licensee’s exercise of one of the rights granted by the license (to copy, modify, or redistribute the work) without observing the applicable conditions exceeds the licensee’s authority, and renders that use unauthorized (and, thus, infringing).

Many of the GPL’s conditions exist to inform downstream users about the legal status of the work and to provide them with the resources necessary to execute the rights granted under the license. To ensure that licensees are able to exercise their power to alter the software, the GPLv3 mandates that the software’s source code also be made available whenever the software is distributed in executable or binary form.

The so called “copyleft” condition, a provision included in one form or another since GPLv1, limits the ability of authors of works derived from GPL-licensed software to release their own work under licenses more restrictive than the GPL. In its current form, the provision requires authors of a modified version of software derived from GPL-licensed code to “license the entire work, as a whole, under this License to anyone who comes into possession of a copy. This License will therefore apply . . . to the whole of the work, and all its parts, regardless of how they are packaged.”

In a license document that otherwise celebrates freedom and choice, the copyleft condition at first seems out of place. It limits, rather than enhances, one potentially important freedom enjoyed by users of the software—the freedom to license derivative software works they create on terms of their choosing. It is surely this loss of freedom that underlies pejorative labels, such as “viral” or “infectious,” that critics sometimes use in describing the GPL or other copyleft licenses. As a programmer, using GPL-licensed code in your own work makes you a FOSS author whether you wish to be or not, because the copyleft condition “infects” your final product, requiring it to be licensed under the GPL as well if it is distributed to others.

Yet just as the law celebrates “those wise restraints that make us free,” the GPL enhances far more freedom than it limits. The copyleft condition feeds the commons. It ensures that FOSS works remain free even as they evolve and protects earlier programmers against later programmers’ proprietary appropriation of their work. Because the copyleft condition applies only to software derived from GPL-licensed code, it is sometimes labeled a “reciprocity” provision. In other words, the cost of reusing GPL-licensed code in one’s own work is that one must either offer that work to all under the GPL, or offer it to no one.

Although the GPL also includes what is labeled a “termination” provision, it is important not to confuse this clause with the issue under consideration herein. The GPL provides, in essence, that if a licensee fails to honor the conditions stated in the license (and fails thereafter to cure her default), that licensee may not thereafter rely on the GPL as a defense to the original licensor’s claim of copyright infringement. In the language of the license, the rights of a defaulting user are said to “terminate.” Stated differently, a user may, by her conduct, forfeit the rights the GPL grants. This is a different issue from the question that this Article considers, namely, whether the Copyright Act permits unilateral termination of open content licenses by the author (or the author’s heirs) irrespective of any breach by the licensee.

Unless the licensee forfeits her rights by breaching a condition of the GPL, however, the license is expressly declared to be perpetual: “All rights granted under this License are granted for the term of copyright on the Pro , and are irrevocable provided the stated conditions are met.” This guarantee makes explicit a promise of permanence that has always been implicit in the peer-production model: it “assur[es] . . . developers that their work could never be taken away from them,” but will remain in the commons in perpetuity. It is precisely this assurance that the Copyright Act’s termination provisions may place in doubt.

A GPL variant used for some FOSS works, the GNU Lesser General Public License (“LGPL”), incorporates most of the provisions of the GPL. The key difference is that the LGPL relaxes the GPL’s copyleft condition as applied to application programs dynamically linked to LGPL-licensed library code. This waiver from the GPL’s strict copyleft condition is believed to encourage the commercial use of LGPL-licensed content.

2. BSD Licenses

In the late 1970s, a group of computer scientists and graduate students at the University of California–Berkeley created the Berkeley Software Distribution (“BSD”), a small collection of software tools they had written for the Unix operating system. Over time, BSD expanded to become a full-fledged Unix-compatible operating system in its own right. BSD was issued under the BSD License, which lent its name to a family of software licenses collectively known today as “BSD-style” licenses.

BSD-style licenses are models of brevity compared with the GPL. They mandate an express copyright notice acknowledging the owner and year of the work and authorize “[r]edistribution and use” of the licensed software “in source and binary forms, with or without modification[.]” This authorization is subject to only three conditions:

for software works distributed in the form of source code, the distribution must include the copyright notice, the listing of conditions from the license template, and a paragraph disclaiming warranties or liability based in contract or tort;

for software works distributed in the form of object code, those same items must be included in the accompanying documentation; and

for either type of work, the name of the owner’s organization or its contributors may not be employed in a manner that suggests endorsement or promotion of products derived from the licensed software.

Because they require crediting the owner but omit other limitations on the copying, modification, or redistribution of the licensed works, BSD-style licenses are often characterized (with tolerable, if imperfect, accuracy) as “attribution-only” licenses. What the BSD-style licenses omit, moreover, is at least as important as what they require. BSD licenses are not reciprocal licenses and have no “copyleft” condition. Nor do they require source code to be distributed with the software. Because derivatives of BSD-licensed works need not be licensed under a BSD-style license (or any other recognized FOSS license) and need not include source code, BSD-derived code may be freely incorporated into proprietary software works distributed in binary-only form.

3. Open-Source Licensing: Constraint and Liberation

The GPL, the LGPL, and BSD-style licenses serve as useful examples from the broader universe of FOSS licenses because of their very different presumptions and effects. They represent points along a continuum which can be viewed from either of two directions.

The licenses might first be characterized according to the constraining force they exert on downstream users of the licensed works. The BSD license is minimally constraining; beyond the obligations to recognize the original author’s copyright and to refrain from implying endorsement or pro- motion, the license places no limitations on a user’s copying, modification, or distribution of the work, including for profit. At the opposite extreme, the GPL is maximally constraining, regulating not only how a work may be

reused, but also how derivatives of the work may be licensed. The GPL thus effectively forbids proprietary or commercial adaptations of GPL-licensed code. The LGPL falls somewhere between, but still effectively permits commercial reuse via linking to an LGPL-licensed library. These three paradigms raise essential questions, such as whether or not attribution should be required, whether derivative works should be required to be licensed under similar terms, and whether commercial reuse of the licensed content ought to be allowed. These questions recur in open-content licensing even outside the software context. The issue is thus not whether to constrain users’ behavior, for all licenses constrain (even if only by maintaining extrinsic constraints imposed by copyright law), but rather which constraints a licensor finds best to effectuate her intent.

The same group of licenses might also be evaluated not for how they affect individual users, but for how they affect the commons. BSD-style licenses are essentially agnostic on the question of growing the commons. Developers may add software to the pool, but anyone else is free to incorporate the licensed content into a proprietary product and need not make their own contributions available for others to reuse. In contrast, the GPL’s reciprocity provision demands that what is borrowed from the commons be returned to the commons: modified versions of GPL-licensed software may be distributed only under the GPL (with source code) for universal copying and modification. Again, the LGPL stands in between, excusing compliance with the GPL’s copyleft condition in some circumstances.

Comparing the licenses’ effects along these two dimensions illuminates the underlying bias of contemporary copyright law in favor of proprietary production. Creating a commons of freely reusable works can be accomplished only with licenses that constrain the commons-defeating features of current law. The default positions of copyright law tend to reflect assumptions that emphasize individual production under a strongly proprietary regime. Those proprietary defaults have only strengthened over time. FOSS licenses inevitably reflect that underlying copyright regime even as they work to construct an alternative. The GPL, for example, aims to make FOSS works freely reusable by restricting behaviors (such as releasing software in binary-only form or incorporating it into proprietary products) that, although not forbidden by copyright law, inhibit the growth of the commons. The GPL backs its commons-creating mandates with the threat of liability for copyright infringement if they are breached, thereby cleverly inverting the proprietary architecture of copyright to preserve a domain free from proprietary control. BSD-style licenses, in contrast, impose fewer constraints on users because licensors who select that license are indifferent as to whether commons-defeating behaviors permitted under current copyright law occur.

There is, in short, an inverse relationship between allowing the exercise of proprietary control over software works and the growth of a freely reusable commons of such works. The GPL, LGPL, and BSD licenses mark several of the possible points along a continuum at which the tradeoff between those two objectives might be effectuated. If FOSS licenses are subject to the Copyright Act’s termination provisions, however, all of those tradeoffs become conditional, contingent, and uncertain. By strengthening the commons-defeating features of copyright’s statutory defaults, the Copyright Act’s termination provisions may pose particular risks to the growth of a software commons.

C. Open-Content Licensing: Non-Software Works

Nothing in the logic of FOSS licensing necessarily confines the creation of an open-content commons to the realm of software, and a number of projects have sought to extend the creative dynamism of the FOSS movement to other productive domains. An overview of some of the commonly used open-content licenses for non-software works reveals both similarities and differences with the software-licensing regime previously discussed. On the one hand, other varieties of open-content licenses must grapple with some of the same tradeoffs that have contributed to the proliferation of FOSS licenses. Issues surrounding attribution, commercial reuse, and reciprocity loom equally large outside the software domain, and diverse licensors might see the issues differently depending upon their attitudes toward the importance of a growing commons. On the other hand, open-content licensing arrangements have become more sophisticated since the earliest versions of the GPL and, as discussed below, one family of model licenses (the Creative Commons licenses) now offers licensors the easy ability to convert their particular policy preferences into specific license terms with almost arbitrary granularity.

1. The GFDL

The GNU Project made one of the earliest attempts to apply the commons-enhancing principles of the GPL outside the context of computer software. The result was the GNU Free Documentation License (“GFDL”), which was promulgated in 2000. An amended GFDL followed in 2002, which added several new definitional provisions to the license and made a few changes to the language of the license conditions. In 2008, the GFDL was again revised, primarily to expand the license’s termination (or forfeiture) provision and to add new language permitting the relicensing of GFDL-licensed works.

The GFDL was written with software documentation in mind, although by its terms the GFDL may be used for any text. The GFDL, like the GPL, grants users the rights to engage in acts that would infringe the author’s copyright absent a license, including the rights to copy, modify, and redistribute the licensed work. Unlike the GPL, the GFDL expressly permits copying for commercial purposes. Like the GPL and other open-content licenses, however, the GFDL attaches conditions to the rights granted.

First, the GFDL is a reciprocal (or “copyleft”) license. Any modified version of a GFDL-licensed work that is distributed to others must itself be released under the GFDL. In addition, the GFDL mandates attribution and the inclusion of an express copyright notice, requires modified works to adopt a new title, forbids certain other modifications of the licensed content, and empowers authors also to forbid modifications of designated sections of the work, thus requiring the specified portions of the work to be reproduced verbatim in any subsequent version. Finally, like many other open-content licenses, the GFDL includes provisions that essentially require users to be informed about their rights. Any derivatives of GFDL-licensed content must include an express public notice alerting users to their GFDL- based permissions to reuse the work and a copy of the GFDL.