Computer software is characterized by features that defy classification within established legal doctrines. In particular, areas of the law that rely on distinctions between the "functional" and the "expressive" have found software problematic. In this article, I briefly describe the areas of law in which this problem is most pressing, sketch how these problems have migrated across legal domains, and indicate the path legal controversy attending software seems likely to take, especially in patent law. The picture that emerges is one of continuing legal controversy regarding software within the law of copyright, patent, and free speech, as well as a blurring among these areas as they are stretched to accommodate this chimeric technology.
For more than 20 years, copyright has been the primary mode of intellectual property protection for software in the U.S. and in much of the rest of the world. Copyright covers original works of authorship, including literary works, dramatic works, audiovisual works, musical compositions, audio recordings, pictorial, graphic and sculptural works, choreographic works, and even architectural works. A copyright arises whenever original expression is fixed for some substantial duration in a tangible medium of expression. This act of fixing original expression gives the copyright owner the exclusive right to reproduce and distribute copies of the work, as well as to adapt, publicly perform, publicly display, and in the case of audio recordings, digitally transmit the work.
The requirement of originality does not mean the work is novel or unprecedented, only that it must originate with the author and is not copied or taken from somewhere else. For this reason, copyright does not extend to facts or mathematical formulae; they are considered to exist in nature, rather than to originate from a human author. Originality also implies a minimum degree of creativity; copyright might cover the selection and arrangement of a compilation of uncopyrightable facts but has been held not to cover a compilation, such as a telephone directory, when the information is predictably arranged in alphabetical order.
Neither does copyright extend to methods, processes, or purely utilitarian items. This rule, now codified in the Copyright Act of 1976 as subsequently amended in 1990, extends back to the 1879 landmark U.S. Supreme Court case of Baker v. Selden, in which the plaintiff Selden asserted copyright in printed forms used to practice a method of accounting described in a book he had written on the subject.1 The Supreme Court reasoned that although Selden could assert copyright in a description of the method, copyright did not extend to the method itself. Anyone else, including the defendant Baker, could legally write his or her own book describing the same method. And since the accounting method could be practiced only by using the forms, the forms were unprotectable as functional items. Protecting the forms under copyright would have been tantamount to protecting the method, and the protection of such items, the Court declared, was the provenance of patent law, rather than of copyright. Absent a patent on the method, Baker could freely use the accounting forms.
Just as copyright seems to have expanded to encompass functional items, patent law now seems to have expanded to encompass expressive subject matter.
The principle of Baker v. Selden infuses copyright with a broad antipathy to protection of functional objects. This general principle is explicitly stated in the statute with regard to pictorial, graphic, and sculptural works, which are covered by copyright only as to their separable, nonfunctional characteristics. The specific rule for architectural works is similar, eschewing copyright protection for purely functional elements of buildings. Yet despite this general rule against protection of functional subject matter, the U.S. Congress chose copyright as the primary mode of protection for software. This choice has since led to enormous difficulties in applying copyright protection to software, since the primary value of software arises from its functionality.
For the past 25 years, courts throughout the U.S. have have struggled with the paradox of applying intellectual property protection that explicitly does not extend to functional items to an item that is primarily functional. The courts have attempted to accommodate software within copyright by excluding from consideration its functional aspects, applying copyright protection to only its expressive portions. But as one commentator has observed, any court honestly applying this test will find that nothing remains for its consideration after subtracting out the functional portions; everything in software is essentially functional.2
Nonetheless, copyright protection for software is sometimes justified by the argument that copyright also covers other representations of functional operations, such as choreographic notation, cookbooks, and how-to books. Each comprises an algorithm of some sort, such as the steps in executing a dance, steps in food preparation, and steps in repairing a toaster. Copyright also extends to blueprints and architectural drawings representing functional artifacts or buildings. Since the copyright statute defines software as "a set of instructions to a machine," the argument goes, software fits perfectly well alongside other instructional texts, such as choreographic notation, dictionaries, and blueprints or similar items.
Unfortunately, this argument demonstrates confusion of two very different definitions of "instruction"; instructions to a human are not the same as instructions to a machine. The instructions in a cookbook may say to add one teaspoon of baking soda to a recipe, but the human reading the instructions has the option of not adding the baking soda, adding some different amount, or substituting another ingredient, such as baking powder. A machine executing instructions has no such latitude, because the software instructions literally become part of the machine performing the operations they embody. Choreography, dictionaries, and blueprints do not configure humans in the manner that software configures hardware; blueprints may tell humans how to build a machine, but software is the machine. No one would mistake choreographic notation for the dance or the blueprint for the building. Instructions to a human may be useful, but they are not functional in the same sense as software.
Following this distinction, some might suggest that source code, like a blueprint, represents a machine or building, while object code is analogous to the machine or building itself. According to this argument, human-readable source code would perhaps be the proper object of copyright protection, whereas machine-readable object code is not. But U.S. courts early in the development of copyright protection for computer software rejected this distinction. As a technical matter, the distinction is blurred and difficult to apply for interpreted languages, such as Basic and Java, and for hybrid languages, such as Perl. As a legal matter, if a programmer in fact fixes copyrightable expression, it makes no difference whether the expression is fixed as source code or as object code, just as it would make no legal difference whether copyrightable expression in a sculpture were fixed in steel or marble. As a purely practical matter, it would make little sense to protect the "expressive" source code version of a program without protecting the "functional" object code version, since mass-marketed programs are generally distributed in the latter form; outside the "open source" movement, a program’s source code is typically held as a closely guarded secret, and its object code is most likely to be pirated.
This point—that there is no practical difference between source and object code for these purposes—becomes even clearer when we consider the fundamental interchangeability of software and hardware. For reasons of cost and efficiency, society has tended to deploy universal machines that can be configured by software to perform a variety of tasks. But we could also hard-wire machines to perform the same tasks, substituting hardware for software. Society has also chosen to place software within the subject matter of copyright, though no one would concede that the equivalent hardware falls within the purview of copyright. On the contrary, if two computer scientists were to present, respectively, for registration to the U.S. Copyright Office the software code for a particular application and a machine hard-wired to perform the equivalent function, the Copyright Office would readily accept the deposit of the code but reject the hard-wired machine. The hardware version would likely be turned away with the advice, much like that in Baker v. Selden, that hardware and similar functional items are the provenance of patent law, not copyright.
Software and Patent
Because patents cover utilitarian articles, it seemed natural to many observers that patent would be the form of intellectual property best suited to cover software. Patentable subject matter includes machines, processes, compositions of matter, and articles of manufacture that are novel, useful, and nonobvious. Unlike copyright, which arises spontaneously when a work is fixed in a tangible medium, a patent is obtained by submitting an application to the Patent Office. The application must describe how to make and use the claimed invention and contain one or more claims delineating the invention covered by the patent. Once the patent application is approved, the inventor may, for a period of 20 years, exclude others from making, using, selling, offering for sale, or importing the claimed invention.
The protection of software under copyright, rather than solely under patent law, is something of a historical accident. During the 1970s, the U.S. Supreme Court considered the patentability of software, initially declaring it unpatentable under a rule that prevents the patenting of abstract ideas or formulae.3 In the view of the Court, algorithms seemed to constitute such abstract ideas or methods, placing software comprised of such algorithms outside patentable subject matter. The Court revisited the issue in 1981, amending its position to hold that software by itself was unpatentable but, if tied to some physical component or substrate, could be patentable.4 This ruling led to elaborate judicial tests to determine the substantiality, or embodiment, of computer programs, as well as to the widespread practice by patent attorneys of claiming software in terms of the state of the machine configured.
Throughout the 1980s and 1990s, as applications for software patents became more common and the importance of the technology increased, lower courts incrementally stretched the Supreme Court’s standard to cover more and more types of software. These courts had little difficulty in at least nominally tying software to a physical substrate, as the Supreme Court had required. To constitute patentable subject matter, software needed simply to produce some physical change in a machine—something a program unquestionably does; the execution of each step in an algorithm places the hardware in a different electrical or magnetic state. Thus, in one case, the stored data in a standard form of computer memory was held to be patentable, since the storage of the data placed the machine in a different state from the one it would otherwise have been.5 Similarly, storing data or software code on a magnetic disk has been recognized as producing a novel and patentable article of manufacture, because the disk with the stored data differs from the disk without it. Under this approach, software might be claimed as either a "machine" or as a "process," because the algorithm produces a sequential series of different machines.
Just as a work need not necessarily communicate with a human to qualify for copyright, a work does not necessarily qualify as protected speech even if it communicates a message to a human.
In the process stretching patent law to accommodate software, the courts went to some lengths to alter or discard certain rules of patent law that would tend to exclude software from being patented. For example, the courts have long held that mere arrangements of lines or symbols would not constitute patentable subject matter. Such "printed matter" was not considered to constitute patentable machines, manufactures, or compositions of matter. Novelty in substance, language, or meaning of printed indicia was similarly considered unpatentable. This is not to say that an invention would be unpatentable simply because it was printed on paper or some other substrate. Rather, to meet the criteria for patentability, the claims must define new features of physical structure or new relations of printed matter to structure. So, for example, some methods of fraud prevention involving detachable menu or railway tickets were held patentable, because they were tied to a physical cardboard ticket structure involved in the claimed process.
This rule tended to reinforce the dividing line between the functional and the aesthetic by removing literary or artistic works from the realm of patents; a poem or painting might be novel and even nonobvious but was considered to be the proper subject of copyright, rather than of patent protection. Forms and labels that had as their sole point of novelty printed markings or indicia were held unpatentable by U.S. courts. So too, accounting or indexing methods involving novel arrangements of symbols on known substrates, such as paper, were unpatentable, as were methods of arranging directory names by phonetic groups when the novelty lay in the arrangements of the names.
Ironically, despite the dictum of the Supreme Court in Baker v. Selden that Selden should have sought a patent for his accounting method, the printed-matter doctrine would likely have denied the uncopyrightable accounting forms of Baker v. Selden patent protection as being unpatentable "printed matter" involving only an arrangement of marks and not operating on a physical substrate.
The requirement of a physical substrate for printed matter evolved into the Supreme Court’s holding that software must be tied to a tangible embodiment to qualify as patentable subject matter. But the printed-matter doctrine potentially posed a severe impediment to the patentability of software, which at some level resolves to indicia or symbols written on a physical substrate, such as paper and magnetic media. The courts have now eliminated this impediment by requiring only a novel arrangement of indicia or structure in the substrate. The difficulty with this approach is that data and software are essentially indistinguishable; as one judge has observed, magnetic media or a compact disc containing music files are novel arrangements of indicia in a physical substrate.6 Under current U.S. law, Selden’s method of using accounting forms would likely be patentable subject matter, so long as it produced "a useful result," but so too would many other arrangements of symbols, including, apparently, word processing files and music CDs. Thus, just as copyright seems to have expanded to encompass functional items, patent law now seems to have expanded to encompass expressive subject matter.
Software As Speech
The hybrid nature of software has surfaced in an additional area of the law with important repercussions for the convergence of patent and copyright. In a series of recent cases, the courts have grappled with the issue of whether or not software comes under the protection of the First Amendment, that is, whether software constitutes protected speech. Several such cases have involved constitutional challenges to federal restrictions on the export of cryptographic software.7, 8 More recently, the issue of First Amendment protection has been raised in a copyright case involving a program capable of circumventing the encryption controls on DVD-format movies; defendants in the suit argued that the program code, which was outlawed under an anti-circumvention statute passed in 1998 by Congress, is protected speech and the statute an unconstitutional prior restraint on such speech.9
In each of these cases, the fundamental question for the courts was whether software is primarily functional or expressive, that is, whether software is "speech." The question of what constitutes speech is not entirely new to First Amendment jurisprudence. The Supreme Court has previously considered whether activities as diverse as nude dancing, flag burning, even parade marching qualify as speech under the U.S. Constitution.9 The Court has at times based its definition of speech on whether the activity is communicative, that is, whether it conveys a message to an audience. Yet conveying a message cannot by itself be the touchstone of First Amendment protection; the majority of people cannot perceive the message in many types of expression, as in, for example, modern art and atonal music.
The Court has indicated that such expressive activity is protected, even if understood by few if any recipients. Thus, the Court has at times tempered its communication requirement with the proviso that speech must convey a message to a particularized audience or entail the potential to convey a message to those who might acquire the knowledge to understand the message.10
To those familiar with programming, the status of software under such a test may seem obvious, at least for source code. Source code is certainly comprehensible to the audience of computer programmers, conveying ideas to other programmers. This argument for code as speech certainly carries with it some superficial appeal. Programmers are undoubtedly expressing something when they compose the code that comprises a program. They think carefully about what they are composing. That careful thought, some of which may involve very innovative ideas, is transposed into the notation of a computer language other programmers can read and decipher to extract the logic that went into its composition.
But as we have seen in the case of copyright, such attempts to map expressive paradigms onto what is primarily a functional item have been highly problematic. The fact that an engineer "reading" program code appears to be reading textual instructions, rather than tracing logic circuits, is an artifact of the medium from which the program is built; the engineer "reading" the program code is extracting design information from a functional machine that happens to be composed of text.
The same characteristics are shared to a greater or lesser extent by other functional items. My running shoes, automobile, table cutlery, and essentially every other object around me in an industrialized society is the embodiment of some artisan’s or engineer’s design. Thus, the final product of such industrial design choices, functional as they may be, in a sense "communicates" the ideas of its originating artisan or engineer and so could be said to constitute a sort of "speech." As the Supreme Court has noted in other contexts, it is possible to find some kernel of expression in almost any activity, including, for example, walking down the street or meeting one’s friends at the shopping mall, though such a kernel is not sufficient to bring the activity within the protection of the First Amendment. Similarly, every human artifact contains some kernel of "expression," but we do not protect them all.
This First Amendment problem is in some sense the inverse of the problem encountered in copyright law; just as a work need not necessarily communicate with a human to qualify for copyright, a work does not necessarily qualify as protected speech even if it communicates a message to a human. Litmus paper conveys a message to a human, but that message is unlikely to qualify as protected expression under either the First Amendment or the Copyright Act. However, the courts have accepted that computer operating systems, which do not convey a message to a human, may qualify as protected expression under either the First Amendment or the Copyright Act. Simultaneously, if either the litmus paper or the operating system produce a useful result, the paper or the system is now the proper subject matter of patent. This convergence of expressive and functional subject matter has brought us to the point that patent law now covers expression, as patent law increasingly extends to software-related patents, including systems of writing and for voice command recognition, not to mention magnetic or structural elements comprising expressive works.
Accommodating Patentable Expression
The peculiar nature of software now demands an accommodation between the law of patents and the law of free speech. But it seems likely that the law of patents is not prepared for this juxtaposition with the First Amendment and may need considerable adjustment after the fashion of copyright. Copyright may have provided a poor guide for accommodating software within the law of the First Amendment but may provide a better guide for accommodating the First Amendment within patent law. Unlike patent law, copyright law has long been required to accommodate free speech issues through, for example, the doctrine of "fair use."
Copyright may have provided a poor guide for accommodating software within the First Amendment but may provide a better guide for accommodating the First Amendment within patent law.
In copyright, fair use serves a variety of functions. First, it is widely held to facilitate worthwhile uses of copyrighted works where the value of the use is exceeded by the transaction costs of negotiating a license. Under such conditions, an unfettered "right to exclude" might deter such valuable uses; the potential users of the work are unlikely to expend more time and money to locate the owner and negotiate a license than they can recover from the licensed use. The fair use doctrine allows the potential user to take the needed portion of the work and make use of it without seeking a license, bypassing the need for deterrent high-cost authorization.
A second, related purpose of fair use is to facilitate socially valuable uses of copyrighted works for which the value is not fully internalized. Commentary, criticism, parody, and other unauthorized uses of intellectual property may encourage public debate and help foster an informed populace, but this value is diffuse and accrues to recipients other than the user of the copyrighted work. In the case in which such positive externalities are present, social welfare would be increased by the use of the work, but potential users may be deterred from doing so because they will not assess the use by its full value. Only a portion of the value of the new work will accrue to the potential user, and transaction costs in excess of that portion will produce the kind of market failure described earlier. In order to encourage creation of the work, fair use may again serve to bypass licensing that appears excessively costly from the perspective of the potential user.
Because copyright has traditionally encompassed expressive subject matter, fair use also functions as a sort of safety valve between the purposes of copyright and the demands of the First Amendment. Copyright is a type of restraint on speech, that is, the property right in the expression restrains others in their use of that expression. The First Amendment typically prohibits such governmental "prior restraints," but in the case of copyright, governmental interference is constitutionally authorized. Fair use mediates between these apparently contradictory constitutional provisions, especially in the instances outlined earlier in which parody and criticism foster an aware and educated populace better able to participate in the exchange.
As copyright has increasingly been forced to deal with functional artifacts within its ambit, fair use has also become an important way to prevent copyright holders from dominating new markets or from impeding development of new products. For example, fair use has been employed to provide the necessary flexibility in copyright law to allow reverse engineering of copyrighted software. The creation of software compatible with a given operating system or application program may require examination of the program’s structure in order to design an interoperable product. The utilitarian functions of computer programs lie in the public domain and may be freely copied unless patented, as outlined earlier. However, companies developing competing or complementary products necessarily create a copy of the program being studied in the course of decompiling it to examine it. This copying during reverse engineering might be considered an infringement, but the courts have consistently held that making intermediate or temporary copies in order to study a program and extract its public-domain information is fair use.11
Although fair use has proved critically important for all these roles vis-á-vis copyright, no such doctrine has emerged to fulfill these roles in patent law. One would think that in the case of functional technology, no less than the case of nonfunctional expressive works, instances might arise in which market failure is present. The cost of locating a patent owner and negotiating a patent license might well deter a potential user from engaging in valuable social use of the claimed invention. In many cases, access to patented technology may also involve positive externalities, in which the social value of certain uses of patented technology may be too diffuse to prompt a potential user to seek a license. The conventional wisdom has long been that patent owners must be free to exercise a right to exclude potential users in precisely the manner that was not permitted to copyright holders. However, such conventional wisdom may have to change in the face of patents now including works that are either overtly expressive or that lie on the uncertain border between function and expression.
Conclusion
Software does not fit comfortably into the traditional legal categories for functional works or expressive works. The law of copyright, then of patent, and now of the First Amendment have been strained past recognition in order to accommodate this technology. Perhaps the most sensible resolution to this difficulty would be to place software into its own unique category—not of patent, copyright, or even protected speech—but one specially designed to fit the characteristics of this unusual legal hybrid. Unfortunately, it is likely too late to turn back the clock and take such a sensible step; so we may expect that legal institutions will continue to classify as either "functional" or "expressive" a technology that may be neither or both.
Join the Discussion (0)
Become a Member or Sign In to Post a Comment