Julie E. Cohen*
68 S. Cal. L. Rev. 1091 (1995)
Copyright 1995 - HTML Version Prepared by the LII with Permission
[p*1135]
The Sega court's decision to sanction decompilation was based on its belief that the program attributes that dictate interoperability requirements are functional features that copyright does not protect and that reverse engineers have a right to copy. n224 It follows that the freedom to decompile established in Sega must be accompanied by the right to produce a final product compatible with the desired computer system. The same principles and policies govern both inquiries. At each stage, the court must weigh considerations of functionality, public access, and creatorship. The balances reached, and the incentives that result, should be consistent. It makes little sense to allow intermediate copying where necessary to understand the requirements for interoperability, only to withhold permission to incorporate the copied work's functional features to the extent necessary to make the reverse engineered product interoperable. Both types of conduct should be allowed, or disallowed, to the same extent and in the same manner. n225
What, then, to make of Sega and Atari? Are they consistent - as I have defined that term - with each other and with the language and overall purpose of the copyright laws? They are not. Atari's misconduct aside, the result of the copyright inquiry in Atari should have been no different from that in Sega. The same principles that dictated that Accolade be allowed to disassemble Sega's programs also dictated that Atari be allowed to design a program that incorporated all of the functional characteristics of the 10NES. Instead, the Atari court relied too heavily on doctrinal formulations developed for analyzing traditional literary works, and in doing so lost sight of exactly which aspects of computer programs copyright does not protect. This part explores the application of the "idea-expression" distinction and its offshoots, the doctrines of merger and scenes a faire, to lock-out routines and other interoperability-related features of computer programs. [p*1136]
Whether the Copyright Act permits duplication of the functional features necessary for interoperability has been said to depend on where the boundary between protected expression and unprotected idea is drawn. n226 The location of the boundary between protectable and unprotectable subject matter is, of course, the central project of copyright. Drawn too narrowly, copyright protection will provide insufficient incentives to invest creative effort; drawn too broadly, copyright in existing works may impoverish future creative efforts. n227 Lost in the formulation of the boundary-drawing project as an idea-expression distinction, however, is the fact that copyright also does not protect systems, processes, procedures, or methods of operation - much more relevant and analytically useful concepts in the computer programming context. n228 In part because of the idea-expression formulation, early cases concerning the scope of copyright protection for computer programs swept too much within copyright's scope. Recent decisions have begun to correct that imbalance. However, considerable uncertainty remains among courts as to which program elements copyright protects, and how to approach the tasks of describing and identifying them.
At first, beginning with Whelan Associates, Inc. v. Jaslow Dental Laboratory, Inc., n229 courts defined originality in computer programming largely without reference to external constraints. The "idea" of a program was defined with reference to the program as a whole, such [p*1137] as, for example, the idea of a word processing program. n230 Most industry commentators and legal scholars argued that efficiency considerations and interoperability requirements, properly understood, operate as functional constraints, and that program subroutines or processes designed to satisfy those constraints therefore should not be considered part of the programmer's creative expression. n231 However, although subsequent courts criticized or rejected various aspects of Whelan, few questioned its basic premise that the creative content of a computer program could be assessed without looking beyond the program itself, n232 and none undertook a major reformulation of the Whelan court's approach to evaluating the "substantial similarity" of nonliteral program elements.
Not until May 1992, less than two months before oral argument in Sega, did a federal appellate court systematically attempt to dismantle and replace the Whelan paradigm. In Computer Associates International, Inc. v. Altai, Inc., n234 the Second Circuit held that program features dictated by efficiency or interoperability constraints must be [p*1138] "filtered" out of the substantial similarity analysis in software infringement cases. n235 In so holding, it became the first appellate court to treat computer programs as fundamentally interdependent and highly utilitarian works, with at least some features driven by functional constraints rather than creative choice. n236 Professor Miller has suggested that Altai's "abstraction-filtration-comparison" test does not constitute a philosophical departure from Whelan, but merely fleshes out issues identified in Whelan. n237 As discussed below, Altai, like Whelan, framed the problem as one of demarcating "idea" - as opposed to "system" or "process" - from "expression." n238 However, one may search Whelan in vain for any recognition that a computer program is dependent, to a degree, upon the functional constraints imposed by its operating environment. Under Whelan, every program element is presumptively protected by copyright to the extent that it conveys information - which, of course, is what program elements do. n239 As a result of this inclusive approach, comparison of the challenged and copyrighted works sweeps broadly, encompassing program elements that are substantially similar by necessity rather than by design. In [p*1139] contrast, under Altai, it is presumed that many elements of a copyrighted program will not be protected because although they convey information, they do so in a way that is necessary to the operation of the program and cannot be expressed in a substantially different manner; "filtration" of these elements must precede any comparison of the challenged work. n240 The upshot is that the comparisons undertaken by Altai and Whelan are materially different. The inclusion of inevitably similar unprotected elements cannot help but affect the degree of perceived similarity between the copyrighted program and the allegedly infringing one. n241
Drawing once again on an analogy to traditional literary works, Professor Miller cautions against reading Altai to require only thin copyright protection for nonliteral program elements. He notes that "the mere fact that the expression is efficient should not, without more, bar protection for original authorship in the programming context any more than it does in prose works. An uncritical application of Altai's language would penalize the most effective (and in some senses the most artistic) programmers." n242 As an initial matter, Altai did not hold that efficient programs would be unprotected, but only that sequences dictated by considerations of efficiency might be. n243 The two concepts are quite different. For example, a program routine for calling a lookup table during a complex calculation is a sequence dictated by considerations of efficiency that might be unprotected under Altai. Lookup tables are the prevailing industry method for performing complex calculations quickly with minimum memory, and there are only so many ways for a program to call one. n244 In contrast, a relational database might be an efficient solution to the problem of storing and retrieving data pertaining to bank loans, but that does not mean that relational database programs are not protected by copyright. There are a number of different ways to write such programs, and several leading programs are available for purchase. n245
More important, the Copyright Act does not protect efficiency per se. It protects "original expression," a term of art that refers to [p*1140] the particular instantiation given to an idea or theme. n246 That instantiation may be efficient, but efficiency is not synonymous with originality, and the expression as a whole may be efficient and original for different reasons. An effective surgeon may be considered an artist within the medical profession, but the procedures the surgeon has developed and perfected are not protected by copyright. If the procedures are novel enough, they may qualify for patent protection; otherwise, they belong in the public domain - even if the surgeon writes a book about them that details how they are to be performed. n247 The fact that other surgeons may describe the procedures as "creative" or "elegant" is beside the point. In the copyright context, those words are terms of art, and section 102(b) of the Act makes clear that they do not apply to systems, procedures, or routines. n248
In other cases, a programmer may be efficient, in the sense of getting a task done, without being at all creative, if the most efficient way to perform the various steps needed to complete the task is well-known and standard within the industry. The fact that the programmer accomplishes the task by writing should not occasion a quantum leap in the level of protection afforded the work. Here again, the treatment of computer programs as literary works obscures the issue; to the extent that industry-standard efficient routines are comparable to literary works at all, they are comparable to the alphabetical arrangement of entries in a telephone directory or dictionary, a convention so commonplace that it has been held uncopyrightable as a matter of law. n249 If the programmer incorporates some idiosyncratic features into the efficient routine, copyright may protect those features, but not the routine itself. Even if the efficient routine is what gives the program its commercial value, copyright does not permit the programmer to complain when the routine is duplicated. n250
[p*1141]
The Sega court invoked Altai as a prelude to its fair use analysis, and listed as unprotected program elements those "dictated by the function to be performed, by considerations of efficiency, or by external factors such as compatibility requirements and industry demands." n251 For the Sega court, however, approval of the Altai approach was almost a foregone conclusion. The Ninth Circuit, which as the "Silicon Valley court" has assumed a central role in the evolution of computer copyright law, had never adopted Whelan, although that fact went largely unnoticed in the academic community. Instead, in previous computer program copyright cases, the court had applied the idiosyncratic "intrinsic-extrinsic" test it had developed for evaluating similarity of all types of challenged and copyrighted works. n252 In Brown Bag Software v. Symantec Corp., n253 decided several months before Altai, the Ninth Circuit explained that the first, or "extrinsic," part of the test requires "analytic dissection" of the copyrighted work to identify the core of protectable expression. n254 In Altai, the Second Circuit relied on Brown Bag to support its "filtration" and "comparison" steps. n255 Thus, although Sega was a literal copying case, not a substantial similarity case, and although the Sega court did not cite Brown Bag to support its approval of Altai, Altai was wholly consistent with Ninth Circuit law.
Although the Third Circuit (the Whelan court) has yet to comment on Altai, every other court to consider the issue has endorsed at least some aspects of the Altai approach to substantial similarity. n256 In the formal sense, Whelan has become a whipping boy, the artifact of an earlier, less technologically sophisticated era. However, that does [p*1142] not necessarily reflect a new unanimity on the question of how to determine which features of a computer program constitute protected expression. To the contrary, the application of Altai's abstraction-filtration-comparison test has proved extremely malleable. As a recent district court decision (now reversed) from the Tenth Circuit illustrates, it is possible to craft a test that, though derived from Altai, nonetheless looks suspiciously like Whelan.
In Gates Rubber Co. v. Bando American, Inc., n257 which involved two computer programs for calculating the dimensions and capacities of industrial belts, a Colorado district court attempted to apply a version of the Altai test. Relying on expert testimony, the court compared the programs and found them substantially similar in many respects. n258 It then applied what it termed the "abstractions" test to determine whether the similarities were excused, and found that many were not. n259 Like the Whelan approach, the court's analysis put the cart before the horse, with predictable result. Because the two programs contained common errors and there was evidence of deliberate copying, the outcome - a judgment of infringement - probably was warranted. n260 However, the court also found infringement in the use of mathematical constants necessary for performing the calculations, and in the engineering modules that performed those calculations using published formulas. n261 It rejected Bando's argument that the formulas and constants could be protected, if at all, only under patent laws, on the ground that "such a holding would tend to fragment further the rather tenuous continuity found in copyright law concerning computer programs." n262 But on that point Bando, copier though it was, was absolutely right; copyright does not protect formulae or processes. n263 Fear of discontinuity is insufficient reason to ignore the plain language of the Copyright Act and is ultimately beside the point; discontinuity is inherent in the statutorily mandated task of separating [p*1143] protected expression from unprotectable ideas and processes, and in the constitutionally mandated separation of subject matter protected by copyright from subject matter protected by patent. The Tenth Circuit reversed, and directed the lower court to reevaluate the two programs after filtering out unprotected matter. n264 However, the district court's opinion demonstrates that a mandate to conduct "filtration" is useless without an understanding of what elements must be filtered out, and why.
It is still too early to determine with precision Altai's impact. It is certain that after Altai, however broadly or narrowly applied, the universe of protected programming elements has shrunk, while that of unprotected elements has expanded. In individual cases, however, the basis for distinguishing between creative and functional programming elements is imperfectly defined, as Gates Rubber illustrates. Without question, some of this imprecision is inherent in the test itself, however formulated. As Judge Learned Hand concluded over sixty years ago, "nobody has ever been able to fix that boundary, and nobody ever can." n265 For software engineers concerned with creating interoperable programs, however, pinpointing the location of that boundary is of critical importance, and a matter of commercial survival. And some of the uncertainty regarding which program elements are subject to filtration is neither necessary nor desirable. We turn now to Atari and the interoperability question, and attempt to develop a clearer test.
Technically, the Atari court did not hold that Atari was barred from duplicating the features of the 10NES that were necessary for interoperability with the NES console; in light of Sega, it could not. n266 The court simply held that Atari had duplicated more than was required to achieve that goal. However, what Atari had duplicated too much of, according to the court, was functionality. The court was unpersuaded by Atari's contention that the functionality in question [p*1144] would materially affect interoperability with future versions of the NES. n267 Essentially, Atari stands for the proposition that "surplus" functionality at the program-to-program interface - functionality not needed for current compatibility with the target system - constitutes protectable, and protected, expression. In light of the fact that the Copyright Act does not protect functionality at all, that result seems curious. n268 This section considers how the court came to reach it, and how the idea-expression distinction both illumines and obscures copyright issues relating to interoperability.
Before Altai, the only significant appellate decision that dealt with the question of copying to achieve interoperability was Apple Computer, Inc. v. Franklin Computer Corp, n269 decided three years before Whelan. Franklin Computer copied Apple's operating system so its own computers would support Apple-compatible applications programs. To justify the copying, Franklin argued, among other things, that the idea of a computer operating system merged with its expression because of the multitude of functional constraints imposed by existing applications programs. n270 Finding Apple's operating system protected and infringed by Franklin's verbatim copying, the court observed: "Franklin may wish to achieve total compatibility with independently developed application programs written for the Apple II, but that is a commercial and competitive objective which does not enter into the somewhat metaphysical issue of whether particular ideas and expressions have merged." n271 Altai represents a major reassessment of that view. n272
In Altai, the Second Circuit clearly and unambiguously held that interoperability-related features must be filtered out of the substantial [p*1145] similarity analysis. n273 However, the court reached that conclusion in a fashion that should have struck industry observers as nothing short of bizarre. It invoked the doctrine of scenes a faire, a rule developed to aid in the substantial similarity analysis of fictional and theatrical works. Under the scenes a faire doctrine, "standard" literary elements or devices are not copyrightable if they are viewed as virtually indispensable to a literary treatment of the subject matter. n274 The Altai court reasoned that interoperability-related features are indispensable to a program written for a particular computing environment, and so should be treated as scenes a faire for purposes of copyright. n275
The analogy between interoperability-related routines and "stock" literary devices is rather far-fetched. The two types of expression serve very different conceptions of necessity, and are valued for different reasons. Stock literary devices function as a sort of shorthand for the communication of perceived cultural truths; they are deemed "necessary" because they greatly facilitate audience recognition of a particular cultural or historical milieu. n276 The value of scenes a faire lies in the particular expressions of the ideas that they represent. It is entirely conceivable that a work that omitted them might nonetheless be extremely effective. Interoperability-related routines, in contrast, are necessary in the most literal sense possible: Without them, the program will not function. Such routines are valued, in short, not for the expression they contain, but for the result they produce.
Although the Altai court set forth a rigorous, systematic method for identifying and excluding interoperability-related program elements, the test it devised does not appear to be rooted in a firm understanding of why those elements are unprotectable in the first place. The reason is deceptively simple. Program elements necessary for interoperability with another computer program are unprotectable not because they represent "ideas," but because they are systems, procedures, or methods of operation - in short, functional things - and so, under section 102(b) of the Copyright Act, unprotectable just as ideas are unprotectable. n277 The Copyright Act does not prohibit, but rather [p*1146] expressly allows, the creation of an exact functional analogue of a copyrighted utilitarian work. To take a well-known example, the lamp in Mazer v. Stein, n278 the copyright protected only the design of the lamp base. It did not protect the lighting mechanism, or the on-off switch, or the method of using a lampshade to reduce glare.
Atari is proof that the "why" of unprotectability matters. The "surplus" functions in the 10NES program did not become expression by virtue of their surplusage. They were designed to perform particular functions at the interface between console and cartridge, whether or not those functions were necessary in the sense of scenes a faire to unlock a particular version of the NES. There is no reason that a competing program that merely duplicated those functions - not the programming sequences by which they were expressed - should be deemed to contain expression substantially similar to the original. In a sense, Atari's characterization of the issue as a question of present versus future interoperability, with the attendant emphasis on competitive positioning, was a tactical error that confused the court. Atari argued that if copyright permitted only duplication of those functional features that were currently necessary for interoperability, Nintendo would be free to reprogram the NES master chip to look for other functional features of the slave chip, thereby freezing out all of Atari's previously released products with each new release of the NES console. n279 However, nothing in the Copyright Act prohibits a copyright owner from altering the expression in its copyrighted work to gain a competitive advantage. Copying to achieve future interoperability is entirely consistent with the language and purpose of the Copyright Act, not because an applications developer has any suprastatutory right to preempt changes in another's proprietary technology, but because all functionality at the interface between computer programs is unprotectable ex ante. n280
Here it is worth returning, briefly, to the Franklin Computer problem. Franklin Computer involved the opposite situation from Sega and Atari. Accolade and Atari produced game programs that were compatible with their competitors' operating systems; Franklin [p*1147] sought to create an operating system that would support applications originally developed for Apple's operating system. Under Altai's scenes a faire approach to interoperability, as applied in Atari, Franklin would have no right to do so, because the interoperability-related routines necessary to run a particular application could not be considered virtually indispensable to any functioning operating system. Under a functionality-based approach to interoperability, the outcome is quite different. An operating system that duplicates the systems and procedures necessary to run particular preexisting applications does not infringe. Although that rule still would not excuse Franklin's outright copying of Apple's entire operating system, Franklin could not be barred from designing its own operating system that provided the required functional base for Apple II-compatible programs. n281
Nor does the doctrine of merger afford sufficient basis for determining the degree to which interoperability-related program features should be excluded from the substantial similarity analysis. Like scenes a faire, merger is a tool for identifying instances in which expression otherwise protectable by copyright is not protected. As the name implies, merger excuses copying when the expression taken is the only way, or one of a very few ways, to convey the underlying idea. n282 The merger doctrine does not tell us how to identify those aspects of a copyrighted work that are ineligible for copyright protection in the first instance. The Altai court understood the limited role [p*1148] that merger plays in the analysis of utilitarian works. n283 However, several courts that have subsequently considered the Altai test have not. The Fifth and Tenth Circuits, in applying Altai's "filtration" step to computerized diagnostic and training systems, found the categories and subcategories within each system to be protectable expression because there were other ways to design diagnostic and training systems. n284 Both ignored the fact that the Copyright Act expressly excludes systems from the subject matter it protects, regardless of whether or not other possible systems exist. n285
Judge Keeton's much publicized "Key Reader" opinion in Lotus Development Corp. v. Borland International, Inc., n286 recently reversed by the First Circuit, illustrates the consequences of confusing the doctrine of idea-expression merger with more fundamental principles that govern copyrightability. The "Key Reader" dispute in Borland involved two spreadsheet programs, Lotus 1-2-3 and Borland's Quattro Pro. Key Reader, a module of Quattro Pro, was designed to read and execute users' short programs, or "macros," written using the Lotus 1-2-3 command hierarchy. To that end, it duplicated the Lotus command structure, including the first letters of the commands. n287 Judge Keeton likened the arrangement of commands chosen by Lotus to the arrangement of facts in a compilation, in which the organizing principle may be protected if sufficiently creative - in other words, if it is one of many possible organizing principles, and so does not merge with the underlying information. n288 Relying on his earlier rulings in the Borland dispute n289 and in Lotus Development Corp. v. Paperback Software International n290 that the Lotus 1-2-3 menu structure was only one of many possible ways to organize spreadsheet commands, he characterized the organization of the command hierarchy as protected expression. n291
[p*1149]
In support of this reasoning, Judge Keeton observed only that a novel could be characterized as a system for communicating ideas, and that "the ability to describe a work as a "system' is not decisive of whether the work is a "system,' or instead is protected expression, under copyright law." n292 Of course, the same is true of the ability to describe a utilitarian work as an exercise in the arrangement of facts and ideas. The fact that metaphor may part company with reality does not excuse inability (or refusal) to tell the difference. n293 Given the division between copyrightable and uncopyrightable subject matter set forth in section 102(b) of the Copyright Act, the fact that it is possible to describe a novel as a system for communicating ideas and the fact that it is possible to describe a computer program as a compilation of information are equally irrelevant. n294
The First Circuit's opinion recognizes the lesson of Baker v. Selden, codified in section 102(b), that a system is not copyrightable subject matter, and that its designer may not invoke the copyright laws to prevent others from practicing it even if other possible systems exist. n295 Only the expression of a system, method, or procedure implicates the copyright laws, and then only to the extent that there are many possible methods of expression, rather than just a few. n296 At that point doctrines such as merger and scenes a faire become relevant. Another well-known example, the jeweled bee pin in Herbert Rosenthal Jewelry Corp. v. Kalpakian, n297 illustrates this distinction. The issue in Kalpakian was whether the defendant's bee pin duplicated protected expression in the plaintiff's pin, or only expression that was necessary to the idea of a bee. In contrast, interoperability concerns infrastructure, not appearance. Kalpakian would be analogous to Atari or to Borland if the jeweler had sued its competitor for using a pin to attach its bee to the wearer's clothing. Had it done so, it [p*1150] would have been laughed out of court. The result for a party who attempts to bar others from practicing a system that happens to be embodied in a computer program should be no different.
It is worth noting that using the merger doctrine to identify unprotected program elements also risks erring too far on the side of underprotection. If interoperability may be defined as an idea with which particular program routines are deemed to have merged, n298 there is no reason that any other functional attribute of a program may not be similarly characterized. Taken to its logical extreme, that approach could sanction outright copying of all arguably functional program features. n299 Sega and Atari illustrate the shortcomings of this approach to interoperability.
The security system used by Sega was relatively simple: a twenty to twenty-five byte "header file" containing object-coded representations of the letters S-E-G-A. n300 Strictly speaking, Accolade's decision to incorporate the header file into its final product was not at issue in the appeal because Sega challenged only Accolade's intermediate copying. n301 However, Sega specifically argued that Accolade's duplication of its header file was not a fair use. n302 The Ninth Circuit rejected this suggestion. Among other reasons, the court noted that the sequence was "probably unprotected under the words and short phrases doctrine." n303 Even from a strict protectionist perspective, the court's conclusion seems unimpeachable. There is only one way to express the "S-E-G-A" sequence in object-coded representations of ASCII characters. Accordingly, even though the letters spelled Sega's name, copying them could not take protected expression from Sega. In copyright parlance, the idea - the name "Sega" - and its expression had merged. In that respect, Sega was not a difficult case.
The difficulty with holding that the merger doctrine excuses copying of all interoperability-related features arises in cases like Atari [p*1151] when more complex code sequences are at issue. Nintendo's 10NES program was considerably more complex than the Sega header file. n304 Accordingly, in an attempt to avoid infringing the expressive elements of the 10NES, Atari devised new programming sequences that would generate the same results as the 10NES at the required points in time. n305 The result was the "Rabbit" program, which Nintendo acknowledged was not a literal copy. n306 If the object-coded version of the 10NES embodied a complete merger of idea and expression, Atari could simply have copied it, just as Accolade copied Sega's header file. As applied to Atari, however, that approach is flawed. As I have explained, "merger" is not simply a term employed to register the presence of unprotectable functional or utilitarian matter. In the copyright context, it has a specific, technical meaning. For idea and expression to merge, there must be only one or a few means of expression available to the author, so that no real choice exists as to the manner of the expression. n307 That is true of many, but not all, functions that an operating system, or any computer program, performs. Atari's Rabbit program demonstrated that there were a number of ways to produce the mathematical results required by the 10NES. A simpler, and sounder, statement of the right to create a compatible program is that the reverse engineer may duplicate systems or procedures necessary for interoperability, because copyright does not protect them. n308
As applied to the 10NES program, or to any other complex program element, the use of the merger doctrine to justify duplication of interoperability-related program elements simply resurrects the reductionist fallacy of Whelan. Under Whelan, all was expression, no matter how utilitarian. If the merger doctrine is construed broadly to excuse copying of literal code whenever that code is "functional," all is utilitarian, no matter how expressive. Neither approach does justice [p*1152] to the unique blend of creativity and functionality present in most programs - even, however attenuated, in object-coded form - and neither recognizes that copyright already allows complete functional duplication of interoperability-related "systems," "procedures," or "methods of operation." Under the analysis set forth in Altai, the concept of merger has a useful contribution to make to the analysis of functional similarities. n309 However, merger is simply a tool for ensuring that copyright does not inadvertently shield expression that has assumed de facto functionality. As to whether others may practice a system, method, or series of procedures for achieving interoperability embodied in a copyrighted computer program, there should be no doubt; those aspects of the program simply are not part of what the author's copyright protects.