Conventional wisdom used to have it that algorithms are not patentable, so computer software is not patentable. More recently, you may have heard that software patents are extremely costly, take years to acquire, and are unnecessary inasmuch as software is protected by copyright. Today, if you were to talk to a patent lawyer, you would hear that patents are simply another tool for proecting computer software, useful, in fact for protecting aspects of computer software that cannot be protected by copyright or trade secret.
The U.S. Patent and Trademark Office maintains that it does not issue patents for computer programs. It does, however, issue patents for program-related inventions. What this means is that software processes within programs are individually patentable. As an example of such a process, consider Claim 1 from Patent No. 4,674,040, Merging of Documents, issued June 16, 1987.
Competition has driven software publishers to incorporate thousands of such functions in a single applications package. Every such function risks infringing someone else's patent. Even if it is implemented in a different way from the technique disclosed in the patent, it infringes as long as it fits the elements of the claim (unless, of course, the patent is proved invalid in court).
The truth is that patent is not just another protective tool that developers are free to take or leave. As long as patents are being sought and granted, patents must be licensed or avoided. Today, everybody must play by the rules of patent.
Independent creation is not a defense to patent infringement - as it is for copyright infringement. This means that you can never be sure that what you create is yours. You are charged with knowledge of all the relevant patents that have been issued. so you must research before you create - a radical change from the paradigm of authorship under copyright.
What is worse, you are, in effect, also charged with knowledge of patents that are not yet issued. Software patents are typically in the pipeline for two or three years before they are granted, and the entire application file remains confidential until the patent issues. While no damages may be assessed for infringement prior to issuance, that is small consolation to an infringing developer who may have a completed product on the market. Since the patent includes the right of use as well as the right of manufacture and sale, the patentee can go after the developer's old customers who are still using the software program, and these customers can in turn sue the developer.
For software developers. the rules seem to have changed suddenly. But in fact the patent office changed its practices eight years ago. In 1981, a 5-4 Supreme Court decision, Diamond vs. Diehr, held that a process for curing rubber, which included a computer program, was patentable subject matter. In the wake of this decision and other appellate decisions during the following year, the patent office, after many years of resisting, began to issue software patents liberally. Because of the delay inherent in the application process, this change of attitude was not immediately evident. By 1986, however, articles were appearing in legal periodicals that noted that patents were issued for everything from subroutines to applications functions to user interfaces.
Software publishers were slow to pick up on these developments. They were not interested in patents, because copyright seemed to offer as much protection as they needed at very little cost. Only in 1988, when letters claiming infringement (or the possibility of infringement) started appearing with disturbing regularity, did a few industry leaders sound an alarm.
In the spring of 1989, newly awarded patents were making news. Headlines appeared in the Wall Street Journal - "Will Software Patents Cramp Creativity?"{1} - and the New York Times - "Software Industry is in an Uproar over a Rush of Patents."{2}
In August, REFAC International Ltd. filed a patent infringement suit against six major spreadsheet publishers-Lotus, Microsoft, Borland, Ashton-Tate, Computer Associates, and Informix. The patent in question (No. 4,398,249) had been filed in 1970 and granted in 1983; it purports to claim a technique called "natural order recalc", which is commonly used in spreadsheets. REFAC does not design, produce, or distribute software; in fact, REFAC is in the business of acquiring and litigating patents. Other suits have been filed over software patents, but none strikes so spectacularly at the heart of the software publishing industry or is so disturbing in its implications.
WHAT'S THE PROBLEM
The patent bar claims there is no problem. True, the patent office is not quite up to speed on issuing patents, but this is always a problem in new industries. Just give them time and everything will be fine. There's basically no difference between software and other industries, no reason to treat software differently. The days when talented hackers wrote software in their attics are over; the industry is growing up, and it will eventually appreciate the patent system - like everybody else.
But the situation is unique. It is well accepted that the importance of patent protection varies considerably from industry to industry. In the case of software, copyright was and is widely accepted, domestically and internationally, as an effective form of protection (in addition to trade secret). There is no evidence that investment has been lacking, that patents are needed to induce investment. Just as the President's Commission on the Patent System commented in 1966:
Indeed, the software industry grew spectacularly over the next two decades without relying on patents. It may be that the industry has developed in a way that is fundamentally incompatible with a patent regime. Perhaps the train left the station 30 years ago.
Even in the very beginning of the software industry, few programmers sought patent protection, and the patent office simply never developed the means to deal with computer programs. Accordingly, the 1966 Presidents Commission on the Patent System recommended against patents for program processes:
Words that could as well have been written in 1989. The patent office's reluctance to incur what even in 1966 appeared to be an enormous administrative burden meant that nobody looked to the patent office for information about software. The noble objective of the patent system is to encourage public disclosure in exchange for a limited monopoly. But the disclosure provided by the few software patents granted had the significance of trees falling in a deserted forest.
Thus, the two fundamental principles behind the patent system - the incentive and the quid pro quo - seem nearly inoperative in the case of software. On the other hand, the emergence of software patents is riddled with problems that extend from the practical to the philosophic. Some of these problems are inherent in the patent system but are exacerbated to an extraordinary degree by the nature of software and the nature of the software industry.
CAPABILITY OF THE PATENT OFFICE
It is widely believed within the industry that most software patents are defective because there is prior art (software or code created earlier) or because the claims are obvious in view of the prior art. It is assumed that the patent office cannot afford to hire top-quality examiners in the field, that the best examiners leave, and that the point system, which determines promotion of examiners, encourages the granting of patents with little regard for quality. Since software patents nonetheless seem to take an average of 30 months from application to the publication of the patent (compared with 20 months for all patents), presumably quality would deteriorate further if efforts were made to bring the processing time in line.
It must be remembered that the entire application process is conducted in secret, strictly between the applicant and the patent examiner; that the standard of nonobviousness is a person of ordinary skill in the art; that the critical decision is made by a bureaucrat in Washington, rather than a peer in the industry. Furthermore, patent denials are subject to review by the new, notoriously pro-patent Court of Appeals for the Federal Circuit, which also hears all appeals from the district courts in patent cases. In virtually all important patent cases, a patent-lawyer-turned judge sits on the panel - a practice not unlike permitting a lawyer to serve on a jury.
The problem of locating prior art goes beyond the lack of a base of software patents. The literature in the field is scant and unorganized. There was long a dearth of scholarly literature in the field, and there is still no bibliographic database. Without printed publications, it is very difficult to prove prior art. In addition, many programs that are used commercially are subject to trade secret restrictions against reverse engineering. When a software process is hidden and contractual restrictions preclude its disclosure, it does not qualify as prior art.
SOFTWARE AND THE SOFTWARE INDUSTRY
Software is extremely complex. As noted, a single applications program may contain thousands of patentable processes. On the other hand, this complexity is relatively easy to create. No machines have to be tooled; no raw materials have to be acquired. An individual with a computer can create the entire program - often entirely from scratch - without incorporating preexisting modules and without consulting manuals or reference works.
In short, the barriers to entry are very low and the market is highly competitive. Dozens, even hundreds of developers may be working on the same process, which looks like over investment from a patent perspective. Under the patent system, only the first to invent and file receives a patent, which enables that developer to prevent all the others from realizing the fruits of their investment. The waste of labor and other resources is a product of both the multitude of independent efforts and the delay inherent in the system. The fact is that these independent efforts continue, the results are incorporated into products, and the products are marketed before the blocking patent is granted and published. Thirty months is a very long time in the software business, longer than the average product cycle.
In addition, all developers, whether or not they are filing for patents, must bear substantial direct costs of searching patents to avoid infringement (even though such searches will be inconclusive), plus the costs of licensing and/or defending against patents held by others. For those who choose to file patents, the ante will be $10,000-$25,000 in direct costs, but this is likely to be only a fraction of the costs in staff time required to properly prepare and prosecute a patent application. For both plaintiffs and defendants, the costs of litigation can easily reach $250,000 just to get through the discovery phase; a full trial may run millions of dollars. These are, of course, fixed costs that are independent of the size of the company.
The patent bar is urging the strategy of defensive patenting - the principle that you must have patents of your own to protect yourself against the patent claims of others. In industries where the market is cornered by a few large players, cross-licensing is a common and efficient practice. But the software industry has thousands of firms of all different sizes. And the principal threats are not from established companies but from companies like REFAC, which have no interest in developing software products and hence no desire to cross-license. While defensive patenting may have speculative value, its principal effect is to accelerate the development of the patent minefield in what amounts to an arms race.
To the extent that patenting takes hold, the effect or small developers will be devastating, and the result will be consolidation of the industry into a few large firms that have the resources to play the game. In fact, most of the companies that have been getting software patents are hardware companies - and now nearly 40 percent are Japanese companies. Quite possibly, the software industry will cease to exist as an independent industry. Developers will need protection from the patent portfolios of hardware companies and the claims of maverick companies with strategic patents; they will be assured of protection only by becoming part of large hardware companies.
While it is impossible to prove empirically, it is widely accepted within the software community that there are diseconomies of scale in software development, that the best, most creative products come from individuals and small teams. Under a patent regime, software development is no longer a place for the individual author with nothing but good ideas and the skills to implement them. It becomes instead an industrial enterprise in which the cost and risks of avoiding and claiming patents are the price of doing business. The consequence will be a loss of diversity - in innovative firms and in the products available to users. Only the very largest companies will be able to innovate rationally; venture investment will be diverted from developing solid market-oriented products to speculation in strategic patents.
SOFTWARE vs. HARDWARE
The real issue is not hardware vs. software. As patent attorneys are fond of pointing out, anything implemented in software can be implemented in hardware. The real issue is broad, abstract patents that cover how information is processed and presented. Nowadays, such processes are usually implemented in software.
Sometime next spring, after much delay, the U.S. Patent and Trademark Office may issue a patent for Cathartic Video. The application. filed in 1965 in the early days of public television by a programmer at WGBH-TV in Boston, shows an implementation using Sophocles' "Oedipus Rex", performed by college students and a black-and-white television set. However, the claims are broad enough to cover almost any classical tragedy on television or home video (which, of course, was unknown at the time). Consider Claim 11:
Incredible, you say? True, there probably is no such application pending. (But there is no way to find out since applications are confidential). However, consider Claim 11 of a very real patent that issued in 1979 for an Interactive Teaching Machine, No. 4,170,832.
The prose used in No. 4,170,832 lacks the stylistic clarity of the WGBH hypothetical but strip away the redundancies, and the underlying concept is revealed as perhaps even more elegant than cathartic video.
The actual implementation of this patent is the combined videotape player and television with six push buttons shown in Figure 1, but the claim seems to cover a very commonplace process in interactive video whether implemented on videodisc or using digital video on a computer.
On a different scale, consider a patent for system, which may be relevant to the proposed National Research and Education Network.{3} Patent No. 4,768,087, "Education Utility" issued August 30, 1988, claims:
The disclosed implementation, shown in Figure 2, is a satellite broadcast system with a telephone return, but the claim covers use of packet-switched data networks.
These examples (and many of the other patents listed in the table, previous page) show both the sweeping nature of claims being allowed by the patent office and the great disparity between the implementation disclosed and the scope of the claims, most of which could be and probably have been implemented in any number of different ways. Here, again, the presupposed trade-off between public disclosure and the grant of monopoly is inoperative: The grant of monopoly far exceeds the disclosure, and the process claimed is self-evident, i.e., it could not possibly be concealed as a trade secret. Thus, the patent office appears to he granting monopolies to whoever first implements common business, educational, and research functions on a computer, and the patentee gives nothing in return but a description of one way of implementing the function.
HIGHER EDUCATION'S STAKE
There are very different perspectives on software patents within higher education. Ronald Bracewell, an electrical engineering professor at Stanford, has obtained a patent on a method of computing Fourier series, No. 4,646,256, "Computer and Method for the Discrete Bracewell Transform", which has been assigned to Stanford University. This patent has stirred controversy because it appears to violate the Supreme Court's rulings on the patentability of mathematical algorithms.{4} In certain specialized fields, where researchers have a fairly clear sense of what is state-of-the-art (and where the standard of ordinary skill is therefore very high), some universities are encouraging the patenting of algorithms. Other universities and researchers oppose this practice and encourage early publication to defeat the possibility of patenting.
When it comes to the development of software to support research and education, universities are essentially in the position of small software companies, that is, the scale of investment is typically small, and the institution cannot normally afford the costs and risks of navigating a patent environment. This means that the commercialization of university-developed software will be inhibited, and that software-intensive cooperative enterprises (such as the National Research and Education Network) will be jeopardized by the risk of patent infringement.
Universities are also prodigious consumers of commercial software. The advent of software patents will discourage acquisition of software from smaller companies that cannot effectively indemnify universities for patent infringement. This means less choice, hence less efficient use of software in education and research.
Furthermore, the costs of operating in a patent environment must be paid by someone. In the short run, the costs may come from industry operating margins. However, in the long run, after the industry shakes out, users will have to foot the bill.
The irrationality introduced by patents also risks upsetting the work done by EDUCOM and many individual schools to develop an atmosphere of respect for intellectual property in the electronic environment. Example after example seems to show that the patent office has issued patents that are ludicrous in breadth and ignorant of common programming practices. Worse, a case can be made that the office is proceeding oblivious to its own limitations, that it has failed of its basic purpose, and that it has proven more responsive to the patent bar than the needs of the industry, let alone the public interest.
More generally, while copying fits closely with academic values supporting originality and condemning plagiarism, patent is based on the primitive ethic of "first in time is first in right". In practice, the patent system seems to lend itself to abusive tactics such as broadcast mailings of threatening letters and harassment of companies through their customers. If patent law is to have moral as well as legal force, its excesses must be pruned and its ambition must be no greater than the system can sustain.
Because of their vital role in the creation and dissemination of knowledge, universities need to be concerned not only about the impact on their operations but about first principles. The use of patents to control information processes raises policy questions that fundamentally challenge the assumptions of the patent system. How should the flow of information be regulated in a free society? What functions of information technology and infrastructure should be franchised through patents? Does freedom of expression extend to functionality as well as content? To interaction as well as linear exposition? Is the patent system meeting its constitutional mandate - "To Promote the Progress of Science and Useful Arts" - or has it lost its bearings in a digital jungle?
The House Subcommittee on Courts, Intellectual Property, and the Administration of Justice under Congressman Robert W. Kastenmeier has started oversight hearings on software protection, and it appears likely that the U.S. Congress Office of Technology Assessment will embark on a new study that will, among other things, scrutinize developments in software patents. The issues are complex, but higher education's enormous stake in information technology demands that the academic community take the lead in developing policies and practices that it, the nation, and the world, can justify and respect.
NOTES
1. March 14. 1989, p. B1
2. May 12. 1989, p. 1.
3. Patent infringement can be a significant risk even to federal activities. See, for example, Edmund L. Andrews, "Patent Case May Cost U.S. Billions", New York Times, April 22, 1989, p. D1.
4. See Edmund L. Andrews, "Equations Patented; Some See a Danger.", New York Times, February 15, 1989, p. D1: Edward N. Zalta, "Are Algorithms Patentable?", Notices of the American Mathematical Society, Volume 35/No. 6 (July/ August 1988). pp. 769-799.