Utility, Innovation and Mathematics Jo Stanley1 Introduction The first section of this paper examines the history of the transition from physical (real world) inventions to ‘virtual’, computer-related ones. It explores the rationale behind the evolution of ‘hybrid’ inventions: those that embed computer software elements in physical environments in order to satisfy the demands of the United States Patent Office. It looks at the distortions engendered by attempts to maintain the ‘physicality’ of computer-related inventions, and how hybrids bring pathologies of their own. Section two examines the current definitions of, and assertions about patent utility, and presents the notion of virtual utility. Although today the battle for ‘virtual innovation’ is largely won (see recent decisions in State Street and Excel), the author has argued elsewhere that the judicial decisions that achieved this move are pragmatic and lack ontological conviction. From an ontological viewpoint little work has been done in the courts to demonstrate a coherent rationale for the change in policy. Instead, there has been a switch to utility as the load bearer of patentability in US law. It replaces the s101 patentable categories in importance. The final skirmish in the debate promises to be in the area of convincing the Patent Office and courts of the ‘virtual utility’ of virtual inventions. Utility carries strong overtones of physicality that are hard to contradict. This paper argues that such connotations make no more sense than the old arguments used against virtual innovation. It suggests that in the real real patent world representative arts such as software modelling of ‘real world’ systems deserve a place in their role as representatives. The final part of the paper looks briefly at the history of some extremely mathematical patents, those from the field of cryptography. Cryptography constitutes a longestablished class of patentable items in the US Patent Office, yet recently there have been complaints that some cryptographic inventions are so close to pure mathematics that they should never have received patents at all. I review some of these patents in the 1 Dr J.A. Stanley, Senior Lecturer Dept of Computer Science Anglia Polytechnic University, and visiting lecturer at Anglia Law School. 1 light of the thirty-year struggle that less mathematical patents have had to gain acceptance as patentable subject matter. 2 Part 1 Statute Law and Case Law Developments in Software Patents 1.1 The Law 35 United States Code, s101 addresses what is patentable. It says this: Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof may obtain patent therefor, subject to the conditions and requirements of this title. Similarly, the corresponding part of the European Patent Convention (paraphrased here) says this: A valid patent is given for the candidate invention having: Novelty (not being anticipated) Inventive step (not being obvious) Capability of industrial application (useful to the public) Both jurisdictions make certain exceptions to patentability. In America these have been established by judicial decision, in Europe they take the form of an exclusions list barring amongst other things: Discoveries, scientific theories and mathematical methods2. Schemes, rules, methods for performing mental acts … doing business and programs for computers3. The American exceptions can be classified as abstracta, and converge closely with their European equivalents. The US business method exception from patentability was felled by a recent decision4. 1.2 What is Patentable? The primary goal of the global patent culture is to foster useful, inventive solutions to acknowledged technical problems that will benefit the public. The history of computerrelated inventions indicates that for some time they were treated as non-technical 2 3 European Patent Convention (EPC) Art 52 (2)(a). EPC Art 52 (2)(c). 3 entities5. Computer programs are difficult to classify6 and liable to fall on the abstract side of patentable7 & 8. In the 19th century patents were predominantly awarded for useful applications of ‘natural forces’9, mostly as machines10 – highly tangible objects. Processes were regarded with more suspicion, being intangible until implemented, and being potentially broad because supportable on many platforms. Mechanical processes were executed by machinery, hence they must be mere ‘functions of a machine’ rather than innovative transformations in their own right11. Chemical processes were admitted to be patentable12. The unpatentable items specified in early American judicial decisions, and recited in all subsequent decisions on subject matter patentability were these: scientific truths13, ideas14, properties15 and natural phenomena (such as electricity as such)16. 4 State Street Bank and Trust Company v Signature Financial Group Inc., (149 F3rd [1998] at 1368). The US example of Johnston. (Appeals Court [1974]), banking software considered to support a ‘liberal art’. (502 F 2nd 765, 769). 6 They are ‘literary works’, hence subject to copyright, and ‘virtual machines’ hence potentially patentable. 7 The US Patent Office 1968 Guidelines: Examination of Patent Applications on Computers Programs (33 Fed Reg. 15, 609 – 610) barring programs reflect this view. These were hastily rescinded in 1969 (34 Fed Reg. 15, 724), following the Appeals Court decision in Prater (infra). But see also the subsequent Supreme Court decision in Gottschalk v Benson, which stigmatised programs as likely to be unpatentable items. This blight lasted for the next decade, (409 US 63 [1972]). 8 See also UK Banks Committee Report which advocated keeping a watching brief over the patentability of programs. Few UK cases surfaced at that time. 9 In a German case challenging the patentability of computer programs, the opinion said: ‘All these methods lack a basic requirement for a patentable invention, namely the creative utilisation of natural forces. (GRUR Int [1968] at 211). 10 Eli Witney’s cotton gin (patented 1794). As a process example: Charles Goodyear’s sulphur treatment of rubber to produce vulcanised product (patented 1844). 11 For example, the US decision on Tarcsky-Hornoch’s invention, a pulse counting circuit and method. (Filed [1960]. Appeal from a decision of the Patent Office, heard in the Appeals Court [1968], 397 F 2 nd at 856). Tarcsky-Hornoch citing the fact that chemical processes did not receive the patentability challenges applied to mechanical ones. The Tarcsky-Hornoch court concluded that up to that case the law on patentablity had been ‘..unclear …whether the dividing line marks a difference between means (patentable) and result (unpatentable) or chemistry and mechanics. (Tarczy-Hornoch at 859). The court explicitly threw out the allegedly confused prior law. 12 Tilghman v Proctor, (102 US 707 [1881]). Application of the hydrolysis of fats. Infringer did not use inventor’s apparatus and was therefore deemed not to infringe. The Court did not restrict the process to one apparatus but allowed it as a genuine process with scope ranging over a number of platforms. 13 Mackay Co v Radio Corps (306 US, 86, 94, [1939]). Invention of V-shaped radio antennae, the angle of the ‘V’ specified using Abrahams’ Law. The inventor owned to having simplified some mathematical expression in his disclosure, but did not declare Abrahams’ as being in the prior art. The Court held that the invention was patentable since it applied a scientific truth. 14 Rubber-Tip Pencil v Howard (87 US 498, 507, [1874]). The Court argued that the idea for erasers attached to pencils per se is not patentable. Basically, the claim failed for excessive breadth. 15 Le Roy v Tatham, (55 US 156, 175), [1852]. The invention was for improving the making of pipes by using lead. The pipes were improved by a property of lead: it melted and hardened on cooling. The court argued that this was a general discovery of a use for that property; not a specific process for exploiting it. 5 4 With the advent of computer technology in the middle of the last century, came an intensification of the desire of the Patent Office and courts to demarcate unpatentable ‘laws of nature’ and scientific principles from patentable inventions that applied these principles. Computer programs were seen as being on the cusp of this division. Computer programs proved elusive – intangible – in the patent context. Implementing processes as they did, they already incurred a black mark for being less substantial than machines. Electronic circuits with a useful purpose had no such problems with intangibility; they could be claimed as apparatus (the synonym for machine in the patent context)17. Yet software is logically equivalent and physically interchangeable with hardware18. A designer may choose a hardware implementation for speed19, or a software implementation for portability or adaptability. In short, the decision is made to satisfy purely functional criteria; the same broad purpose may be served by either solution. Also considered alarmingly radical was the fact that computer programs, dealing only with data - that pale surrogate for real world activity - should be patentable. The transformations programs achieve yield values in variables, not actual commodities. Worse, programs frequently use mathematics to implement their functions20. When it came to inventions with a mathematical heart to their novelty, the mathematics involved was allied to a mental process, as well as being too general or abstract to be patentable. The Patent Office took the ‘point of novelty’ approach, arguing as follows: 1. If a claim is to a process that is solely mental, then it is not patentable. 2. If the novelty of the claimed process resides in the mental steps, and the associated physical elements are old in the art, then the whole process is unpatentable. O’Reilly v Morse (56 US 62, 117). The Court argued that Morse had not invented electro-magnetism, but only a subset of the machinery that was enabled thereby to print at a distance. His claim therefore outstripped his actual invention. 17 All Tarcsky-Hornoch’s machine claims were accepted, only the method ones were rejected. 18 Prima facie it is. Many arguments are suppressed, however, by this statement. 19 Most of the early cryptography patents show algorithms implemented as chips. 20 In the past, the Patent Office has even used the argument that the operation of a computer at a level below that which the claimed computerised solution addresses provides evidence of the mathematical nature of the solution. Application of Walter, (618 F 2nd [1980] 758, 761). The Patent Office argued in Walter that a computer operates on binary signals. Base two arithmetic is imposed by the architecture of the computer itself. Since these manipulations are the way that a computer operates to effect a process, then this must mean that the claimed process is mathematical. 16 5 3. If the physical elements are new then the novelty of the mental element is immaterial, (don’t care), and the process is a candidate for patent21. The second ‘rule’ above epitomises the point-of-novelty approach. This was Patent Office policy when the first computer-related inventions started to appear as regular candidates in the 1950s. Patent examiners frequently strengthened their case by holding that if a process could be implemented by a mind or a machine then the claims to it were vague, (even if a machine implementation was described in the patent application22). The Appeals Court challenged the Patent Office’s point-of-novelty approach in Prater23. It argued that only purely mental processes were unpatentable24. In Prater no exclusively human judgement was involved. The Court further argued in Musgrave25 that if any process showed truth-functional steps (where output relies solely on inputs together with a clerical procedure) – which a program will - then it is potentially patentable. The Court discredited the point-of-novelty test, (rule 2 above), on grounds of its illogicality, and held that it should no longer control computer cases26. The Appeals Court continued in this vein through a number of seminal cases between 1968 and 1971. Towards the end of this series of cases the Appeals Court was confidently allowing claims to programs, so long as they were new and useful, simply because software technology was firmly in the mainstream of ‘the technological arts’27. 21 These are the infamous Abrams Rules. Abrams invention (1950s) was in the field of petroleum prospecting. There was a calculation using bore-hole pressure readings as input. It determined whether petrol-bearing strata were likely to be present in the vicinity of the bore-hole. The ‘method’ offered no actual (real world) way to determine pressures, and consequently the application failed. 22 See Reduction of Data From Spectral Analysis. Charles Prater. (Filed [1961]). 23 First Appeals Court hearing of Application of Charles Prater. (Prater 1, [1968]. 415 F 2nd at 1378). Rehearing, (Prater 2, [1969], 415 F 2nd at 1393). 24 A purely mental process may well be one that shows a propositional attitude: believing, fearing or hoping about a proposition. This ‘aboutness’ is expressed as ‘intentionality’ by John Searle, who uses the term to characterise a human mind as compared to a machine. (Machines do not fear or hope or desire about a procedure that they carry out). Intentionality: An Essay in the Philosophy of Mind, Searle, [1983]. In any case, ‘purely mental acts’ are private to the mind that carries them out. They cannot be inspected, let alone used, by another person. Therefore they cannot have the required utility to be patentable. 25 Corrections for Seismic Data Obtained from Expanding Spread. Filed in 1965. (Appeals Court hearing: 431 F2nd 882 [1970]). 26 Said the court: ‘A given process including both “physical” and “mental” steps could be statutory during the infancy of the field of technology to which it pertained, when the physical steps were new, and non-statutory at some later time after the physical steps became old’. That is, the statutory nature of the invention varies across time, which is clearly false. (Musgrave at 889). 27 That term is used in Foster [March 1971], 438 F 2nd at 1011, Benson, [May 1971], 441 F 2nd at 682, and McIlroy [May 1971], (three weeks after Benson), 442 F 2 nd at 1397. 6 The Appeals Court then suffered a set-back in developing its jurisprudence. The Supreme Court reviewed its decision in Benson28. The highest court barred the patenting of ‘mathematical algorithms’, of which Benson’s number converter was said to be an example. The definition of ‘mathematical algorithms’ was never that clear to commentators or to the Appeals Court29. They were conflated with computer programs, and this was taken by all concerned as a slur on the patentability of programs. Following Benson, applications for patents on computer programs slumped and remained low for a decade30. After the body blow of Benson, the Appeals Court’s new task was to ‘decide around’ that case, and rescue innovative software where this seemed justifiable. 1.3 Deciding Around Benson These were some of the strategies the Appeals Court used: 1.3.1 Category Myopia If software is installed on a hardware platform and claimed as a machine its chances of patentability are improved. The Benson decision only rejected computation processes, therefore the court was entitled to cordon off processes embodied by programs and claimed as such as the only unpatentable category31. But note the uneasiness displayed by Appeals Court Judge Rich, dissenting in Johnston from approving the machine claims: The point is that the machine and process claims are really directed to the same invention … 28 Gottschalk v Benson, 409 US at 273. [November 1972], reviewing Appeals Court decision (441 F 2 nd [1971], at 682). Concerning a number converter that changed a number represented in Binary Coded Decimal (BCD) into pure binary. The use was in a telephone switch station. The benefits of this converter that were puffed over and above solutions already in the art were these. It cut down the repetitive computerised stores and retrievals of intermediate data, cut down the overall number of steps taken, reduced the number of hardware components required, and diminished the chance of error in number manipulations. Burroughs Corp. subsequently patented a ‘Number Conversion Apparatus’: US 4,342,026 [1982]. Their specification states: ‘many systems for converting between BCD and binary have been proposed and are well known in the art.’ They cited patents for converters from 1957 (2,814,437) up to 1970 (3,505,675). 29 Note Judge Rich’s comments in Christensen (immediately post-Benson). 478 F 2nd at 1392 [1973], at 1396. 30 Keith Witek’s software patentability index. (Pers com). 31 By ‘judicial exception’. Application of Johnston. Before the Court of Appeals ([1974], 502 F 2 nd 765). Johnson’s invention was a program that processed bank account transactions. All 5 claims at issue were to a system (a machine). The Appeals Court over-turned the Patent Office’s rejection of Johnston’s claims. 7 He pointed out that the invention was, in fact: being sold as a computer program to banks so that they can perform data processing services. Once categories are specifically listed, they provide opportunity for misconstruction of terms when new technological objects confront them. The centralist approach of Judge Rich sought to expose the actuality of the invention32. 1.3.2 Carving Computational Levels System software was distinguished from application software in Chatfield33. So long as software enhances the performance of its platform rather than merely processing data it may be deemed patentable. This is a mirror of the European ‘technical effect’ argument. The Chatfield court also distinguished the ‘Benson’ algorithm from the ‘Mainstream’34 algorithm. The former had mathematical content, the latter was ‘just another form of process’35. I suggest that this was a seriously wrong turning in the jurisprudential development. In fact, an algorithm is not ‘just another process’. It has symbolic connotations that ought not to be misrepresented: An algorithm can be applied to any of a class of certain symbolic inputs, and in a finite time and number of steps can eventuate in a result in corresponding symbolic (Frederick Adams) 36. output 1.3.3 Holism: the Physical Bookends Argument If there is sufficient ‘real world’ (that is, physical) activity that uses the results of any calculations (called post solution activity), a software invention may be patentable37. This 32 Judge Rich subsequently down-played the importance of s101 categorisation in State Street; Judge Plager even more so in Excel. 33 Application of Chatfield. 545 F 2nd, 152 [1976]. Chatfield invented system software that allocated resources in a multi-user system. The innovation was that priorities were dynamically assigned while programs were running, as opposed to being statically predetermined. 34 My term. 35 In Chatfield Judge Markey said: ‘The broader definition of an algorithm is “a step-by-step procedure for solving a problem” … It is axiomatic that inventive minds seek … step-by-step solutions. (545 F 2nd [1976]), at 156. 36 ‘[An algorithm is] an effective or clerical procedure. It is a step-by-step recipe for computing the value of a function. It determines what is to be done at each step without requiring any ingenuity of any person, or any machine, executing it’. (Stewart Shapiro). 8 safe harbour remains in the MPEP Guidelines for computer-related inventions38, and practising attorneys recommend the use of manoeuvres that meet such demands39. It stands for the judicial view that the hybridised invention must be treated as a whole, and not dissected. Historically, the problem that arose from this mindset was as follows. The ‘diplomatic’ use of real world bookends tacked on to a basically mathematical solution made the Patent Office and courts very sensitive to the liberties that might be taken in the name of holism40. At the time, the Patent Office had, to an extent, replaced the point-of–novelty test by the Blue Pencil Rule41. This strategy allowed a claim to overcome the s101 hurdle as patentable subject matter so long as one statutory (ie physical) element existed in the claim. It subsequently normalised by ‘pencilling out’ the non-statutory elements for the purpose of the s102 (novelty) and s103 (non-obviousness) inquiries. If the mathematics is removed from an invention, what is lost may very well be the prime source of novelty. What is definitely lost is the interface between the mathematics and the physical world which may be a source of novelty in its own right. This move is dissectionist and anti-holistic. The most extreme (and aberrant) decision in this mindset was Flook42 in the Supreme Court. In that case not only was the mathematics extirpated from a hydrocarbon cracking process (which the program embodying the mathematics controlled) at the start of the analysis, but it was declared a ‘known’ mathematical relation, when there was no evidence to that effect43. The Patent Office was developing a test that essentially instructed thus: 1. Identify a mathematical algorithm in a claim. 2. Delete it from the claim. 3. Ask, does the remainder The term ‘post solution activity’ was coined in the Appeals Court hearing of Flook (559 F 2nd [1977], at 21). An invention that continually (the innovation) recalculated the alarm limit in a petroleum-cracking plant. 38 S1V B 2b, citing Diehr. 39 Keith Witek, Developing a Comprehensive Software Claim Drafting Strategy for US Software Patents. Berkeley J Tec & L, [1996]). Gregory Stobbs, [1996 and 2000 eds] Software Patents. 40 For example: In re Richman 195 USPQ 340 [1977]. 41 Carl Moy Statutory Subject Matter and Hybrid Claiming, John Marshall JCIL, 277. 42 Parker v Flook, 437 US, at 451. [Decided June 22nd 1978]. 43 According to Chisum that was an illicit manoeuvre that should have been purged as ‘antithetical to patent law’. The Patentability of Algorithms. University of Pittsburgh Law Review, [1986], 970, at 995. Carl Moy suggests that the Flook decision shows the Court’s intention to demarcate the domain of principles and scientific truths (where it considered Flook’s equation belonged), as immune from patent monopoly. 37 9 of the claim address statutory subject matter44? If so it passes the s101 categorisation test. This clearly reduces the threshold of s101 as compared to point-of–novelty. In Flook, the Court asked this question: is the invention minus the maths novel? That is, is oil refining novel? The answer was, not surprisingly, ‘no’. In the Appeals Court we see the beginnings of the Freeman, Walter, Abele test for the ‘process with a hole in it’: Any mathematics is removed, the statutory nature of the residue is tested, then the mathematics is allowed to limit the claim. (See Abele45 and also Diehr46 in the Supreme Court). During this time the focus was on s101 statutory subject matter issues. The s102 and 103 inquiries may well have been starved out47. 1.3.4 The Fate of Expert Systems under the Blue Pencil Rule Under Blue Pencil, expert diagnostic programs and business methods were bound to fail the patent demand for real-world, structural correlates to any calculations being done by software. The decision in Meyer illustrates this point48. Meyer’s invention49 was an expert system to complement a neurologist’s diagnostic skills. A battery of tests was performed on a patient and the results computed as a probability of normal function or malfunction of various parts of the patient’s nervous system. Using the ABC pattern of analysis we split the invention into three phases: A is data gathering (by physiological testing). B is a mathematical analysis of the test results – the core of the invention. C is a clinical action taken by a doctor informed by the analysis at B. A was held to be not convincingly ‘real 44 Gable and Leaheey: The Strength of Patent Protection of Computer Products, (17, Rutgers [1991] at 106), citing Barett: Patentable Subject Matter: Mathematical Algorithms and Computer Programs, (1106 Official Gazette PTO [1989]). 45 Abele’s invention allowed reduced exposure of a patient to X-irradiation in a computed tomography application. Artefacts thrown up by the limited exposure were removed by a mathematical weighting function. In the Appeals Court (684 F 2 nd, [1982]), Judge Nies conceptually removed the mathematical part and found the remains to be statutory. The invention was then considered as a whole, and the mathematics approved as limiting the claims. The Patent Office examiner, however, had operated under the Flook principle, looked for novelty in the tomography without the maths, and decided it was missing. 46 Diehr’s invention provided a dynamic control for traditional rubber curing. Previously, the total time allowed to heat rubber in moulds and effect the cure was statically set, and could not be adjusted during the cure. The cure time was calculated using the Arrhenius equation, which was the mathematics in the method, and known to be used in the field. (450 US [1981], 175). 47 Diehr’s invention, for example, is just an iterative version of rubber-curing control, and appears to be rather obvious. Also Judge Newman, dissenting, in Schrader found that ‘the record does not show analysis in terms of sections 102 and 103. It is not reasonable to believe that the activity described is devoid of prior art …’. 48 688 F 2nd 789 [1982]. 49 Process and Apparatus for Identifying Locations of Probable Malfunction. Filed [1974]. 10 world’50, but the mere foddering of variables with values. B is ‘mathematics’ and therefore unpatentable. C, of course, is beyond the ambit of any patent. 1.4 Recent History by Cases 1.4.1 Iwahashi51 Iwahashi invented ‘An Auto-Correlation Circuit for Use in Pattern Recognition’52. As the specification put it: state of the art units for calculation of the auto-correlation coefficients have the disadvantage of requiring expensive multipliers and also complex circuitry. Rather than use a multiplication step to get the square of the sum of two numbers, the Iwahashi unit provided a look-up table, implemented as a ROM. The Appeals Court allowed that Iwahashi’s invention was patentable because it claimed the invention as a specific machine. All the hardware was general except for the ROM, which the Court considered to be ‘specific’, and to make the whole specific thereby. Iwahashi’s claims were functional (that is, in ‘means-for’ format – each element of hardware being described as ‘a means for’ implementing a function). Under such format the examiner is coerced (by s112.6) to scrutinise the invention’s written description to find, for each function: corresponding structure … described in the specification and equivalents thereof53. The Court looked for ‘specific structure’54 and found the ROM. Iwahashi stands for one of two propositions, according to your viewpoint. Either that some piece of specific structure supporting a mathematical innovation ratifies that invention as a proper patentable machine, Or that a mathematical process supported by a general and non-innovative platform can be called statutory, thanks one piece of ‘nominal hardware’55 (a ROM in this case56). Data gathering was discredited as insufficiently ‘real world’ in Richman, (563 F 2 nd 1026 [1977]). 888 F 2nd, at 1370 [1989]. 52 Filed [1982]. 53 35 US Code s112.6. 54 The origins of this ‘specific structure’ requirement are questioned by Irah Donner, Throwing Out Baby Benson, (Jurimetrics Journal [Winter 1993] at 247). 50 51 11 1.4.2 Arrhythmia57 The issue in this case was whether claims to an output of signals that merely represent physical entities (in this case a patient’s heart condition), as opposed to being physical entities (beyond being ‘signals’58) are valid. Arrhythmia’s invention sought to pinpoint exactly which patients recovering from heart attack were at high risk of the potentially lethal tachycardia (fast heart-rate). If this risk were known, then ameliorative drugs, (with their strong side effects), could be used in a focused way and only where necessary. The invention used a mathematical implementation of a Butterworth filter to suppress noise on the heart signals. Judge Newman said this: The resultant output [of the filter] is not an abstract number, but a signal related to the patient’s heart activity. Arrhythmia stands for one of two propositions, according to your viewpoint. Either that signals that represent physical things of claimed origin59 (human heart rate etc) are legitimate physical entities for patent purposes60, Or that the bounds of patentability are being pushed back a little further. It is now sufficient for the mathematical treatment of data (the novel item in Arrhythmia) to be couched in terms of ‘signals’ that represent physical things rather being physical things. 1.4.3 Alappat61 It is reported that Francis Gray (internal counsel for Tektronix, the assignee of the Alappat invention) brought this case for the sole purpose of seeking a ruling on whether Richard Stern’s term. The use of a ROM by the way was utterly obvious. 57 Arrhythmia Research Technology Inc v Corazonix Corp. (District Court ND Tex, [1990]). Appeals Court (958 F 2nd 1053 [1992]). 58 The Patent Office always looked favourably on ‘signals’ as a token of tangibility for patentability. 59 Note also that those claims in Abele that failed to declare the origin of the ‘signals’ as being from Xirradiation of a human body failed to convince the court of their validity. 60 MPEP Guidelines. 1V B 2b. 61 In re Alappat. (33 F 3rd [1994]). 55 56 12 digital circuitry governed by mathematical formulae could be patented62. Alappat, an en banc decision63, in effect said that it could. Digital oscilloscopes suffer the flaw of displaying traces that rise or fall steeply as jagged and discontinuous, an effect called ‘aliasing’. Although anti-aliasing devices were known in the art before Alappat’s invention 64, Alappat made the following advance. If the pixels adjacent to the wave trace were subdued in intensity, (due to a novel mathematical treatment of the data), the eye/brain axis perceived the trace as ‘smoothed’. Although the data was derived from, and the benefit was felt in, an oscilloscope, the oscilloscope was not claimed. The specification did not disclose, nor was it claimed, where the input data – the vector list - was to come from or how it was to be generated.65 However, the content of the claim preamble linked into the elements of the claims sufficiently to satisfy Judge Rich that the data was adequately sourced. Here is the preamble to claim 15: A rasteriser for converting vector list data representing sample magnitudes of an input waveform into anti-aliased pixel illumination intensity data to be displayed on a display means comprising: …..66 Alappat stands for one of two propositions, according to your viewpoint. Either, provided sufficient structure (albeit non-specific) is supplied in the written description pointed to by a means-for claim, there is no reason to bar an invention with a ‘useful, concrete and tangible result’. Or, a method of mathematically processing data qua data (representing numbers of unclaimed physical origin) is patentable if claimed as a (structurally supported) ‘means machine’. 62 Reported by Gregory Aharohian (Internet Patent News Service). The full complement of judges, in this case eleven, was present. 64 Patent subsequently issued as patent number 5,440,676, [1995]. 65 Comment of the dissent in Alappat, led by Judge Archer, at 1563. 66 Note the link to this claim element: ‘means for outputting illumination intensity data as a pre-determined function of the normalised vertical distance and elevation.’ 63 13 The Alappat dissent saw the majority’s decision as a sacrifice of the s101 hurdle: the statutory subject matter test, since essentially mathematics had been shown patentable. We also see the inception of the ‘useful, concrete and tangible result’ triplet, reiterated in State Street and Excel. One reading of Alappat is that there is an attempt here to re-unite utility and innovation. 1.4.4 In Memory: Lowry and Beauregard A program carried on a disk defines a legitimate ‘manufacture’ for s101 purposes. Lowry’s invention67 concerned an object-oriented data structure. Claims 1 to 5 were directed to the ‘memory’ that stored the data structure. The court maintained that: The claims require specific electronic structural elements which impart a physical organisation on the information stored in memory. Lowry’s invention manages information. … Thus Lowry’s claims define functional characteristics of the memory68. More than mere abstraction, the data structures are specific electrical or magnetic structural elements in memory. While this is true as far as it goes, a software engineer would simply regard a data structure as part of a program in which the program declaration pledges that memory will be used in a specified manner in response to program instructions. Any notion of electrical or magnetic structure is beside the point. Beauregard’s invention69 was a program, claimed on a disk, that overcame the graphical complexity of filling in a complex polygon, such as might be found on a map of Norwegian fjords. Complicated algorithms for filling such shapes were already in the prior art, but these were slow to execute. The essential novelty of the program was that it initially inspected the shape to fill, classified it, and selected the fastest fill algorithm for a member of that class. 67 Data Processing System Having a Data Structure with a Single Simple Primitive. Assignee was DEC. In re Lowry. Appeals Court [1994]. (32 F 3rd at 1579). 68 Lowry at 1583. 69 System and Method for Utilising Fast Polygon Fill Routine in a Graphic Display. US patent 4,962,468, [1990]. Assignee: IBM. 14 Jeffrey Draeger70 argues that the Patent Office Guidelines and the courts part company in their views of what it is allowable to claim on a disk. The Examination Guidelines state that patentable subject matter of the ilk of programs and data structures must: Become structurally AND functionally interrelated to the carrier medium71. Lowry approved what was perceived to be the physical stamp of organisation that a data structure lays upon memory. It is an actual enablement that allows more efficient access to data. For Draeger a program is a different matter to a data structure. It is only stored in memory of necessity, prior to its animation; it leaves no structural imprint on memory. According to Draeger, the claim of the disk-carried program to patentability is as a functional component of the machine system as a whole72. Hence the Guidelines go beyond Lowry, the seminal decision, yet they do not carry the legal weight that the court decisions carry. The most the Guidelines ought to stipulate, claims Draeger, should be: structural OR functional interrelation [between program and disk]. Thus Lowry does not warrant Beauregard. In this mindset you cannot argue for the functional entity (the program) on the strength of the structural one (the data structure). Beyond these issues lies the problem that if a disk is claimed, there is no new utility in the fact that a program (or anything else, music, for instance73) is committed to disk, or the way the material is committed to disk, or the way the machine reads it. Neither is a disk a functional entity without the co-operation of a machine. Lee Hollaar makes a distinction between kinds of memory. He separates executable from repository memory74. He objects that a program on media only has the potential for function (for utility). It needs to be integrated with a hardware/software platform, provided with linking loaders and other software, in order to display its intended transformation of values ‘kinetically’ (my term). He says: The only memory where information is functional is executable memory. Are Beauregard’s Claims Really Valid, John Marshall JCIL [Fall 1998]. Guidelines Part 1V.B.1. See the grounding of the program-substrate interaction argument in the printed matter exception. 72 Id, at 361. This is the ‘cam for the machine’ argument offered by IBM in Beauregard. 73 The US Patent Office has been obliged by the carrier medium solution to distinguish, in the [1996] Guidelines, non-functional descriptive material (music, play readings etc) from functional descriptive material (computer programs). 70 71 15 A program on a disk, from his viewpoint, is just data, and might as well be music. Although current MPEP Examination Guidelines75 allow transitive (on-a-disk) function as actual function to cover this case, they also say that media-borne programs are only patentable if the underlying process is also patentable. A drawback to a memory claim is loss of scope, since media rapidly go out of style, and a claim needs to be as broad as possible. On the other hand, to claim ‘any and every medium’ risks getting default treatment as a process, whose claim to patentability is in any event less robust. Richard Stern76 has suggested that a disk claim should only be valid if the corresponding process is valid. Stern suggests the carrier disk claim must match a ‘hypothetical’ process or machine counterpart77 under this test. But if the process claim would have been patentable, then the chief benefit in the carrier disk claim strategy appears to be to extend the reach of a direct infringement action78. 1.4.5 State Street: Tangible Assets Commentators hold out State Street79 as the case that swept away the business method exception, yet a careful reading of the case indicates that it was the mathematical algorithm argument that was most significantly quenched80. Signature wrote software called the ‘Hub and Spoke’ financial management system. It was designed to implement economies of scale in administration, but primarily to make tax savings. License negotiations with State Street broke down and State Street tried to have Signature’s claims invalidated. 74 Justice Douglas was Right (AIPLA Quart J, [1996]). Part 1V.B.1. They speak of ‘functional descriptive material’ which ‘imparts functionality when encoded on computer-readable media’, and give no patentable weight to non-functional descriptive material (traditional literary texts and music). 76 An Attempt to Rationalise Floppy Disk Claims. John Marshall JCIL [Fall 1998]). 77 Stern supra at 198. 78 If software is claimed as a process then only the end-user may be sued for direct infringement. The copyist distributor is, at worst, a contributory or indirect infringer. The same applies to software installed on a machine, if execution of the software is an inherent requirement voiced in the claim. But if the copyist distributor has a product – a disk – in his hand, then he infringes directly, and liability is strict. 79 State Street Bank and Trust Company v Signature Financial Group Inc., (149 F3rd [1998] at 1368). Signature’s Patent: no 5,193,0536. Filed [1991]. Issued [1993]. 80 The business method move by Judge Rich was a reiteration of Judge Newman’s suggestion in Schrader that ‘a method of doing business’ was a ‘fuzzy concept’ which should be ‘discarded as error-prone, redundant and obsolete’. This followed her detailed analysis of so-called business method precedent cases. She demonstrates that their BM aspect was tangential to the main thrust of the decisions. Yet they were habitually cited in any business method context. (22 F3rd [1994], 290, at 298). 75 16 One important fact in the case was that Signature’s claims were to a machine, using means-for language. Judge Rich had only to use the technique he had already used in Alappat and Iwahashi of coupling each means-claimed function to a structural element present in the written description of the invention, to demonstrate sufficient structural support to validate the claim. In State Street the significance of the s101 patentable categories was minimised: The question of whether a claim encompasses statutory subject matter should not focus on which of the four categories of subject matter a claim is directed to ….81 The court’s conclusion was that: the transformation of data, representing discrete dollar amounts, by a machine through a series of mathematical calculations into a final share price, constitutes a practical application of a mathematical algorithm, formula or calculation, because it produces ‘a useful, concrete and tangible result’. State Street stands for one of two propositions, according to your viewpoint. Either that a useful, concrete and tangible result has evolved to override considerations of s101 type, judicial exceptions and all Or that the force of technological expansion in the software art has coerced the courts to abandon their Canutist stance; they can no longer sustain the recruitment of ancillary artefacts to produce a specious hybrid out of software. Tailpiece The Signature patent application failed in Europe82. The grounds were that the invention lacked technical character. Beresford expresses the view that this might have been mended by the submission of specific algorithms for solving the complex administrative tasks. These tasks would presumably comprise the technical problem83. 1.4.6 Excel84: Utility Dominates Subject Matter The word ‘useful’ occurs fourteen times in the Excel Opinion as the primary justification for the patentability of an invention with mathematical content. For instance: 81 State Street at 1375. European application number: 92908680.9-2201. 83 E-commerce and Business Method Patents in Europe (IBC Conference [October 2000]). 84 AT&T Corp. v Excel Communications, Inc. (172 F 3rd [1999] 1352). 82 17 Even though a mathematical algorithm is not patentable in isolation, a process that applies an equation to a new and useful end is at the very least not barred at the threshold by s101’. AT&T were the holders of US Patent 5,333,184, for an invention concerned with billing long distance calls. A customer can choose which interexchange carrier (IXC) they will principally use (their ‘PIC’). Records are passed to a message accumulation system for processing and billing. AT&T’s invention added a field to the record expressing a correlation between the recipient’s and the caller’s IXCs, to enable differential billing. The essence of the claimed method was to set values in the PIC record, apply a function and modify the bill figure. Judge Plager emphasised the ‘mainstream algorithm’ approach85: the judicially-defined proscription against patenting a ‘mathematical algorithm’, to the extent such a proscription still exists, is narrowly limited to mathematical algorithms in the abstract. The judge justified State Street as following Diehr. But Diehr had kept a sharp eye on the host process of rubber curing, and the holistic nature of the claims. State Street blatantly used the tangible hardware platform as just that; the focus of the invention was the calculating software. Excel approved State Street as ending the bar on numberoutputting inventions if there was a ‘useful concrete and tangible result’. Alappat had said that the invention claims should be assessed as a whole. Alappat had claimed a ‘means machine’, though the focus was the mathematics. The Excel court confronted claims to a method, unsupported by tangible artefacts. Hence Judge Plager strongly opined that the form of the claim was immaterial86. According to the court, the AT&T PIC indicator ‘represents information about PICs’. Its utility is in the achievement of a tuned billing method. Said the court: it neither claims nor pre-empts the Boolean functions it uses: AT&T’s claimed process employs … PICs as data, applies Boolean algebra to those data to determine the value of the PIC indicator, and applies that value through switching and recording mechanisms to create a signal useful for billing purposes. 85 Background Section B. 18 The ‘mechanisms’ and ‘signals’ presumably fulfil the tangibility requirement. Excel had argued that there was no physical transformation in AT&T’s invention. The Federal Circuit answered that point, citing Arrhythmia: The steps [in Arrhythmia] transformed physical, electrical signals from one form into another form – a number representing a signal related to the patient’s heart activity, a non-abstract output. … Arrhythmia’s method claims satisfied s101 because the algorithm … [produced] a number which had a specific meaning87. The Appeals Court reversed the District Court, remanding the claims for consideration under sections 102, 103, and 112. Tailpiece Although comprising patentable subject matter, the AT&T billing method was subsequently found obvious in light of MCI’s ‘Friends and Family’ program. Conclusion This review of case law shows some of the problems created by the introduction of artefacts and linguistic euphemisms designed to implicate tangibility in claims for virtual inventions. What is at issue here is to show a proof of specific utility in a virtual device. In this paper it is taken as a given that such utility will redound in the real world as a human benefit. Where utility finds a focus in incidental artefacts, it loses its focus in the very novelty being claimed, yet we know that the condition of patentability is conjunct: novelty AND utility. An holistic view of the invention, expressed in the claims, and assessed in the Patent Office and Courts is often a valuable and appropriate strategy. However, there comes a point when that approach runs out of road; certainly the merger of novel virtual solutions with artefacts becomes distorting. That point seems to be where so-called virtual inventions service a virtual world used for great benefit by humans. The next section explores what is understood by utility. Also citing State Street (at 1372): ‘for the purposes of s101 analysis, it is of little relevance whether claim 1 is directed to a “machine” or a “process”…’ 87 But note the words of Newman in Schrader: All mathematical algorithms transform data, and thus serve as a process to convert initial conditions or inputs into solutions or outputs…Data representing bid prices … do not differ … from data representing electrocardiogram signals… or parameters in a process for rubber curing, (at 297). 86 19 Part 2 2.1 The Nature of Utility Attempting to Define Utility Here are two dictionary definitions of utility: 1. Ability, capacity or power of action, to satisfy needs or gratify desires of the majority, or of the human race as a whole. (OED). 2. Fitness for some purpose; profitability for some desired end. Serviceableness. (Websters). The general rule for laws of nature (embracing mathematics), as expressed in Mackay Radio v Telegraph Co88 says this: while a scientific truth, or mathematical expression of it, is not a patentable invention, a novel and useful structure created with [its] aid may be. Many would hold that the utility requirement of patent law demands corporeality in the invention. In the rule that follows, the implication is that corporeality is opposed to abstraction, and allied to use: an idea of itself is not patentable.89 The corporeality of a computer program is frugally satisfied by the presence of signals on the wire. Beyond that, a particular use can be specified. In the European case of Vicom, the Board of Appeal complained that: ‘a method of digitally filtering data’ remains an abstract so long as it is not specified what physical entity is represented by the data and forms the subject of a technical process90. Yet the ‘physical entity’ referred to the Vicom turns out to be a satellite-transmitted picture which can be sharpened by using mathematical procedures. The grant of a patent is rightly viewed as a contract between grantee and government91 & 92 . There must exist, at the time of claiming an invention, some expressed utility for the public: 88 306 US [1939], 86, at 94. Rubber Tip Pencil v Howard. 87 US (20 Wall.), [1874], 498, at 507. 90 Decision T208/84 [1986] (2 EPOR [1987], 74 at 79). The claim was amended to ‘image data’ and was passed. 91 Genesis of Copyright and Patent Law Bugbee, at 10. 89 20 ‘a patent is not a hunting licence’93. In Brenner94, the Supreme Court cited Justice Story, reading him as describing the entry level for utility as: not doing damage. The Brenner court went on to demand ‘specific utility’ otherwise there was just a ‘monopoly of knowledge’, which was impermissible: The basic quid pro quo contemplated by the Constitution and Congress for granting a monopoly is the benefit derived by the public from an invention with substantial utility. Chisum’s Treatise spells out these elements for the utility of an invention: First it must be operable and capable of use to perform the functions and secure the result intended.95 Second it must operate to achieve some minimum human purpose. Third, it must achieve a human purpose that is not illegal, immoral or contrary to public policy96. There seems to be nothing in these three elements of the core definition to challenge the specific utility found in certain mathematical relations. 2.2 Utility and the Statute Two sections of the US Code address the utility requirement: S101: ‘new and useful’. S112: The specification shall contain a written description of the invention, and of the manner and process of making and using it, in such full, clear, concise and exact terms as to enable any 92 A monopoly, on the other hand, has the connotation of taking out of public usewhat they once had automatic right to. (Comment of Chief Justice Taney in Charles Rivers v Warren Bridge [1837]. Coke (1628) traces this usage). Patent and monopoly are currently used interchangeably. 93 Fortas J in Brenner v Manson, (Supreme Court. 383, US 519 [1966]). 94 Brenner v Manson, 383 US 519 [1966]. 95 Having designated functionality. An example where this was not the case was a purported ‘perpetual motion machine’ which, under trials by the National Bureau of Standards failed in its goal of running perpetually. (Newman v Quigg, 877 F 2nd 1575 [1989]). 96 As Justice Story put it, a patent must not achieve utility to ‘poison people, or promote debauchery, or facilitate private assassination’. Example case decisions include: a one-armed bandit machine denied a patent (Ex parte Murphy, 200 USPQ 801 [1977]), a toy racecourse denied a patent [1889], (encouraged betting), and a pinball machine allowed a patent, [1941] (did not necessarily promote gambling). 21 person skilled in the art to which it pertains … to make and use the same … S101’s use of the term ‘useful’ is considered to be mainly historic and general. S112, on the other hand, demands sufficient description for the skilled artisan to make and use the invention. The relevant rule97 states that a specific embodiment of the precise invention claimed must be described. It is the invention claimed that must have the utility. The updated MPEP Guidelines [2002] cryptically say this (2106): Usefulness ‘requires significant functionality’. ‘the mere fact that the claim may satisfy the utility requirement of s101 does not mean that a useful result is achieved under the practical application requirement as expressed by the useful, concrete and tangible (UTC) result. The Guidelines go on to admit that ‘the courts have yet to define the terms useful, concrete and tangible’. 2.3 The Notion of Virtual Utility Rochelle Dreyfuss has commented that: Both patent law and copyright distinguish between principles and instantiations of those principles. An instantiation can be privately owned, but an abstraction must go into the public domain. This begs the following questions, which I deal with in detail elsewhere98: 1. What if the instantiation of a theoretical/explanatory law or principle (abstract) is in fact a local/phenomenological law (also ‘abstract’), with a highly specific application in a patentable invention? This has the appearance of an abstraction with utility. 2. Is specific mathematics ‘technical’ or non-technical? 3. What is ‘technical’ anyhow? 2.4 97 Machines99: a Class of Useful Things The Code of Federal Regulations: CFR s1.71. 22 In 1967 Minsky said this: ‘Machine’ cannot usefully be defined as ‘a member of [a certain class of physical objects]’. For the decision as to whether something is a machine depends on what that thing is actually used for, and not just its composition or structure. I suggest that a more flexible approach to the term ‘machine’ would have assisted the transition between conceptualising the machine as heavy, industrial plant and conceiving of it as the software virtual machine. Computer Science is grounded in ideas of the ‘levelled machine’100. Each level has its own concerns, each takes input, processes it and produces output for another level to consume and process. A software machine is able to function on variable thickness platforms: (that is hardware or a hardware/software mix). 2.5 Software as a Process If software is viewed as a process it is seen as a platform-independent transformer of input into output. Cochrane v Deener101 is considered the lead case that made the oneto-many relation between process and platform explicit, and liberated the process from its platform. The legacy of that case in software decisions, unfortunately, has been that courts took a valid process to be always one that treats of a physical input. It was therefore argued by the Patent Office in the early software cases that the input to software, being data, was insufficiently 'physical'. The language used in the Cochrane reasoning has no connotations of physicality, and the notion that it meant physicality has been challenged (see the early software cases). Moy102, however, maintains that the physicality constraint has tended to act as ‘a bulwark’ against the expansion of the patent system beyond its traditional limits, Recent software decisions effectively refute 98 J.A. Stanley, Thesis: Invention and Representation. Also see J.A. Stanley, Are the Laws of Nature Patentable After All? (SLS Conference, September 2003). 99 The Turing machine has been described as ‘the most famous abstract model of a computer’ (Representation and Reality, Introduction, Hilary Putnam, [1988], MIT). 100 Two examples are (1) the ISO OSI 7-layered model for networking, and (2) Tanenbaum’s own levelled model for conceptualising software in levels from user to hardware. Structured Computer Organisation, Andrew Tanenbaum. 101 94 US 780 [1876]. 102 Revision to Deller on Patents. 23 the physicality approach, reasoning that data represent actual things, which seems only to be a convenient truism. 2.6 Claim Structure in the Service of Virtual Utility: The Substitution of Physical Constraints by Virtual Ones If software is a representative, then surely all its elements - the properties of its data substrate and the its output - ought to be accepted as virtual. Alappat was controversial for this very reason. The Federal Circuit dissent in that case complained that the provenance of the data (the oscilloscope) was unspecified in the claims, as was its target use103. Similarly in State Street the data was simply numerical, only constrained so far as it represented ‘dollar amounts’104. I suggest that the physical book-ends of a claim to an algorithm could be removed (though this is not prescriptive) if these claiming devices were used: 1. Data type, range and the specific source of input data, also the target use of the output of an algorithm should all be delineated in an enhanced preamble to the claim, which is a definite part of the claim. 2. In support of point 1, the intra-claim nature of the preamble can be confirmed, as in Alappat, by strategies such as reference between the preamble and the elements of the claim body. 2.7 Field of Use At the moment, a claim’s preamble is used principally to express the ‘field of use’. Its possible range of roles seems at the moment uncertain105. There are, however, two recent straws in the wind that indicate its possible range of scope. 103 Although its data derived from an oscilloscope, and its output was used by one, the oscilloscope itself was not claimed. 104 Which led the District Court to reject State Street’s action, and declare the claims invalid through insufficient mapping to structural correlates under the prevalent Freeman Walter Abele test, subsequently thrown out by the Federal Circuit in its hearing of State Street. 105 Chisum comments on preamble (at s8.06[1][d]): ‘if the preamble is a limitation, it defines the scope of the claimed subject matter. It can be relied on to distinguish the claim from the prior art. Rosenberg says that the preamble can be made directed to the inventor’s own work (as opposed to the prior art context – which is the usual thing) by an explanation of the fact (Reading v Baker 748 F 2nd 645 [1984].) Otherwise, preamble is not considered part of the claim. He concludes that there is ‘no litmus test either way’. (Citing Judge Nies in Corning Glass v Sumitomo [1989] CAFC. See also Patent Law in Biotechnology, Chemical sand Pharmaceuticals, Harold Wegner (2nd ed.) 24 In the case of Rowe v Dror, a general purpose ‘balloon catheter’, already in the art, was compared to an invention which was a ‘balloon angioplasty catheter’, and referenced in the preamble to a claim106. In construing the application, the Federal Circuit picked up the fact that the claim, limited by the pre-amble referred to equipment that had to cope with the extra functionality of expanding an arterial stenosis, which a standard balloon catheter does not. The preamble provided the required functional specificity and therefore constituted a proper limitation on the claim107. Secondly Alappat’s claim 15 preamble gave the field of use of the rasteriser: A rasteriser for converting vector list data representing sample magnitudes of an input waveform into anti-aliased pixel illumination intensity data to be displayed on a display means comprising ... Section 3: How did Cryptography Get Away with it For So Long? 3.0 Introduction The art of cryptography allows the development of methods to transmit messages securely on an insecure channel. The best of the encryption algorithms run very slowly. A variety of hardware and software implementations are usually available for any of them. Ciphers are coded representations of representatives. The representatives are words and numbers in binary format. Thus the real world is very far removed from these objects. Yet cryptography has been a well-established patentable class for at least 25 years. Under the US Patent Office Classification108 it is ‘Class 380, CRYPTOGRAPHY’. Subclass 28 concerns algorithms encoding ‘signals’. Subclass 30 (which is indented within subclass 28) concerns public key encryption specifically. I will briefly outline the content and benefit of three cryptographic algorithms, considering the utility of each. They are, the RSA public key patent109, Schlafly’s patent110, and Hoffstein’s patent.111 & 112 106 Rowe v Dror, (112 F 3rd 473 [1997]. The claim was therefore patentably distinguishable from the prior art (ie unobvious) on the basis of what the preamble said. 108 Refer to the Classification Definitions [Dec 2000], US PTO web site. 109 Rivest, Shamir and Adleman: US Pat# 4,405,829 (Class 380/30). Filed [December 1977], issued [September 1983]. Assignee MIT. 107 25 3.1 The RSA Public Key Patent The most secure method of encryption rests on the notion that the key used to decrypt a message is not a direct inverse of the key used to encrypt it, (the asymmetric solution). For this purpose a one-way function is necessary. The unpredictable behaviour of prime numbers under functions such as exponentiation and modular reduction make them peculiarly fitted to these methods. A user creates a public key (two numbers N and e, a modulus and an exponent respectively). These the user releases to the public who might wish to correspond with her. Also in the public domain is the one-way function that will be needed to encrypt the message. The N number, however, is a composite; it is made by multiplying two large prime numbers (p and q) together. The user does not release these two numbers, but can generate her deciphering key from them. Once she has done this, p and q can be discarded. The deciphering key (d) is a function of the same broad sort as the enciphering key, involving an exponentiation and a modular reduction. The success of the method relies on the difficulty of factoring N to try to find p and q and hence d. This is notoriously difficult to do, and any attack on privacy relies on brute force trial and error in detecting p and q. Such an attack is so impotent because of the computation time needed to run the attacking algorithms (10s or 100s of years). Therefore the encryption method can be called ‘computationally infeasible to break’. This is the benefit of the RSA patent. There is no guarantee that one day an alternative attack (not using factoring) will not succeed against RSA. Whitfield Diffie held out the principle of just such an asymmetric solution to the encoding, decoding problem113, he could not, however, implement such a solution. Rivest, Shamir and Adleman114 were able to implement it in their RSA Public Key patent, issued in 1983. Twenty-six of the forty RSA claims are to a cryptographic communications system; in other words, a machine. There are references to ‘channels’, ‘lines’ and ‘registers’, and other general physical elements. 110 Roger Schlafly: US Pat# 5,373,560. (Class 380/28). Filed [August 1993], issued [December 1994]. Jeffrey Hoffstein et al: US Pat# 6,081,597 (Class 380/28). Filed [August 1997], issued [June 2000]. Assignee NTRU Cryptosystems. 112 For technical details, see the works of Bruce Schneier and William Stallings. See also Monopoly, Mathematics and Nomological Machines, J. Stanley, forthcoming submission to Jurimetrics Journal, [2004]. 113 New Directions in Cryptography, Diffie et al, IEEE Trans Info Theory, Vol IT-22, num 6 [Nov,1976]. Stanford group. 111 26 How was it that the piece of specific mathematics that RSA embodies was not regarded as a ‘mathematical algorithm’ and rejected in the Patent Office? The two relevant Supreme Court decisions of the time of RSA’s filing and search were Benson and Flook. Both were deeply unpromising one would have thought. Yet the patent issued. This is what Flinn and Jordan wrote in 1997, when RSA had run its patent period: Meeting the FWA test involved reciting the cryptographic algorithm as an element within a physical device. In the RSA patent, this approach is stretched near if not beyond the breaking point. The only elements of the physical device which are not descriptions of the algorithm are ‘communications channels’, ‘an encoding means’ and ‘a decoding means’115. The authors comment that this strategy comes ‘as close as possible to claiming the algorithm by claiming the use of the algorithm in essentially all possible machines or with all possible processes’. Had the mathematics of the algorithm been examined as such, perhaps some notion of where mathematics is too broad to claim could have been developed in the Office. 3.2 Roger Schlafly’s Patent Since encryption relies on modular reduction, any method to make this more efficient is welcome in the art. Roger Schafly’s invention was a method to reduce the number of data-manipulations in a public key system without jeopardising the veracity of the output. The method depended, amongst other things on the specific properties of the numbers used as moduli. Schafly’s method is not more mathematical than RSA, but Schafly’s patent has attracted criticism for claiming two prime numbers albeit in a specific utility context. Individual primes vary in their specific properties, some lend themselves to such a method as Schlafly claimed, others do not. Two of Schlafly’s claims indicate setting an input port to specific primes in order to achieve the patent’s goal. The question again is: ‘if the claims are to specific mathematics – even numbers themselves - for particular use are they still too broad to succeed’? 3.3 Hoffstein’s Patents116 114 MIT group. CyberLaw [July 9th 1997]. 116 In addition to the patent mentioned above, also consider US Patent 6,298,137. Also see Hoffstein’s NTRUE web site. 115 27 Jeffrey Hoffstein elucidated mathematics to generate easily created, shorter encryption keys, and provided an algorithm that allowed a speed of encryption and decryption up to two orders of magnitude faster than the prior art. Was the method too general for patentability, as certain contributors to the Internet Patent News have suggested? Conclusions 1. The diffusion of the focus of invention in software to include physical artefacts is a solution that baffles the central question of whether a mathematical claim is excessively broad and ought not to be allowed, or whether it is so specific as to be allowable within its utility ambit. 2. Utility ought to be located at the very focus of invention, even if that is mathematical. 3. Specific utility voiced within a claim, and made explicit in the claim preamble, rather than reliant on added artefacts, is the best protection of patent quality. 4. Nowhere is the possible development of the role of the claim preamble discussed as a methodology that might help to solve the problem discussed here. 28