If you're thinking about getting involved in e-commerce, you'd best be aware that you've missed your chance. Thanks to a 1996 U.S. Federal appeals court decision (State Street Bank & Trust vs. Signature Financial Group), it's now possible to obtain broadly phrased patents on Internet business processes. These patents enable a firm to obtain what amounts to a government-guaranteed monopoly on the abstract concept of how to carry out a certain type of business on the Internet. Since then, the U.S. Patent and Trademark Office (PTO) has awarded hundreds of patents for e-commerce business processes, including Amazon's patent on its one-click ordering technology. As the Amazon case illustrates, businesses that obtain such patents will not hesitate to use them against competitors—and Amazon would just love to get rid of that pesky upstart, Barnes & Noble, which wasn't quite as quick to realize the potential of selling books on the Internet. Within the next few months, savvy patent seekers are sure to win patents on every conceivable method of doing business by e-commerce. If you're not aware of such patents, and start making some money, you'll hear from their lawyers. Count on it.

Defenders of the U.S. patent system concede that some recent patents seem too broadly phrased. They point out that the courts may overturn some of the recently granted Internet patents. Of course, that's no consolation to those who may have to give up their businesses rather than face insolvency in prolonged patent litigation; the cost of contesting a patent infringement generally exceeds $500,000. But you should be aware of an even more chilling possibility: namely, that the courts may very well sustain such patents. That is precisely the conclusion reached by legal scholar Francisc Keeley-Domokos (1999), and other prominent legal scholars concur (Melges 1999). After tracing the long, tortuous line of legal reasoning that led the Federal Court to the State Street decision, these scholars conclude that the Court's reasoning was sound—and that many, if not most, Internet patents will withstand a court challenge.

How could sound legal reasoning lead to a patent policy that so obviously undermines competition, discourages innovation, and distorts market dynamics in an important, emerging industry? As you'll learn in this essay, you don't get the answer until you trace this reasoning back to the earliest decision that enabled patent seekers to win government-guaranteed monopolies for software algorithms—that is, the first U.S. Supreme Court decision concerning computer software (Gottschalk vs. Benson, 1972). In this decision, the computer-illiterate judges committed an error of monumental proportion in defining one of the basic concepts of computer science, the algorithm. Their error led directly to the lower court decisions and PTO policies that transformed virtually all software algorithms into patentable subject property—and in the end, as you'll see, to the concept of patents on abstract Internet business patents.

Legal scholars will say that the long chain of legal reasoning since the 1972 Benson decision is too well-established to be overturned. But this is tantamount to saying that, if the U.S. Supreme Court once ruled that the earth is flat, then all subsequent courts must proceed as if this is true. In the end, I believe Internet patents can be successfully challenged only by laying bare the erroneous thinking that underlies the very concept of software patents. Please read on; everyone who understands essential computer concepts bears an important responsibility to educate the public, as well as our lawmakers, about what's truly at stake. If you're reading this from outside the U.S., please be aware that these patents can be written so that the U.S. can compel foreign countries to honor them. That's true even if your country's laws haven't yet been fully molded in the image of the out-of-control U.S. intellectual property system. The points I'm about to make matter to everyone, everywhere, who's concerned in any way about the future of computing and electronic commerce.

Let's start with a bit of background. As you're probably aware, the U.S. Constitution authorizes Congress “to promote the Progress of Science and useful Arts, by securing for limited Times to Authors and Inventors the exclusive Right to their respective Writings and Discoveries.” But what does this mean?

Judges and attorneys who favor software patents will argue that a literal reading of the Constitution discloses only that such rights should be limited in duration, but this is obviously false. The Constitution clearly states that that copyrights and patents should promote the progress of science and technology. It is also obvious that the Constitution means to distinguish intellectual property rights from the type of rights granted to owners of tangible property, such as real estate. In the “takings” clause, the Constitution plainly states that the government may not deprive citizens of their tangible property without paying fair compensation. However, no such compensation is owed when copyright and patent holders lose their protection. Therefore, the Constitution argues that Congress promotes science and technology not only by granting temporary monopolies to copyright and patent holders, but also by depriving them of these rights after a certain amount of time has expired. What the Constitution envisions, in short, is a thriving and growing public domain of ideas, knowledge, and techniques. In the public domain, no one person has a right to exclude others from using ideas, information, techniques, or knowledge in a particular way (Benkler 1999).

From the writings of the Framers and the earliest patent and copyright laws, as well as two centuries of jurisprudence, it is clear that the Constitution does not offer protection for “anything under the sun made by man”, as a more recent (and misguided) Congressional committee put it. Authors and inventors cannot seek protection for ideas, scientific truths, abstract concepts, and the mental steps that are the stock in trade of expertise in science and technology. These must always be part of the public domain. If intellectual property rights are expanded to the point that they reduce rather than enlarge the public domain, then they fail to measure up to the expressed constitutional standard, i.e., promoting the progress of science and technology. How effective would your college education have been if your professors had to pay licensing fees every time they taught you a fact or a mathematical formula?

Copyright law serves the constitutional purpose admirably. As you may be aware, copyright protects only the expression of an idea, not the idea itself. But patents are risky. A patent seeker may try to win a patent, not only on the practical implementation of an idea but also the idea itself. This is precisely what telegraph inventor Samuel Morse tried to do in 1853. After winning seven patents on telegraph-related devices, Morse sought an eighth patent on the abstract concept of using electromagnetism to produce printed characters at a distance. The U.S. Supreme Court refused to award the eighth patent to Morse, pointing out that it amounted to a patent on an abstract concept—that of electromagnetism. Such a patent could not serve to promote science and technology. It would do the opposite, because it would enable Morse to levy what amounts to a tax on any new method of electromagnetism for communication purposes. An inappropriately broad patent, in sum, fails to promote science and technology because it enables the patent holder to eliminate competitors who may very well have come up with better technology. For the same reasons, the courts have long refused to grant patents on abstract business concepts, such as a certain type of accounting system that is required in order to do business in a specific field. Such a patent is inimical to the public good because it enables the patent holder, in effect, to win a monopoly on a specific type of business.

In short, what is patentable is not the idea itself, but rather the practical application of an idea in a genuinely innovative way. As the U.S. Supreme Court has recently affirmed, Congress and the Patent Office may not enable patent seekers to obtain patents that, in effect, remove existing knowledge from the public domain. Such knowledge includes the truths of science, which are given by nature and, as such, cannot be discovered. An inventor may claim to have made a considerable investment in research that disclosed a new scientific truth, and seek to win patent protection for the truth itself, not just the practical application of this truth. But such actions would impede the progress of science. For this reason, the U.S. courts have long held that scientific truths, even if newly encountered, must be treated as if they are, in principle, prior art, and therefore not subject to patent protection.

Still with me? OK, you're about to find out how we got into this mess. Let's start with patent attorney Giles S. Rich, who in 1952 chaired a congressional commission that rewrote the Patent Act.

The Patent Act's language, dating back to 1790, enumerated patentable subject matter in the following terms: “any new or useful art, machine, manufacture, or process of nature”. In place of “art”, the Giles-led commission recommended that Congress use the word “process” instead. This was no insignificant updating. The history of technology reveals that the term “art” nicely captures the Constitution's intent in that, in this context, it refers to the non-scientific skills and techniques that a successful practitioner uses. In contrast, “process” could include scientific processes, such as those employed in industry. It seems reasonable enough to extend patent protection to industrial processes, but you shouldn't be naïve about Giles' intentions. In the early 1950s, Giles embarked on a single-minded, decades-long campaign to greatly expand the scope of patent protection to “anything under the sun made by man”, as the Giles-led commission put it. Subsequently, Giles fought for the legislation and jurisprudence that enable biotechnology firms to win patent protections for engineered life forms.

Enter computer software—and a great deal of confusion. In a computer program presented for patent protection, it is far from obvious (especially to those new to computing) just how to separate the abstract concepts from their practical application. In the 1972 Benson decision, the computer-illiterate judges tried to deal with a program that converted decimal numbers into binary numbers. Doing its job as charged, the Court attempted to determine whether the patent seeker was attempting to win a patent, not only on the specific application of the underlying idea, but on the idea itself. In its analysis, the Court determined that the underlying idea was the program's algorithm, which it defined as “a procedure for solving a mathematical problem”. The Court rejected the patent, pointing out that granting patent protection to this “mathematical algorithm” would be tantamount to granting a state-guaranteed monopoly on a scientific truth. Scientific truths are not subject to patent protection, the Court affirmed.

The Benson decision paved the way for the patentability of virtually all software algorithms. Almost immediately, the lower courts (and subsequently, the Patent Office) interpreted the Benson decision to mean that only “mathematical algorithms”, those that solve an equation for numerical purposes, are excluded from patentability. All other algorithms became subject to patent protection, even if they appeared to use mathematics. The key determining factor in an algorithm's patentability, the lower courts concluded, was whether the algorithm was intended to perform a numerical computation rather than control or supervise a practical process. In the end, the courts' reasoning led them to embrace patents for most mathematical algorithms as well, excepting only those with no known practical application, such as (presumably) an artificial life simulation.

What's wrong with the Benson decision is immediately obvious to anyone with a modicum of computer literacy. As Turing proved in 1937, there is no defensible boundary that separates mathematical from non-mathematical algorithms. On the contrary, any mathematical or logical problem that is capable of solution can be solved by a simple symbolic processor, which is capable of only a few basic operations, such as advancing and retracting a strip of paper and making or erasing marks on this paper. In short, there is no boundary that clearly separates mathematical from non-mathematical algorithms, such that the mathematical algorithms constitute a program developer's stock in trade of abstract concepts. On the contrary, the programmer's stock in trade consists of algorithms, most of which are used for purposes other than solving equations. For example, when you study programming, you examine a whole series of algorithms that perform tasks such as sorting and searching. Properly defined, an algorithm is a step-by-step procedure for solving a specified problem. Whether the problem is mathematical in nature or not has nothing to do with the level of abstraction involved.

To restrict the abstract conceptual knowledge of programmers and software engineers to mathematical algorithms is not only theoretically indefensible, but flies in the face of everyday professional practice. When you develop a program, you move from a functional specification, a statement of what the program is supposed to do, to a structural specification, a statement of how a particular algorithm, independent of any specific programming language, should be employed in the program's design. To accomplish this, programmers use paper-and-pencil techniques such as pseudo-code or flow charts, or programs that emulate these paper-and-pencil techniques.

Here's the key point. By restricting the abstract concepts of program development to a non-existent class of “mathematical algorithms”, the Benson decision enabled patent seekers, in effect, to win patents on the conceptual knowledge and mental steps that programmers use to create software. I'll return to this point in a bit, but first let's look at how the Court's reasoning in Benson led to Internet patents.

Dystopian novelist George Orwell liked to point out that, in the future, the government will twist and distort the meaning of words in order to enslave the citizenry. If you think he was nuts, read on.

Here's how we got to State Street, folks. The basic reasoning:

Incidentally, the judge who wrote for the State Street majority was none other than our old friend Giles S. Rich, then in his nineties. Rich survived long enough to interpret the very language that he himself had injected into U.S. patent law, namely, the substitution of “process” for the former word “art”. He died in June, 1999. The rest of us are stuck with the mess.

You can protest Internet patents until you're blue in the face, and boycott all you like, but you're wasting your time. The court's reasoning in the State Street decision is sound, if you accept the definition of “algorithm” laid down by the Benson decision. It's quite likely that the courts will affirm Internet patents. We're not going to get rid of Internet patents until the computer profession exposes the profound error that underlies this whole dysfunctional tradition of jurisprudence. That means getting rid of software patents, too.

Defenders of Internet and software patents will argue that strong intellectual property protection is needed for economic advancement. In this, they are correct, but they're leaving something out. Strong science is needed, too. In their boundless greed, some patent seekers will try to lock up the abstract concepts, mental processes, and scientific truths used by practitioners so that they can effectively monopolize a certain way of doing business. This is precisely what Amazon has achieved in its patent on one-click technology. The company defends its patent by stating that it spent thousands of hours developing the one-click method, which boils down to the use of garden-variety web programming to retain customer information from session to session. But all e-commerce sites spend thousands of hours developing their pages, and all e-commerce sites use cookies to retain customer information. What Amazon has obtained is a patent on the conceptual knowledge and mental processes that Web developers employ every day to create e-commerce sites.

Amazon's patent could be contested by looking for evidence of prior art, which would invalidate the company's patent claim; but such a solution is piecemeal. It does not address the underlying problem, which is essentially this: in obtaining patents on any software algorithm, patent seekers are succeeding, in effect, in obtaining a monopoly on the mental processes and abstract concepts routinely employed by software developers. Our attack on the patentability of software algorithms should begin at this level and proceed in accordance with the First Amendment. The U.S. Constitution guarantees not only freedom of speech, but also freedom of thought. By obtaining broadly phrased patents that restrain others from using the stock-in-trade mental steps and concepts used to develop web sites, Internet patent holders are violating the constitutionally guaranteed freedom of software developers to engage in the thought processes characteristic of their profession.

Internet patent holders will reply that their algorithms are novel, and therefore deserve protection. They are, after all, easily copied by others. This argument must be refuted as well, and once again, Turing points the way. His 1937 paper proves that algorithms, like the laws of physics, are timeless, immutable truths given by nature. If this were not the case, Turing's equations could not have shown that a certain class of problems exists which algorithms (and by extension, computers) will never be able to solve. Turing was able to project the science of algorithms forward to the point at which he could discern its ultimate limitations, just as Einstein was able to discern, thanks to his equations, that it is not possible to travel faster than the speed of light. Neither prediction could have been made unless both were dealing with the underlying truths of nature. Either you see this, or you don't; and if you don't, go back to college.

As two centuries of U.S. jurisprudence establish beyond question, the truths of nature cannot be subject to patent protection. Any such truth that is discovered in the process of technological development should be regarded, therefore, as if it were prior art. In short, there's a better way to combat Internet patents than chasing after old journals and Internet posts in search of prior art. Because algorithms are part of the truths given by nature, all algorithms, even novel ones, are prior art by definition. It therefore follows that there is no such thing as a valid patent on a software algorithm, including one for which no evidence of previous practice can be found to exist.

Internet patent holders will further argue that patents are needed to compensate those who try to discover new algorithms. This is the most absurd argument of all—and it is deeply offensive, as well, coming as it does from companies that may very well have accomplished little more than blending an existing business practice with the level of web publishing knowledge found in an introductory textbook. Most of the algorithms now in existence were invented before they became eligible for patent protection. Patent protections were not needed then. They are not needed now. There is no evidence that withdrawing patent protection for software algorithms would discourage further scientific progress in this area.

The whole lengthy, dysfunctional, Orwellian edifice of legal reasoning that supports Internet patents has a huge central flaw. It rests, ultimately, on a monumental misunderstanding of the science of algorithms (Newell 1986). The computing community must explain this to the public and lawmakers. And we must speak with a unified voice: Abolish software patents, and abolish them now.