SOFI

SIMSON L. GARFINKEL firm RICHARD M. STALLMAN mos MITCHELL KAPOR serve their case! sole l Lotu shee Why Patents Are Bad patei whic whei for Software since cept ticip, one Patents can't protect with or invigorate the pear! computer software ligen sheel "i% Septembeer 1990, users of the industry; they can longer be able to automatically cor- thep ,opular XyVVrite word processing only cripple it. rect common spelling errors by near] ...... 4-a M _ in Ai~~~~trhints otter ~~~~nroc~cin a th~ae not' e ht ft- tp program got a Ulstur1ll1 ltLLUtI Il F",aall~r, C11%,aF"`1 u"L 1 LL L%. gram . "~ T .1 " "-----" ·-- -'-·-L----- - the .tmail trom Xy1uest, Inc., me mlsspelle word. In audtion, to ex- is exi program's publisher: pand abbreviations stored in your so el personal dictionary, you will have prob] "In June of 1987, we introduced an automatic cor- to press control-R or another designated hot key." r rection and abbreviation expansion feature in XyWrite netw III Plus. Unbeknownst to us, a patent application for a XyQuest had been bitten by a software patent- on a related capability had been filed in 1984 and was sub- one of the more than two thousand patents on com- thou sequently granted in 1988. The company holding the puter algorithms and software techniques that have pater patent contacted us in late 1989 and apprised us of the been granted by the U.S. Patent and Trademark Office softm existence of their patent. since the mid- 1980s. The owner of the patent, Produc- comI We have decided to modify XyWrite III Plus so tivity Software, had given XyQuest a choice: license any c that it cannot be construed as infringing. The newest the patent or take a popular feature out of XyWrite, is de, version of XyWrite III Plus (3.56) incorporates two XyQuest's flagship product. If XyQuest refused, a I significant changes that address this issue: You will no costly patent-infringement lawsuit was sure to follow. appli Some choice. legal XyQuest tried to license the patent, says Jim Adel- com] son, vice president for marketing, but Productivity poter Software kept changing its terms. First Productivity Simson Garfinkel is a senior editor at NeXTWORLD magazine fecti) (San Francisco) and the coauthor of the book PracticalUNIX said that XyQuest could keep the feature in some ver- estab Security. Richard M. Stallman is one of the founders of the sions of XyWrite, but not in others. Then the company that v ~,J.,eague for Programming Freedom (Cambridge, Massachusetts) said that XyQuest could use one part of the "inven- d the recipient of the Association for Computing Machinery's tion," but not other parts. And Productivity Software ,race Hopper Award. Mitchell Kapor, a founder of the Lotus Wh. Development Corporation, is president of the Electronic Frontier kept increasing the amount of money it wanted. Xy- Softl Foundation (Cambridge, Massachusetts). Quest finally gave up and took the feature out. stead

'50 ISSUES IN SCIENCE AND TECHNOLOGY FALL 1P .4w SOFTWARE PATENTS

XyQuest was lucky it had that option. Other structions that tell a computer how to carry out a firms-including some of the nation's largest and specific task in a program. Thousands of instructions most profitable software publishers-have been make up any one computer program. But whereas the served with notice of patents that strike to the heart of unique combination of algorithms and techniques in a their corporate vitality. In one of the most publicized program is considered an "expression" (like a book or cases, a company called Refac International-whose a song) and is covered by copyright law, the algorithms sole business is acquiring and litigating patents-sued and techniques themselves are treated as procedures Lotus, Microsoft, Ashton-Tate, and three other spread- eligible for patenting. sheet publishers, claiming they had all infringed on The judicial basis for this eligibility is tenuous at patent number 4,398,249, which spells out the order in best. U.S. law does not allow inventors, no matter how which to recalculate the values in a complicated model brilliant they are, to patent the laws of nature, and in when one parameter in the model changes. (Refac has two Supreme Court cases (Gottschalkv. Benson, 1972, since dropped its claims against all the companies ex- and Parker v. Flook, 1978) the Court extended this cept Lotus, but only because company lawyers an- principle to computer algorithms and software techni- ticipated a better chance of success if they faced just ques. But in the 1981 case Diamond v.Diehr, the Court one opponent.) said that a patent could be granted for an industrial Patent 4,398,249 does not have anything to do process that was controlled by certain computer algo- with spreadsheets in particular; the technique also ap- rithms, and the Patent Office seems to have taken that pears in some graphics drawing and artificial intel- decision as a green light on the patentability of algo- ligence programs. And the idea that values in a spread- rithms and techniques in general. sheet should be recalculated in the order specified by Software patents are now being granted at an the patent is so obvious that it has probably occurred to alarming rate-by some counts, more than a thousand by nearly everyone who has written a spreadsheet pro- are issued each year. Unfortunately, most of the he gram. But the Patent Office's standard for obviousness patents have about as much cleverness and originality ,x- is extremely low; patents have been granted for ideas as a recipe for boiled rice-simple in itself but a vital }ur so elementary that they could have been answers to part of many sophisticated dishes. Many cover very yve problems in a first-year programming course. small and specific algorithms or techniques that are Practically once a month, the nation's computer used in a wide variety of programs. Frequently the "in- networks are abuzz with news of another patent issued ventions" mentioned in a patent application have been on a fundamental concept that is widely used. Al- independently formulated and are already in use by m- though the Patent Office isn't supposed to grant other programmers when the application is filed. ive patents on ideas, that's essentially what it's doing with When the Patent Office grants a patent on an algo- ice software patents, carving up the intellectual domain of rithm or technique, it is telling programmers that they 1C- and handing little pieces to virtually may not use a particular method for solving a problem : --any ------rnmnnnvI------that------files----·------an adnlication. And the practice· without the permission of the idea's "owner." To them, ite,te, is devastating America's software industry. patenting an algorithm or technique is like patenting a 1,I, a If Congress does not act quickly to redefine the series of musical notes or a chord progression, then )w.M.. applicability of patent law to computer programs, the forcing composers to purchase a "musical sequence legal minefield confronting the introduction of new license." el-l- computer programs will be so intimidating-and ity potentially so costly-that small companies will ef- Systems at odds 'ity fectively be barred from the marketplace, while large, The traditional rationale for patents is that protection er- established firms will become embroiled in litigation of inventions will spur innovation and aid in the dis- my that will have a stultifying effect on the entire industry. semination of information about technical advances. an-m-- By prohibiting others from copying an invention, are What's being patented? patents allow inventors to recoup their investment in CY-Cy- Software patents do not cover entire programs; in- development while at the same time revealing the stead, they cover algorithms and techniques-the in- workings of the new invention to the public.

.OGY FALL 1991 S1 so]

But there's evidence that the - mon. In another case, the journal acc patent system is backfiring in the IEEE Computer in June 1984 pub- WI computer industry; indeed, the sys- The Pate,nt-OJficeis lished a highly efficient algorithm vel tem itself seems unsuited to the up the for performing data compression; ma nature of software development. unbeknownst to the journal's ed- prc Today's computer programs are so intellectlgal I domain itors or readers, the authors of the col complex that they contain literally ofcomph r science article had simultaneously applied cot thousands of algorithms and tech- tte science for apatent on their invention. In me niques, each considered patentable and hariding little the following year, numerous pro7 the hv the Patent Aff-'.r .£to'ltAr, T. one onyAn ~ li es , WyLL CUIU WUkI U13 it reasonable to expect a software t o any tributed for performing the so- ha) company to license each of those Compan *hatfiles called "LZW data compression." pas patents, or even to bring such a The compression system was even legallymarketlace? risky To product make things into eventhe an apl)litcation. adopted as a national standard and pri o mpa Dronosed as an international one. fins morecomputerlicat complicated, theed, thePatent Of- Then, in 1985, the Patent Office pat fice has also granted patentspatdevelopments on awarded patent number 4,558,302 anc combinations of algorithms and techniquestechn that to one of the authors of the article. Now Unisys, the 6or produce a particular feature. For example, Apple was holder of the patent, is demanding royalties for the use wa! suedsued because its Hypercard program allegedlyalleged violates of the algorithm. Although programs incorporating the nor patent number 4,736,308, which covers a specific algorithm are still in the public domain, using these exi ~cechniqueechnique that, insimplified simplified terms, terms, entailsentail scrolling programs means risking a lawsuit. through a database displaying selected partspa of each Not only is the patent approval process slow, but mu line of text. Separately, thethe scrolling and displaydis func- the search for "prior art"-the criterion the Patent Of- Pat tionstit reasons are arbleubiquitous fixturesof of computecomputer program- fice uses to determine whether an invention already exit ming,m buting, combining them without a licenselicens from the exists at the time of a patent application-is all but im- cats holder ofof patent 4,736,308 is now apparentapparently illegal. possible to conduct in the realm of computer software. in t Another problemproblem with patenting softwaresofthe is the After more than 25 years, the Patent Office has not de- dec amountamount of time it takes to do so. The two tovento five years veloped a system for classifying patents on algorithms dor required to file for and obtain a patent are acceptableac if and techniques, and no such system may be workable. alla a companycompany isis patenting, thesay, the formula fofor Valium, Just as mathematicians are sometimes unaware that whichhasn't hasn't changed in mmore ore than 20 years.years But in the essentially identical mental processes are being used Ba software industry, companies that don't continuallyc in separate areas of mathematics under different ter- Eve bringbring out newions versionsof algoritheir of theirand programs techn go out of minology, different parts of computer science fre- pub business. Success forfeature. them depends Forthem on spottingspot needs quently reinvent the same algorithm to serve different or I and developingits solutionssolutionscauseHypercard as program quickly asas poslegepossible. purposes. It is unreasonable to expect that a patent ex- pro! Unfortunately, conductingconducting a patentvers search is a aminer, pressed for time, would recognize all such the slow, deliberative processthat, that, when harnessedhail to duplication. For example, IBM was issued a patent on whi software development,devparatelnt, couldstop stopinnovad innovation di in its the same data-compression algorithm that Unisys sup- plei tracks. And because patent applications are confiden- posedly owns. The Patent Office was probably not fror tial, there isubiquitous simply no wayfor for computercomputer programmerspro aware of granting two patents for the same algorithm pate to ensureto that what theythey writewrite will not violatevice some because the descriptions in the patents themselves are tec patentpatent that is yet to,736,308 be issued.is yet Thus XyQXyQuest "rein- quite different even though the formulas are mathe- vented" its automaticproblem spelling-error spelling-errorwits automatic correctioncorrecti system matically equivalent. tion ' -idad brought the the productto to market betweenbetweet the time The search for prior art is complicated by the fact sma aatrequiat ProductivityPred to file for Software and obit had patentfiled afor are its sac application that the literature of computer science is unbelievably soft and andbeen awardedawarded thethe patent. large. It contains not only academic journals, but also patc Such examplesexaoftwareindustry, companiesare becoming are increasiincreasiincreasingly corn- users' manuals, published source code, and popular net

52 ISSUES IN SCIENCE AND TECHNOLOGY FALL SOFTWARE PATENTS

accounts in magazines for computer enthusiasts. the cost of a patent search for each technique and com- Whereas alteam of chemists working at a major uni- bination of techniques that the new program uses versity might produce 20 or 30 pages of published could easily equal or even exceed that. And the cost of material per year, a single programmer might easily a single patent suit can be more than a million dollars. produce a hundred times that much. The situation be- "I'm not familiar with any type of ligation that is comes even more complex in the case of patented any more costly than patent litigation," says R. Duff combinations of algorithms and techniques. Program- Thompson, vice president and general counsel of the mers often publish new algorithms and techniques, but WordPerfect Corporation. But Thompson's greatest they almost never publish new ways of combining old fear is that software patents will wipe out young, in- ones. Although individual algorithms and techniques dependent programmers, who until now have been have been combined in many different ways in the the software industry's source of inspiration. Imagine past, there's no good way to establish that history. what happens, says Thompson, when "some 23-year- The inability to search the literature thoroughly for old kid who has a terrific idea in a piece of software is prior art is crucial, because unless an examiner can hammered by a demand letter from someone holding a find prior art, he or she is all but obligated to issue the patent." patent. As a result, many patents have been granted- As for aiding the exchange of information, the and successfully defended in court-that are not expansion of software patents could mean instead "original," even by the Patent Office's definition. It the end of software developed at universities and was simply the case that neither the patent examiner distributed without charge-software that has been nor the defendants in the lawsuit knew of the prior art's a mainstay of computer users in universities, corpora- existence. tions, and government for years. Many such pro- Some members of the commercial software com- grams-the X Window system, the EMACS text munity are now proposing the creation of a "Software editor, the "compress" file-compression utility, and Patent Institute" to identify software's prior art that others-appear to be in violation of existing patents. existed before 1980. But even if such an institute could Patents could also mean an end to public-domain catalogue every discovery made by every programmer software, which has played an important part in in the , it makes no sense to arbitrarily making computers affordable to public schools. There declare that only pre-1980 work is in the public is obviously no way that an author who distributes a domain. Besides, what would be the purpose? To program for free could arrange to pay for royalties if !: allow the patenting of nature's mathematical laws? one of the hundreds of techniques that were combined to create the program happens to be patented. Bad for business Few programmers and entrepreneurs believe that Even when patents are known in advance, software patents are necessary for their profession. Instead, the publishers have generally not licensed the algorithms impetus for patents on algorithms and techniques :qi or techniques; instead, they I try to rewrite their comes from two outside sources: managers of large I ?i,h1 programs to avoid using the particular procedure that c(oiWpiSff3ise-gpaents as a meaiiiri uiUIpiiulg the patent describes. Sometimes this isn't possible, in over their competitors without having to develop supe- which case companies have often chosen to avoid im- rior products, and patent attorneys, who see the poten- plementing new features altogether. It seems clear tial for greatly expanding their business. ii: from the evidence of the last few years that software Today, most patenting by companies is done to patents are actually preventing the adoption of new have something to trade or as a defense against other technology, rather than encouraging it. patent-infringement suits. Attorneys advise that And they don't seem to be encouraging innova- patenting software may strengthen competitive posi- tion, either. Software patents pose a special danger to tion. Although this approach will work for large com- small companies, which often form the vanguard of panies such as Microsoft, Apple, and IBM, small and software development but can't afford the cost of even mid-sized companies can't play in their league. A patent searches or litigation. The programming of a future startup will be forced to pay whatever price the !iii: new product can cost a few hundred thousand dollars; giants choose to impose.

FALL 1991 53 ii S

Copyright and - owes nothing to any previous work. trade secrecy No author in areas other than The best argument against the wis- The mo.st effective software is held to such an unrealis- dom of software patents may be course 'action tically high standard. history itself. Lotus, Microsoft, 70 The U.S. patent system was WordPerfect, and Novell all be- is to e IcZ9urage created because the framers of the came world leaders in the software C ..... _,n uhalls1chn intntrr nn thp ctrnnth CofgreSIst o amend wouldConstitution discourage hoped trade that secrecv.patents of their products. None of these pate?nt law When techniques are kept secret companies needed patents to se- d for commercial advantage, they cure funding or maintain their mar- to d rsa,Hlow may never become available for ket position. Indeed, all made their softwai reI )atents. others to use and may even be lost. fortunes before the current ex- But although trade secrecy is a plosion of software patents began. problem for software, as it is for Clearly patents are not necessary to other fields, it is not a problem that ensure the development of computer programs. And patents help to correct. for those who want more control over what they see Many of the useful developments in the field of as their property, the computer industry has already software consist of new features such as the automatic adopted two other systems: copyright and trade correction and abbreviation expansion feature in Xy- secrecy. Write III Plus. Since it is impossible to keep a pro- Today, nearly all programs are copyrighted. gram's features secret from the users of the program, Copyright prohibits the users of a software program there is no possibility of trade secrecy and thus no need from making copies of it (for example, to give to their for measures to discourage it. Techniques used inter- friends) without the permission of the individual or nally in a software system can be kept secret, but in the company that licenses the program. It prevents one past, the important ones rarely were. It was normal for company from appropriating another company's work computer scientists in the commercial as well as the and selling it as its own. But the existence of a copy- academic world to publish their discoveries. Once right doesn't prevent other programmers from using again, since secrecy about techniques was not a sig- algorithms or techniques contained in the program in nificant problem, there is little to be gained by adopt- their own work. A single software technique can be ing the patent system to discourage it. implemented in different ways to do totally different The place where trade secrecy is used extensively jobs; copyright only prohibits appropriating the actual in software is in the "source code" for programs. In code that a particular programmer wrote. computer programming, trade secrets are kept by dis- In general, copyrighting and patenting are thought tributing programs in "machine code," the virtually in- to apply to very different kinds of material: the former decipherable translation of programming languages to the expression of ideas, and the latter to a process that computers read. It is extremely difficult for that achieves a certain result. Until just a few years another programmer to glean from a machine-code ago, computer algorithms and techniques were widely program the original steps written by the program's seen as unpatentable. And as author. But software patents haven't done anything to policy analyst Brian Kahin notes, this is the first time limit this form of trade secrecy. By withholding the in history that an industry in which copyright was source code, companies keep secret not a particular widely established was suddenly subjected to pa- technique, but the way that they have combined :tenting. dozens of techniques to produce a design for a com- :: 4:i.Indeed, without conscious action by Congress or plete system. Patenting the whole design is impractical 0- -_+ +I -- _ __+ E· ^- -..· · I1 ~_+~l 2uplumv, OUIL, LMO SE lul luallivinaLLI IUM vV and ineffective. Even companies that have software vare publishing-if you write a program, you own patents still distribute programs in machine code only. vill change. The new rule will be that you might Thus, in no area do software patents significantly what you write-if it is so revolutionary that it reduce trade secrecy. SOFTWARE PATENTS

Reversing direction tee that the Court would decide to change Patent Many policymakers assume that any increase in intel- Office practice or to do anything about existing lectual property protection must be good for whoever patents. The most effective course of action, therefore, works in the field. As we've tried to show, this is as- is to encourage Congress to amend the patent law to as suredly not the case in the field of computer program- disallow software patents and, if possible, invali- he ming. Nearly all programmers view patents as an un- date those that have already been awarded. The its welcome intrusion, limiting both their ability to do House Subcommittee on Intellectual Property and the by. their work and their freedom of expression. Administration of Justice, chaired by Representative ret At this point, so many patents have been issued by William J. Hughes (D-N.J.), should take the lead by ey the Patent and Trademark Office that the prospect of scheduling hearings on the subject and calling for a 'or overturning them by finding prior art, one at a time, is congressionally sponsored economic analysis of the Ist. almost unthinkable. Even if the Patent Office learns to effect of software patents on the industry. ;a understand software better in the future, the mistakes The computer industry grew to be vibrant and 'or that are being made now will follow the industry into healthy without patents. Unless those who want iat the next century unless there is a dramatic turnaround software patents can demonstrate that they are neces- in policy. sary to the health of the industry, Congress should feel of The U.S. Patent and Trademark Office recently justified in eliminating this barrier to innovation. tic established an Advisory Commission on Patent Law ,Y- Reform that is charged with examining a number of o0- issues, including software patents--or what it prefers Recommended reading m, to call patents on "computer-program-related inven- Brian Kahin, "The Software Patent Crisis," Technol- ed tions." Unfortunately, the commission's subcommit- ogy Review (April 1990): 53-58. er- tee on software does not include any prominent Mitchell Kapor, Testimony at Hearings before U.S. he software industry representatives who have expressed House of Representatives, Subcommittee on for doubts about software patents. But the subcommittee Courts, Intellectual Property and the Administra- he is required to consider public comment. The com- tion of Justice, of the Committee on the Judiciary ce mission's final report is not due until August 1992, so (March 5, 1990). ig- there is still time to make one's voice heard. Pamela Samuelson, "Benson Revisited: Should Patent Pt- Although influencing the Patent Office might Protection Be Available for Algorithms and Other produce some benefits, the really necessary reforms Computer Program-Related Inventions?" Emory ,ly are likely to come only through intervention by the Law Journal (Fall 1990): 1025-1154. In Supreme Court or Congress. Waiting for Court ac- Pamela Samuelson, "Should Program Algorithms Be is- tion is not the answer: No one can force the Supreme Patented?" Communications of the ACM (August in- Court to rule on a relevant case, and there is no guaran- 1990): 23-27. ,es 'or de i's1'S to he tar ed m-M_ :al,at Iretre ly. tly

FALL 1991 55