Julie E. Cohen*
68 S. Cal. L. Rev. 1091 (1995)
Copyright 1995 - HTML Version Prepared by the LII with Permission
As Sega and Atari illustrate, any debate over permissible uses of knowledge gained through decompilation becomes purely academic if decompilation is not itself permissible. This part evaluates the Sega court's resolution of that question. n76 Although many commentators [p*1105] have praised the Sega decision as forward-thinking, n77 the fair use analysis adopted by the Sega court also has drawn some high-powered criticism. Most notably, in his recent comprehensive review of computer copyright law, Professor Arthur Miller assails the Ninth Circuit's application of the fair use doctrine as misguided and "singularly ill-suited to vindicating the public interest." n78 Even a recent student note by an unabashed fan of thin copyright protection for software interface specifications finds the court's analysis "strained." n79 This reception doubtless would come as no surprise to the Sega court, which acknowledged that the result it reached - allowing "wholesale copying" by a competitor intent on producing a competing product - "may seem incongruous at first blush." n80 Are the critics' reactions warranted? Careful consideration of the nature of computer programs and the patterns of innovation and dissemination of new developments within the computer industry suggests that they are not. Sega is faithful to both the letter and the spirit of the copyright laws.
In determining whether Accolade's copying was a fair use, the Sega court engaged in a lengthy analysis of the four factors enumerated in the fair use provision, section 107 of the Copyright Act. n81 The [p*1106] application of these factors to computer programs raises several novel questions, and brings other unresolved issues concerning the scope of the fair use doctrine into sharp focus. Ultimately, the answers to these questions turn on, and require decisions about, the purpose and role of fair use in the overall scheme of copyright protection. This part analyzes the individual statutory fair use factors as they relate to lock-out, and then considers the implications of the decompilation debate, and the Sega court's resolution of it, for an overarching vision of fair use.
Conceptually, the Sega court's analysis began and ended with the second statutory factor: the nature of the copyrighted work. The court observed that when computer programs are distributed in object code form, the only means of access to their unprotected functional features, even for trained programmers, necessarily involves preparing human-readable copies or derivative works. n82 Accordingly, core principles of copyright law would seem to require that reverse engineers be allowed to keep records of their progress; otherwise, "the owner of the copyright gains a de facto monopoly over the functional aspects of his work - aspects that were expressly denied copyright protection by Congress." n83
[p*1107]
1. "One of These Things Is Not Like the Others": Computer Programs as Literary Works
Professor Miller's objection to the Sega court's analysis of the second statutory factor is that computer programs are "literary works" under the Copyright Act and therefore should be treated no differently from other literary works for fair use purposes. n84 That is, intermediate copying of a computer program's creative content - an inevitable consequence of decompilation because protected and unprotected portions cannot be distinguished until they have been translated into human-readable form - should be prohibited, because such copying would not be allowed for other literary works. The assumption implicit in this argument - that intermediate copying of a traditional literary work's creative content can never be a fair use - is addressed below in the discussion of the first statutory factor. n85 As to the second statutory factor, the objection that computer programs are classified as literary works, while accurate as a statement of positive law, n86 is so broad as to be virtually meaningless as a guide for courts struggling to apply section 107 in the computer software context. To the extent that generalizations about the nature of "literary works" are possible, however, what they reveal is that the statutory classification of computer programs as literary works confuses more often than it clarifies.
First, the classification of computer programs as "literary works" is staggeringly uninformative. As defined by the Copyright Act, "literary works" include all works "expressed in words, numbers, or other verbal or numerical symbols or indicia" n87 - in other words, not only novels and essays, but also textbooks, reference works, directories, greeting cards, and everything in between. The proportion of creative, protectable expression in these works varies enormously. n88 Thus, to state that a computer program is, legislatively speaking, a "literary work" proves nothing about the scope of the protection courts should afford it. The copyright protection for which the work is eligible is a function of the work's relative proportions of creative and noncreative content.
[p*1108]
Assuming for the moment that computer programs are properly viewed as literary works, n89 then to which types of literary work should computer programs be compared? Professor Miller acknowledges that "the scope of protection given to different types of literary works may vary." n90 Yet he consistently compares computer programs to works of literature such as "Steinbeck's [The] Grapes of Wrath, Hemingway's The Sun Also Rises, or Miller's Death of a Salesman" without once explaining why they should not instead (or also) be compared to the Physician's Desk Reference or the Pacific Bell Yellow Pages. n91 The parallels between computer programs and literary classics are far from obvious. A computer program is, first and foremost, a series of instructions to the computer to execute a given task. n92 The instructions themselves may be written or arranged with more or less creativity, but that is not their primary significance. In this respect, a successful program is more analogous to a well-designed, easy-to-use directory of information, or to a cookbook, than to a novel or a play. Given this defining characteristic of computer programs, there is no logical basis for Professor Miller's conclusion that the scope of protection afforded computer programs under Sega differs from that afforded other literary works not only in scope, but in kind. n93 Traditional literary works exist on a continuum of protection; if computer programs are best characterized as literary works, it certainly would be reasonable to conclude that computer programs constitute a new endpoint on that continuum.
A far more reasonable conclusion, however, is that computer programs do not lie on the literary works continuum at all. Even among highly utilitarian literary works, the barriers to access created by distribution of computer programs in object code form have no analogue. n94 Thus, it is by no means obvious that computer programs can [p*1109] or should be compared with literary works rather than with some other category of copyrightable works or viewed as sui generis in many critical respects. n95 As Professor Miller's commentary illustrates, treating computer programs as literary works too easily complicates the task of determining the scope of software copyright by importing into the analysis preconceptions of marginal relevance. The Sega court, in contrast, treated computer programs simply as "utilitarian works," and so avoided that pitfall. n96 Arguably, one of the lessons of Sega is that the classification of computer programs as literary works is inappropriate and breeds confusion.
The rationale for the statutory classification of computer programs as literary works, which originated in the Copyright Act of 1976, is unclear. Apparently, neither Congress nor CONTU n97 deemed it worthy of discussion. n98 It appears that both Congress and CONTU simply concluded that because they are written or typed (as opposed to sculpted, drawn, or rendered in musical notes), computer programs are more similar to literary works than to works in the other categories of copyrightable works listed in section 102(a) of the Act. Both legally and factually, that conclusion is dubious.
[p*1110]
As a matter of copyright law, denominating computer programs "writings" tells us nothing about how to categorize them, because constitutionally every work protected by the copyright laws is a "writing." n99 As a practical matter, a rule that all works expressed in "numerical symbols or indicia" are "literary works" also encompasses audiovisual and musical works created and expressed digitally. n100 In terms of use, computer programs also exhibit similarities to works in several of the other statutory categories. To the extent that computer programs constitute a script for the computer to follow, they may be argued to resemble "dramatic works" that are "performed" by the computer. n101 In other ways, computer programs are analogous to "pictorial, graphic ... or sculptural works," n102 in that they constitute a map or set of blueprints for accomplishing a task. n103
Comparison to other statutory categories is more than an exercise in semantics. Each change in the statutory reference point conjures up a slightly different body of precedent and different variations on the basic approach to identifying what the copyright in the work protects. n104 The difficulty of selecting the statutory category of protected works to which computer programs are most analogous, and of finding a good fit in any category, suggests that computer programs may be most appropriately regarded as sui generis forms of creative expression.
That conclusion is not new; the unique nature of computer programs has long been a rallying cry for advocates of a wholly sui generis system of intellectual property protection. n105 My intent here is more modest: at minimum, sections 101 and 102(a) of the Copyright Act should be amended to ensure that computer programs are properly viewed as unique - a ninth category of copyrightable works. As the debate over the Sega decision illustrates, the epistemological consequences of the current classification of computer programs are not [p*1111] trivial. Part III demonstrates that those consequences become even more significant during evaluation of the alleged copier's final product for substantial similarity to the copyrighted work. By signaling courts to abandon preconceptions about "literary works" and to adopt a more flexible, open-minded approach to computer copyright cases, an amendment acknowledging sui generis status would encourage more thoughtful decisionmaking regarding the scope of copyright protection available.
2. Defining "Publication" in the Context of Machine-Readable Works
For fair use purposes, the nature of the copied work is determined in part by whether it was published or unpublished when the copying occurred. n106 Traditionally, courts have accorded unpublished works much greater protection and have been less willing to treat copying of such works as fair use. n107 In an effort to turn the characteristics of object code to its advantage, Sega argued that its programs should be considered unpublished because they were distributed for public use only in object code form. The Ninth Circuit summarily rejected that argument, n108 but it deserves more than summary treatment. Whether and when computer programs distributed in object code form become published works for purposes of copyright is a question of great significance in assessing the level of protection that the Copyright Act affords them. Once again, the search for answers suggests that a traditional copyright concept developed in the context of artistic and literary works - here, publication - is unhelpful when analyzing computer programs.
[p*1112]
The rationale for concluding that Sega's program was a published work was that once a program is distributed for public use it is published for purposes of copyright. n109 At first glance, classification of computer programs distributed for public use in object code form as "published" works is entirely consistent with the treatment of other machine-readable works under the Act. For example, musical works on record, compact disc, or cassette are deemed published when sold, even though the works cannot be played without stereo equipment. n110 However, musical works distributed in machine-readable form and computer programs differ in one significant respect. Playing a machine-readable musical work discloses its substance, while using a computer program need not. Because computer programs are functional rather than artistic works, they may be distributed to and used by the public without disclosing the manner in which they are written or the methods by which they operate. Conversely, computer programs in human-readable form cannot perform the functions they describe; thus, there would be no consumer market for them.
Professor Miller suggests that Congress was aware of the peculiar problem posed by computer programs when it amended the fair use statute in 1992 to state that the fact that a work is unpublished will not automatically preclude a finding of fair use. n111 The bill's sponsor, Senator Simon, noted that the amendment was "not intended to provide new fair use access" through decompilation, nor to "broaden the fair use of unpublished computer programs." n112 These statements are less significant than Professor Miller makes them seem. As is its wont when considering section 107, Congress took a cautious approach to assessing the current state of the law. It stated only that it did not intend to "alter" fair use access to unpublished works; it did not attempt to state the current rule or dictate what it should be. n113 Moreover, nowhere did Senator Simon, or anyone else, suggest that [p*1113] publicly distributed object-coded computer programs are "unpublished." n114 Thus, the legislative history simply returns us to the initial problem.
A rule that public distribution in any form constitutes publication makes sense given the rationale for heightened protection for unpublished works. The unpublished work doctrine protects the author's right of creative control. n115 Allowing the author to determine when a work is ready for release also protects the public, by assuring sufficient time to polish the work to the author's standards. n116 A commercial (or not-for-profit) distribution of the work signifies a decision that the work has met the author's standards and is ready for release. The greater protection accorded to unpublished works also allows the author to reap the first commercial benefits from distribution of the work (or to elect to forgo those profits for not-for-profit distribution). n117 All of these rationales seem to apply with equal force to computer programs. It might be argued that computer programs are different from most other copyrighted works that are distributed to the public, in that versions released to customers often are subject to continuing upgrades and other revisions, both as the "author" deems necessary and in response to customer complaints and requests. However, the mere fact that the programmer may have an ongoing relationship with the program after its release should not call into question the program's "published" status. Many textbooks, casebooks, and treatises also are updated on an ongoing basis, without thereby losing their unquestioned status as published works. And the rationale for considering such works published applies even so, because the author's initial decision to release the work, and any economic benefit gained thereby, cannot be changed by later events.
Finally, a "public distribution equals publication" rule also is consistent with other aspects of the copyright treatment of computer programs. The bare fact that the copyright afforded a program extends to [p*1114] the object code mandates that public distribution of a program in object code form be considered a publication of the program. n118 Any other rule would, in effect, confer heightened protection on object code - an incorrect result, given that copyright protection only extends to the zeros and ones of object code because they are derived from the human expression contained in the original source code. n119 The Copyright Act should not be used to bootstrap de facto trade secret protection for publicly distributed works.
Still unaddressed by the foregoing discussion, however, is what "public distribution" means. Both Sega and the case on which it relied, Lewis Galoob Toys, Inc. v. Nintendo of America, Inc., n120 involved programs distributed to retail customers. n121 As a result, neither court considered the variety of other ways in which computer programs are distributed and the application of section 107 to those programs. n122 Whether the limited decompilation privilege established in Sega applies to programs not distributed directly to the general public is a more difficult question. It is conceivable, for example, that a program with limited distribution to a small number of licensees, subject to contractual restrictions on disclosure, could be considered unpublished. Referring back to the purposes of the unpublished work doctrine, however, the reasons that programs available for retail purchase should be considered published works apply with equal force to programs distributed on a more limited basis. By definition, any distribution to customers or distributors, however small, still reflects the author's choice and the author's decision that the program is suitable for release. Extending the Lewis Galoob Toys ruling to any distribution of a copyrighted computer program thus would preserve both the author's right of control and the public's interest. By the same token, that reasoning would not apply to releases known as "beta test copies," which are distributed on a trial basis with the understanding that they are unfinished, prerelease products. n123
[p*1115]
As a practical matter, however, limited distribution programs typically are licensed rather than sold, n124 and the price of licensing a proprietary program usually includes an agreement not to reverse engineer the program. A question that courts and litigants eventually must confront is whether a contractual restriction on reverse engineering is valid. Contracts that alter the existing balance of common law property rights are commonplace, and license agreements for proprietary computer programs that alter the balance of rights established by the Copyright Act follow in that tradition. Copyright's debt to common law property rights and the theories of ownership in which they are rooted is significant. n125 However, modern-day copyright is substantially a creature of public policy. Arguably, to the extent that private contracts frustrate that policy - for example, by divesting licensees of a right of access to unprotectable information - they are unenforceable. n126 The courts have yet to resolve this question. Their answer will determine whether the reverse engineering privilege established in Sega applies to all computer programs, or only to some.
The first statutory fair use factor is the purpose and character of the use, "including whether such use is of a commercial nature or is for nonprofit educational purposes." n127 Accolade is a commercial actor, and its ultimate purpose in copying Sega's code was unquestionably commercial. For the district court in Sega, the fair use analysis began and ended there. n128 The Ninth Circuit rejected the district court's bright-line approach to the purpose and character test in favor [p*1116] of a detailed, fact-specific analysis. n129 Ultimately, the court concluded that this statutory factor weighed in Accolade's favor, because Accolade had copied Sega's microcode solely in order to study its unprotected elements. n130
The Ninth Circuit's more circumspect approach to the purpose and character inquiry has since been squarely vindicated. In Campbell v. Acuff-Rose Music, Inc., n131 its first fair use opinion in nine years, the Supreme Court warned against "elevating commerciality to hard presumptive significance," and cited Sega with approval. n132 The Court's substantive analysis of the purpose and character test also tends to support the Ninth Circuit's conclusion that Accolade's purpose in decompiling Sega's copyrighted code was "legitimate [and] essentially non-exploitative." n133 Together, Acuff-Rose and Sega suggest a conception of the first statutory fair use factor that is less rigid than the simplistic commercial/noncommercial distinction and far better suited to identifying permissible uses of copyrighted material within the predominantly commercial field of computer programming.
The district court in Sega based its approach to the purpose and character inquiry on the Supreme Court's opinions in Sony Corp. v. Universal City Studios, Inc. n134 and Harper & Row Publishers, Inc. v. Nation Enterprises. n135 In Sony, the first of the two decisions, the Court remarked that "every commercial use of copyrighted material is presumptively an unfair exploitation of the monopoly privilege that belongs to the owner of the copyright." n136 This statement is noteworthy for two reasons. First, because Sony did not involve a commercial use of copyrighted material, the Court's remark was dictum. Second, the Court cited no authority whatsoever for the sweeping proposition that every commercial use is presumptively unfair. Indeed, section [p*1117] 107 itself is to the contrary; several of the activities listed in its preamble as examples of fair use - news reporting, comment, and criticism - are generally viewed as commercial endeavors. n137 The following year, in Harper & Row, which did involve a commercial use of copyrighted material, the Court invoked the Sony dictum to support its finding that the challenged use was unfair. n138 Invented out of whole cloth in Sony and then cited and reinforced in Harper & Row, the Court's sweeping generalization about an entire class of uses became a lodestar of virtually every fair use decision handed down by the lower courts over the next decade, Sega included. n139
Copyright scholars were nearly unanimous in criticizing the commercial/noncommercial distinction as both simplistic and inherently ambiguous. n140 Acuff-Rose signals the Court's response to a decade's worth of criticism: a full-blown retreat. Writing for the Court, Justice Souter went to great lengths to characterize his analysis of the first fair use factor as entirely consistent with Sony and Harper & Row. n141 However, the Acuff-Rose opinion owes far more to the dissents in those cases and to the scholarly criticism that followed them. In noting the commercial nature of news reporting, criticism, and other fair [p*1118] uses enumerated in section 107, Justice Souter relied on Justice Brennan's dissent in Harper & Row. n142 Following Justice Brennan's lead, he expressly acknowledged that whether a use is, broadly speaking, "commercial" in nature is not the sole determinant of its purpose and character. The new standard the Court set forth for evaluating purpose and character derives from the writings of Judge Pierre Leval, one of the fair use doctrine's most thoughtful critics. The Court observed that the statutory distinction between "commercial" and "noncommercial" uses is, to a considerable degree, intended as shorthand for uses that do or do not promote the purposes of copyright. Borrowing Judge Leval's terminology, it reasoned that "the more transformative the new work, the less will be the significance of other factors, like commercialism, that may weigh against a finding of fair use." n143
Acuff-Rose's treatment of the first statutory factor signals a sea change in the jurisprudence of fair use. However, the copying in Sega raises issues that Acuff-Rose did not address. First, there is a far more complex relationship between commercial activity and innovation in the realm of creative expression than most courts have yet acknowledged. For some types of copyrightable works, including computer programs, creativity and commercial endeavor are inextricably intertwined. Developing computer programs is expensive. Research and development efforts may require significant investments of equipment, personnel, and time. Those costs can only increase when product development must be expedited to beat a close competitor to the market. As a result, many of the most creative computer programming innovations come from the corporate sector. n144 In recent years, research consortia, "technology transfer" programs, and other joint ventures sponsored by corporate investors have become the preferred [p*1119] methods of innovation. n145 In this environment, the commercial purpose and character test is more than inapt. Applied without an understanding of the unique constraints inherent in a form of creative expression that requires a research and development budget, the test threatens to remove the protection of fair use from an entire class of copyrightable works, and so undermine incentives for further innovation. In short, the first statutory factor cannot be applied to all types of copyrightable works in the same way. In Sega, the Ninth Circuit implicitly acknowledged this. n146 With respect to the first statutory fair use factor, the first lesson of Sega is that a fair use analysis must take into account the mechanisms by which new works of a particular type are ordinarily created.
Second, the "particular use" challenged by Sega was an intermediate, not an ultimate, one - copying as an essential but preliminary step to developing a competing but hopefully noninfringing product. n147 While Accolade's ultimate purpose was unquestionably commercial, its intermediate purpose was to gain knowledge and understanding of certain functional principles. n148 The case thus required the court to answer a novel question: To which of Accolade's purposes does the first statutory factor refer - or, can the intermediate step of copying solely to gain understanding be viewed as a fair use? Doctrinally speaking, it is in this respect that Sega was a case of first impression. While other cases had considered whether intermediate copying is an infringement, no previous case had considered the fair use defense in the context of intermediate copying. n149 The Ninth Circuit concluded, largely without discussion, that Accolade's immediate purpose was dispositive, rather than its ultimate, unquestionably commercial one. n150 Professor Miller, in contrast, focuses entirely on Accolade's long-term commercial goal of competing with Sega in the market for Genesis-compatible games. He argues that the copier's [p*1120] long range commercial goals should determine the outcome, even if the final product is noninfringing. n151 Yet his consideration of Accolade's intermediate purpose is as cursory as the Ninth Circuit's consideration of its ultimate purpose.
The language of section 107 is instructive in this regard. The first statutory fair use factor seems to require only that a court evaluate the purpose and character of the use that is challenged as infringing - here, the intermediate use. n152 The preamble further suggests that, in general, privileged uses will be those that are intermediate in some fundamental sense. n153 To the extent that criticism, comment, news reporting, teaching, scholarship, and research all involve the use of copyrighted materials, they do so as a means to a different end, whether that end is the creation of a new work or simply the attainment of new understanding that may lead to the creation of new works in the future. In each case, the copier does not profit or benefit from distribution of the chosen portions of the copied work, but rather from the original contribution added or from the knowledge gained, which may then be applied to the copier's own creative projects. The copied work serves as raw material for both endeavors.
Self-evidently, not all intermediate uses will be fair ones. Some intermediate uses involve no more than steps toward unauthorized cutting and pasting of another's creative material; in that case, the copied work is both the raw material and, essentially, the final product. Thus, for example, the use of a scanner to scan works into a computer for redistribution would not, standing alone, be a fair use, but simply an unauthorized appropriation. However, as the foregoing discussion illustrates, a rule that privileges only transformative uses is too narrow. Under the transformative use standard as outlined by Judge Leval, to be fair, a use must seek to comment on the copied material in some meaningful way. Yet the inclusion of teaching and research among uses that are presumptively fair establishes that pedagogical uses of copied material can be protected and that no new work incorporating portions of the copied work need be created. n154 Thus, the language of section 107 suggests that in appropriate circumstances, a [p*1121] use that simply enables understanding of the copied material may also qualify as a fair use. n155
How does Accolade's use of Sega's copyrighted work fare under the "enabling use" standard I have suggested? Assuming that Accolade conformed to prescribed procedures for reverse engineering (we will later consider ways to ensure that the copier adheres to those procedures), Accolade sought only to understand Sega's work, not to comment on or "transform" it. That motivation cannot be a reason to hold Accolade's use unfair. Logically, whether a fair use has occurred cannot turn solely on whether a new work is created that comments on protected portions of the copied work. It would be odd if a use that does not seek or rely on copyrighted material at all, other than to understand it, were penalized more harshly than uses that do seek and rely on creative material. n156
Logic aside, the copier's motive is centrally relevant to consideration of the purpose and character of its use. Although it is hornbook law that neither the copier's motive nor the nature of the use is relevant to a determination of whether the copying has infringed the owner's exclusive rights under the Copyright Act, n157 fair use requires a different, inherently equitable analysis. Motive alone will not determine whether a transformative or enabling use has occurred, but it is indisputably relevant to any analysis conducted according to an "equitable rule of reason." n158 Equity may consider whether a copyrighted material was intended to transform, to gain access to knowledge not otherwise available, or merely to exploit.
Returning to Accolade's motive, we must consider whether the motive of gaining access is legitimate in the eyes of the copyright laws. Professor Miller argues that "the law imposes no duty on authors to provide access to the ideas in a copyrighted work." n159 His views on [p*1122] access are rooted in his belief that "the base objective of copyright is for society to benefit from the availability of creative works - that the progress of science and the useful arts be "promoted' - whether or not the literal expressions or underlying ideas of those works are directly available to the public." n160 As a practical matter - and the Copyright Act is nothing if not practical in intention - this view ignores the fact that access to existing works by authors is closely related to the continued availability of new works to the public. n161 Inspiration does not occur in a vacuum. As the Supreme Court has recognized, the freedom to build on the public domain elements of existing works promotes copyright's overall purpose of promoting innovation. n162 By necessary implication, the Constitution and the Copyright Act mandate a right of access to those elements. n163
Professor Miller is, of course, correct that the Copyright Act contains no express provision mandating the accessibility of ideas. Before the advent of computer programs, such a provision would have been meaningless. n164 However, whether the law mandates accessibility and whether it allows access are separate questions. Permission to gain access is implicit in the statutory provision that copyright protection will not, under any circumstances, be granted to facts, functional principles, or ideas - so that the flow of new works will stimulate, not preclude, further innovation. n165 Allowing copying to gain access to a program's functional elements thus does not frustrate the purpose of copyright, but furthers it.
[p*1123]
Professor Miller's final criticism with respect to the purpose and character of decompilation is a practical one. He argues that under the regime established by Sega, courts will be unable to detect and punish thefts of creative material in the competitor's final product, because a competitor could "electronically massage the copy until every trace of that illicit reproduction is obscured." n166 This observation betrays a telling unfamiliarity with the actual process of computer programming. n167 Even if Professor Miller is right, however, the "massaging" process would not obviate the need for creative effort. Because elements of a computer program's structure, sequence, and organization may be protected by copyright, it is likely that a great deal of effort would be required to remove all traces of creative expression and idiosyncratic style. n168 Moreover, the incentives to "overmassage" the copy, and thereby avoid a judgment of infringement, are considerable.
Arguably, however, a competitor who uses a copyrighted computer program as a template for producing a program with the same functionality is no different from an aspiring suspense novelist who writes with the works of John LeCarre, Robert Ludlum, and Tom Clancy arrayed on the desk, or a romance novelist who consults Danielle Steele and Judith Krantz in the course of developing an "original" plot line. The only difference, once again, is that the programmer must decompile the program to understand what it is doing. The quantum of originality required to bring a work within the ambit of the Copyright Act is very small. n169 A work may be "derivative," critically speaking, without being a derivative work. In short, even if computer programs are properly classified as "literary works," there are many more Danielle Steeles among programmers than Ernest Hemingways. More to the point, a pre-existing program, like a pre-existing novel, may be consulted for ideas, systems, procedures, and methods of operation - elements that copyright does not protect. n170 The second programmer who also imitates protected expression might infringe, but to bar programmers from consulting the copyrighted [p*1124] program at all would confer more protection on computer programs than on other copyrighted works.
Assuming, however, that allowing the competitor continued access to the copied work creates too great a risk, there is a simple enough solution: make "clean room" programming a precondition for a finding of fair use. Under a clean room protocol, the task of decompiling the copyrighted program and that of developing a new program are carried out by two different teams of programmers. The second team, charged with program development, is provided with the functional specifications extracted from the decompiled program by the first team, but no more. n171 Accolade used clean room procedures, and that fact weighed heavily in its favor. n172 Courts assessing decompilation could easily require that clean room procedures be followed and documented; the burden would then shift to the copyright owner to show, as it must for a finding of infringement with respect to the copier's final product, that protected material was taken.
The Ninth Circuit in Sega agreed with the district court that the third statutory factor, the amount and substantiality of the copying, weighed against Accolade. n173 However, the court noted that the factor was "of very little weight" given the limited nature of Accolade's ultimate use of Sega's code. n174 The court's dismissive treatment of the third factor is consistent with the case law, which indicates that the amount copied is perhaps the least critical factor of the four. n175 However, as Professor Miller notes, the court's sudden focus on ultimate use is inconsistent with its approach to the purpose and character inquiry. n176 Where the use alleged to be fair is intermediate, the court should consider the amount and substantiality of the copying done at the intermediate stage.
As the Sega court observed, evaluation of the amount and substantiality of Accolade's intermediate copying did not bode well for [p*1125] Accolade, because it had copied Sega's entire game program during the course of its reverse engineering efforts. n177 However, to hold that Accolade had copied "too much" would overlook the fact that until Accolade had decompiled the entire program it could not know whether it had all the information necessary to produce Genesis-compatible games. The object-coded representation of a computer program produced by a decompiler lists program steps in the order in which they are coded, not the order in which they are executed. n178 For example, a series of interoperability-related instructions performed during the startup of a game program may be dispersed throughout the program microcode, linked by "jump" commands that tell the computer to skip to a different portion of the microcode. n179 The reverse engineer must decompile the entire program to locate those instructions. Other interoperability-related instructions may be performed while the game program is running, and those also may proceed via "jump" commands. Once again, the reverse engineer cannot know whether all steps necessary for interoperability have been located without checking the entire program. Thus, decompiling Sega's entire program was not an indulgence, but a necessity.
Generally speaking, legal scholars have agreed that courts evaluating claimed fair uses should consider the amount and substantiality of the copying in light of the nature of the use and the other statutory factors. n180 The foregoing analysis is consistent with this consensus. The third statutory factor still weighs against the reverse engineer who uses decompilation to discover interoperability requirements, but given the other characteristics of that use, the fact that it entails copying the entire program should not preclude a finding that the use is a fair one.
The Sega district court's analysis of the fourth statutory factor, the effect of the unauthorized copying on the market for the copied work, n181 paralleled its analysis of the first. Judge Caulfield read Harper & Row to establish a presumption that heavily favored the [p*1126] copyright owner, Sega. n182 Again, the Ninth Circuit indicated that fair use requires a more fact-specific approach, n183 and again, the Acuff-Rose Court later agreed. n184 This time, however, Acuff-Rose does less to dispel the prevailing confusion about how courts should evaluate the market effects flowing from "commercial" uses.
Judge Caulfield relied on the Harper & Row Court's statement that "fair use, when properly applied, is limited to copying by others which does not materially impair the marketability of the work which is copied." n185 Many fair use decisions by lower courts subsequent to Harper & Row have read "materially" out of this test, finding challenged uses unfair if they would have any effect at all on the market for the copyrighted work. n186 Literally, of course, any use of a copyrighted work has some effect on the market for that work. However, just as section 107 does not require that every use with a commercial purpose be found unfair, neither does it require that any market effect preclude a finding of fair use. It merely instructs courts to consider that effect as one factor among many. n187 In particular, as the Supreme Court recognized in Acuff-Rose, the fact that a work is used in a commercial or for-profit setting does not create a presumption of market harm. n188 The question remains, as always, at what point - short of every use, or every "commercial" use, of copyrighted material - to draw the line.
The Acuff-Rose Court distinguished between "potentially remediable displacement and unremediable disparagement" in the form of criticism. n189 Defendants' parody of the plaintiffs' copyrighted song fell into the latter category, and was held potentially fair, pending further factfinding. In light of the high intellectual and First Amendment values placed on criticism in all its forms, the Court's conclusion seems unimpeachable. However, Acuff-Rose should not be read to [p*1127] indicate that every instance of "potentially remediable displacement" is unfair. First, as a technical matter, that question was not before the Court. Given the Court's express rejection of the Eleventh Circuit's attempt to read Sony and Harper & Row as establishing bright-line rules for separating fair from unfair uses, reading its remarks on displacement to create such a rule would amount to willful misunderstanding. n190 Second and more important, the "transformative use" criterion adopted by the Court implicitly broadens the category of permissible uses. There are many conceivable commercial uses of a work (including research and news reporting) that do not necessarily constitute "disparagement" in the sense of parody or unfavorable criticism. Finally, as a matter of logic, the fact that a particular use is "potentially remediable" cannot alone be grounds for a finding of unfairness; many fair use disputes have arisen precisely because a license to use the work has been denied. n191
The question left unresolved by Acuff-Rose, but directly addressed in Sega, is: Under what circumstances is a use that occasions displacement of the copyrighted work fair? The Sega court held that analysis of market effect must include consideration of the extent to which a given use simply enables a competitor to enter the market with another work of the same type. n192 For Professor Miller, that conclusion is tantamount to sanctioning piracy. n193 However, as discussed above, the connection between use of the copyrighted work and competition with the copyright owner is indirect. n194 As the Sega court recognized, simple common sense dictates that the absolute rule applied in Harper & Row, which involved a "scoop" of the heart of a copyrighted work that threatened to supplant the market for the work entirely, cannot logically be extended to works that are "the same" only to the extent that both are compatible with the same computer operating system. n195 To exclude those competitors from the market for that reason would effectively protect not only the copyrighted work's expression, but the underlying ideas as well. Of course, allowing access [p*1128] to interoperability requirements affects the market for the copyrighted work, because it facilitates increased competition in that market. However, the purpose of copyright - to encourage the production and distribution of creative works - is best served by allowing such competition, not by blocking new market entrants. n196
If a use that enables production of a competing product is unfair, then the result of the fair use analysis would have been very different if Accolade had simply decompiled Sega's operating system rather than its games. In that case, disassembly would have resulted in the development of a complementary product rather than a competing one. Accolade's final product, however, would be the same, as would its effect - whatever that may have been - on the market for Sega's games. Moreover, as applied to computer programs, the term "displacement" may be misleading. As a result of the interdependence among applications programs and operating systems, the consequences of a "displacing" use are by no means one-sided. At the very least, it is possible that the increased availability of compatible games translated into increased sales of Sega's console, which in turn would translate into increased demand for all Genesis-compatible games. If so, the work created through copying complements and supplements the original. This does not justify holding all such uses fair, but it is reason enough to hold that not all such uses are unfair per se. n197
For Professor Miller, the Sega court's emphasis on access to unprotected functional principles is, "at bottom ... an argument for standardization." n198 This conclusion mistakenly conflates two quite different concepts. As at least one commentator has observed, there [p*1129] is a world of difference between removing legal barriers to compatibility and making system incompatibility illegal. n199 While mandated standardization would foreclose, or at least inhibit, the development of new computing methods and operating systems, its polar opposite, the complete absence of interoperability, also would prove detrimental to innovation, and to consumers. n200 Taken to extremes, complete acompatibility would require consumers to buy a different computer system - or, in the future, install a different set-top box - for each application desired. More realistically, selective licensing would make available a limited number of preselected combinations of copyrighted products, depriving consumers of the opportunity to mix and match individual applications according to their individual tastes. Only two parties will have the power to determine what combinations of copyrighted applications will be made available: the manufacturer of the uncopyrighted, and in most cases uncopyrightable, hardware system, and (in the case of the set-top box) the telecommunications provider. Although some manufacturers of computers and operating systems release interoperability-related information, it has become clear that others will not. The video game industry, which is immensely profitable, is simply the most egregious and the earliest example. n201 Clearly, the effective monopoly over functional principles that results from such a policy is not a situation anticipated by Congress or by the members of CONTU. However, given that copyright does not protect functional features, such a monopoly cannot be what Congress would have intended. n202
[p*1130]
As the foregoing discussion makes clear, the controversy over Sega's application of the fair use doctrine to the reverse engineering of computer interoperability requirements concerns more than just a narrow and highly technical subject. It is also a debate about the purposes of copyright protection and the role of the fair use defense within the larger statutory scheme. The task of crafting copyright doctrines that are responsive to the needs of new technologies, as well as to those of more traditional forms of creative expression, cannot proceed without a clear and shared understanding of what copyright protection seeks to accomplish, and how the particular rule at issue furthers that agenda.
Professor Fisher identifies four objectives of "copyright law in general and the doctrine of fair use in particular" that emerge from Sony and Harper & Row:
(a) advancing social utility by increasing the supply of intellectual products and facilitating their distribution; (b) enforcing an author's natural right to a reasonable portion of the fruits of his labor; (c) protecting an author's interest in controlling the way in which his creations are presented to the world; and (d) aligning the law with custom and popular conceptions of decent behavior. n203
Certainly, to varying degrees all four are objectives of copyright law. With respect to fair use in particular, however, the list is overinclusive.
A finding of fair use effects a contraction of the scope of copyright protection. With each successful assertion of the defense, a use of a copyrighted work that ordinarily would constitute infringement entitling the copyright owner to redress is held, instead, to be lawful. n204 The second and third objectives identified by Professor Fisher, in contrast, are reasons for according copyright protection broader, not narrower, scope. Natural rights justifications for copyright and theories of "moral rights" or "artistic integrity" n205 are not concerned [p*1131] with questions of larger social utility, in particular the increase in social utility that might result from allowing access to and limited fair use of copyrighted works. n206 Accordingly, natural rights concepts can never be affirmative justifications for a finding of fair use. They are important only in the negative sense, as reasons for declining to find fair use in particular cases. n207
The fourth objective, tailoring the scope of copyright protection to custom or community standards of acceptable conduct, may favor either expanding or contracting the scope of copyright protection, depending on the circumstances. Unlike natural rights and moral rights theories, therefore, custom is a plausible underpinning for the fair use doctrine. A community standards or "fairness" approach also is consistent with the fair use doctrine's equitable origin and intent. n208 A legislatively sanctioned element of "gut instinct" lies at the core of every fair use determination. n209 Finally, to a significant degree, a fairness justification for fair use reflects reality. Professor Weinreb, in particular, has shown that custom and community standards go a long way toward explaining what judges actually do in fair use cases. n210
However, recourse to community standards of fairness can resolve only those cases in which the community standard or "customary practice" invoked truly is a shared one and transcends the particular commercial interests and agendas of those involved in the case. Sega is a case in point. In some sectors of the computer programming industry, reverse engineering is an accepted method of innovation and competition. In others, it is viewed as barely one step above industrial espionage. n211 In general, at least in this country, the smaller software companies are of the former persuasion, while the [p*1132] larger, established manufacturers of hardware and operating systems - those with the strongest interest in keeping operating systems proprietary - espouse the latter. n212 Had the Ninth Circuit attempted to resolve Sega based solely on the parties' representations as to accepted practice, its task would have been impossible. Moreover, a choice between values advanced by competing business lobbies smacks of mere politics, and lacks legitimacy. If considerations of accepted practice are the primary factor motivating judicial decisionmaking in the area of fair use, the results may be perceived as unprincipled and, to both authors and would-be infringers, readily manipulated. In cases where a real, widely shared consensus exists, that consensus can assist judges in evaluating claims of fair use. n213 In cases where representations regarding accepted practice simply reflect the commercial agendas of the parties, notions of custom and fair play cannot supply enduring and principled rules for determining when the boundaries of copyright protection should contract.
Of the four objectives of the copyright laws identified by Professor Fisher, only the first - increasing the supply of creative works and facilitating their distribution to the public - remains as a potentially viable unifying justification for fair use. In a very real sense, therefore, the debate over fair use reduces to a debate over how this goal is best achieved. n214 As a threshold matter, we must consider whether protecting authors' exclusive rights does not in itself accomplish this goal. Clearly, it does - that is the rationale for according authors copyright protection in the first place. n215 However, the existence of the fair use privilege reflects an implicit consensus that protecting authors' rights to the hilt is not always enough to ensure adequate supply and distribution of creative works. If it were, the scope of copyright [p*1133] protection with the fair use doctrine would be coextensive with the scope of copyright protection without it.
Thus, the fair use doctrine represents a determination that affirmative measures to increase the supply and distribution of creative works may occasionally be necessary. The forms taken by these exceptions to the general rules governing infringement will depend on how "supply" and "distribution" are conceived. As discussed above, commentators differ as to whether the Copyright Act (and the Copyright Clause) were intended only to secure the widespread availability of creative works to the public or also to secure access to the expressive and nonexpressive building blocks used to create them. n216 I have argued that particularly in the case of computer programs, access and public availability are inseparably related. If interoperability-related information is denied to programmers, the flow of new creative works into the market may slow to a trickle. Fair use thus must promote access to creative works as well as their distribution for public consumption.
The "transformative use" criterion developed by Judge Leval and adopted by the Acuff-Rose Court serves this dual purpose, for it rewards the use of copyrighted works as raw material in the creative process. n217 However, I have argued for a broader conception of fair use than either Judge Leval or the Acuff-Rose Court recognized. It appears that Judge Leval would not privilege an "enabling use" aimed only at understanding a work unless that use could be directly traced to the production of a new creative work. n218 Yet, section 107's enumeration of teaching and research as presumptively fair uses suggests that exploration alone, without more, may be fair, and that interpretation makes sense. n219 Privileging the acquisition of knowledge, even without a close temporal connection to new expression, creates a society in which authors and the creative process can flourish. n220 New creative output, however remote in time, still must meet the standard for noninfringement. Fear of hypothetical future infringement cannot [p*1134] justify limiting access to the information and expression contained in existing works.
Of course, fair use should privilege enabling uses only to the extent that it does not thereby undermine the other objectives of the copyright law so greatly that it produces a net disincentive to create and disseminate new works. n221 Once again, this is a slippery standard. As a matter of logic, whenever the fairness of a particular use is disputed, it is likely that a finding of fair use will undermine at least one of the other goals to some degree. If fair use is to have any scope at all, there must be occasions when the other goals of copyright should give way, but it is important as well to identify those occasions when they should not. Although fair use is inherently irreducible to general rules, certain broad parameters suggest themselves.
First, the use should not offend fundamental, universally shared community standards regarding commercial fair play. In contrast, community standards of fairness should play a lesser role when there is significant, pervasive disagreement - or no widespread public opinion at all - on the commercial fairness of the use. n222
Second, to be fair, a use should not strike at the core of the author's "natural right to a reasonable portion of the fruits of his labor." n223 Thus, for example, an incorporation of any of an author's creative expression into a final product should not constitute an attempt to appropriate the heart of the original work and supplant market demand for that work. However, as the debate over decompilation and interoperability demonstrates, the same rule should not extend to a use of creative material that enables creation of a new work that simply competes with the original in the market for works of its type. In particular, where copying is necessary to gain access to and understanding of the ideas and principles embodied in a work, copying solely to gain understanding should be deemed a fair use.