|Analysis||Compensate?||Sui Generis||Inflation||Tax Evasion||Sue?!||catlu|
Multinational companies in the fields of telecommunications (e.g. Alcatel, Siemens), computer electronics (ex. Thomson Multimedia), Aerospace & Defense (ex. Dassault, Matra, Thales), software (e.g. IBM) have been continuously lobbying European governments in order to get software patents fully legalised in Europe. The European Patent Office, most national patent offices as well as industrial property professionals are equally lobbying European governments.
Plans to legalise software patents in Europe have met opposition from various groups. 300 mostly small software companies from the Eurolinux Alliance have raised more than 90.000 signatures through an electronic petition. Most political parties in Europe have taken public positions against software patents. In some countries, parliaments have blocked governments from taking pro-patent positions within international bodies.
Multiple government-sponsored studies have been conducted in Europe to assess the economic and legal impact of software patents. Reports written from an economic perspective tend to conclude that the introduction of patents in the software economy would either generate no noticeable impact or lead to less innovation and competition. Reports written from a patent law perspective tend to conclude that there should be no explicit limitation of patentability and that the European practice should come in line with that of the United States of America. Some reports (IPI and the 2 German studies) combine both conclusions: software patents probably have a negative economic impact but should be legalised. Three reports (DG-MKT, IPR Helpdesk, Fraunhofer) include a review of opinion among software companies which shows that an overwhelming majority (80% to 90%) of software companies is happy with copyright and more or less hostile to software patents, with the noticeable exception of large software corporations (ex. IBM, Microsoft). Only one report explores alternative scenarios to the patentability of software: full exclusion, reduced enforcability, sui generis rights.
Among recent emerging solutions to the ongoing debate, some groups such as the French Academiy of Technologies have proposed to quickly and fully legalise the patenting of "computer-implementable inventions" (logical functionalities) and at the same time adopt compensatory regulation in order to reduce the potential abuses of such patents. The purpose of this article is to assess the financial costs of this approach and compare it with other approaches.
These two common ideas are quite wrong. Software patents are granted to people who do not necessarily develop and publish software. Also, software patents do not protect software authors against imitation.
For example, what the reader is currently reading is a textual composition of concepts, such as argumentation chains and rhetorical forms. Much effort went both into designing these concepts and combining them into a structured work. Copyright protects the orginal combination of concepts which defines this work: we hereby grant the reader a license to produce integral facsimile copies of this article but we forbid plagiarism or reusing parts of this article without permission. Copyright does not protect however the innovative or less-innovative concepts which this article in based on: readers are free to write original articles based on the same argumentation chains or rhetorical forms. The main argument for not granting monopolies on concepts or ideas in our societies is: promoting creation. Copyright would actually become quite useless if authors had to ask permission to hundreds of concept owners each time they wanted to create and publish an original work. Protecting ideas and concepts would just act as a barrier to creation.
Just like this article, a computer program is also a textual composition of concepts. Rather than argumentation or rhetorics, computer programs are based on logical functionalities. Software copyright protects the orginal combination of logical functionalities but not the logical functionalities per se. Advocates of software patents suggest that designing a functionality is the important part of a computer program, the rest consisting mainly of easy "coding". Opponents to software patents on the contrary argue that logical functionalities tend to be fairly easy to design, whereas the bulk of the skill and sweat of a programmer goes into weaving these logical functionalities together into a well-formed tissue. In either case, subjecting software to both patents and copyright means allowing the lesser achievement to stand in the way of the greater one, thereby risking to stifle innovation rather than promote it.
Claims are the most important part of a patent since they define how the patent may be enforced in case of dispute. For example, new applications of an existing patented molecule may be claimed in a new patent by another inventor. The owner of the initial patent may produce the patented molecule but will need a license in order to be allowed to use it for his newly patented application.
One should note that patentable inventions in chemistry, mechanics etc. may contain in their description some steps which are achieved under program control. Such inventions are not considered here to refer to software innovations, as long as the invention-relevant problem solution involves physical causality and not merely logical functionality.
While the description and the main claim often disclose an abstract rule of organisation and calculation, the dependent claims may describe a wide range of concrete objects or processes. Usually the description will use terminology of microelectronics (ex. "associative memory") or "software engineering" (ex. "database") rather than their mathematical equivalents (e.g. "indexed set") or their business administration equivalents (e.g. "directory book"). In practise this does not restrict the application of the patent claims, because software is the standard way of putting program and business logic to work, and the metaphorical language of "software engineering" is an equivalent of mathematical language. Where this still is not enough to cover a competitor's similar solution, patent courts may further enlarge the claim scope by applying the "doctrine of equivalence".
Thus, software patents on the description side teach us nothing but intellectual methods (logical functionalities, rules of organisation and calculation), while on the claim side they may succeed in monopolising a broad range of material objects and processes.
The EPO has never had a definition of what is "technical". The question was treated as a matter of examiners' intuition. "It is hard to define what is a camel, but when you see it you know it" was one common excuse for this lack of rigor. Yet there seemed to be a reliable consensus of customary law on what is technical. German courts were not satisfied with intution and created a definition, which has been in use for 30 years now:
Meanwhile, the EPO has however adopted a different approach to the EPC. In 1985, the EPO deleted the above-cited definition of "computer program" from the Examination Guidelines and replaced it with ambiguous wordings. The intuition on what is technical also faded away. In 1998, Mark Schar, a leading EPO judge, redefined the "technical invention" as "any practical and repeatable solution", explicitely rejecting the German definition which required a direct use of physical forces. Thereby he correctly described the EPO's de facto practise of patenting business methods and mathematical methods of even the abstractest kind. Since the early 90s, patents for such methods have regularly been granted, as long as they were dressed up as "system and process for ..." and no explicit mention of "computer program" was made in the claims. Starting in 1997, even claims to "computer program products" were granted, and "computer programs" were no longer considered to be "computer programs as such" if they could be said to have a "technical effect" such as running on a computer, or a "further technical effect", such as helping to improve computing efficiency.
However there is also a tendency in the EPO to refuse at least some software patents on grounds of not having a "technical effect". This was the case with a "pension benefit system" in 2000. In such cases, the EPO first conducts a prior art search. Then, in the context of assessing "inventive step" (non-obviousness), it demands that there be a "technical solution to a technical problem", based on EPO intuitions. In a recent press release, the EPO has furthermore signalled that it is overloaded with search requests for business methods and therefore will not conduct novelty searches if it is evident from the beginning that the contribution is not "technical". This again seems to indicate a certain willingness to return to "differential examination", i.e. the common sense of looking for a technical teaching before examining the novelty of that teaching, and to return to an understanding of "technical character" which no longer includes all "practical repeatable solutions" but on an intuitively perceived "technical character", which is somehow vaguely related to the mind-matter boundary and to Art 52 EPC. Therefore, at present, words should be carefully chosen whenever filing software patents at the European Patent Office. For example, it is better to use the word "database" than equivalents like "indexed set" or "directory book".
One should note that software patents granted by the European Patent Office have currently little value. Disputes show that high level national courts have largely enforced, even recently, the original interpretation of the EPC which excludes software patents. Some national judges, patent examinors, academics and patent attorneys, all known for their deep involvement in the patent system, have publicly expressed their worries that the current practice of the EPO could be illegal. Owners of European software patents tend not to use and enforce them because the risk of having the patent voided by a national court is much too high. This situation may sound absurd but it has a great advantage since it creates some kind of self-regulation which prevents abuses from software patent owners.
The only limitation currently lies in the requirement that the business method must be automatable, i.e. "repeatable" according to Mark Schar's invention concept. Thus, while the EPO granted a patent on the principle of learning languages by comparing student pronounciation with a digitally stored master sample, this covers only program logic which automates this principle, not application of the principle by human interaction between teachers and pupils.
Software patents tend to block interoperability because they allow to create monopolies on file formats, network protocols or interfaces. For example, Apple Computer was granted an exclusive license on the Sorenson patents on digital video compression. Apple markets a product called Quicktime which allows to view digital video compressed according to the Sorenson method on Windows and MacOS only. Other competing products (ex. Real, Windows Media Player, KDE Multimedia, XMMS, etc.) can not view digital video compressed this way because of the exclusive licensing of the Sorenson patents to Apple.
It is therefore desirable that software patent regulation measures enforce fair licensing practices on all patents required for interoperability.
As a reminder, most packaged software is developed by small teams of developers, typically 1 to 5 engineers. Training existing engineers to industrial property is long and reduces their availability for innovation tasks. Hiring a full time industrial property expert is too costly. Sharing a full time industrial property expert is difficult. Fully outsourcing industrial property is often inefficient thus too expensive. Industrial property is often experienced as a useless burden rather than ahnything else by small software publishers as the Fraunhofer study showed.
The risk for a small software publisher is tremendous. All currently published software likely include software patent infringements without their author being aware of it. Simple universal methods, such as publishing a database through a Web server, are covered by more than 10 patents, some of which are very similar. Patent attorneys are often unable to assess the risk of litigation on those patents. Although probably small, this risk relates to disputes costing millions of Euros. Small software publishers suffer much more than big corporations from this risk, not only because of their smaller financial size, but also because they do not usually own large patent portfolios which large corporations use to eliminate litigations risks with their competitors.
It is therefore desirable that software patent regulation measures allow to lower the access costs to industrial property and eliminate litigation risks for small software publishers.
Compensatory regulation should provide cheap or near-free insurance services to small and independent service shops in order to let them compete with large service shops.
Software patents are not a threat to free / opensource software developped by companies such as IBM, HP, etc., which constitutes the minority of opensource software. Software patents are however a threat to the vast majority of free / opensource software which is developped by individuals. A substantial amount of free / opensource software has already been taken off the net because of patent threats. Even in Europe, large corporations have threatened individual programmers. E.g. Thomson Multimedia threatened a Norwegian who published original implementations of patented algorithms (MP3). This case is quite typical: Thomson Multimedia had probably no legal grounds to start a dispute and win because the Norwegian programmer had written nothing but a computer program as such, which cannot be a patentable invention according to European law. However, a legal dispute with Thomson Multimedia is very costly and frightening. This is why Thomson Multimedia succeeded in forcing an individual programmer to remove a program from the Internet. This is also why European Linux distributions such as SuSE and Mandrake come without MP3 encoders and with ugly font rendering mechanisms for TrueType fonts.
Threatening opensource / free software developers with software patents is quite common in the United States. Sourceforge, a web site which hosts most opensource / free software projects, receives every week dozens of letters threatening to start a dispute for patent infringement. Up to now, Sourceforge has used the first amendment act in the United States, which protects freedom of speech, to reject such claims.
However, it is not certain that there is anything such as the first amendment act in Europe to protect the publication of free / opensource software. Exceptions to infringement, namely private use or research, may not apply to the publication of free/ opensource software because free / opensource software often constitutes a serious commercial competitor. Companies such as Microsoft consider software patents as a strategic tool to fight against opensource software.
It is therefore desirable that software patent regulation measures at least make sure that opensource software can be freely distributed and eliminate legal risks for individual developers.
Rather than financing innovation, software patents seem to finance a litigation economy based on the filing of more or less trivial (and often even old) software ideas at patent offices. Small software publishers, who are essential for competition and innovation in the software economy, are the most vulnerable targets. In order to garantee that software patents do not excessively impede innovation and competition, compensatory regulation would need to address this problem.
The French Academy of Technology did not provide any detail on what those measures really mean. In order to assess their cost, we shall introduce hereafter original details for those 4 measures and add a fifth one: a patent insurance, which could be considered as an incentive for SMEs in combination with a garantee fund.
If patents are bought at 200,000 EUR from national research centers and universities, 50,000 EUR of which are given to researchers as an incentive, if 20 persons are hired to select patents which deserve to be bought and 20 other persons are in charge of patent busting, a Garantee Fund costs about 200,000,000 EUR a year. Such a garantee fund would also optimise the effect of patents on innovation since it allows to combine property approaches with recompense approaches.
One should note that 1,000 patents a year is not that much. According to the USPTO database, a company such as IBM owns more than 30,000 patents and files more than 3,000 patents every year in the United States. Those figures are probably underestimated if one considers patents owned by IBM through a subsidiary or bought by IBM from third parties. It is very likely that IBM owns as much a 100,000 patents. Therefore, 1,000 patents a year is required to cover the basic technologies of software within a few years. 5,000 patents a year would probably be required, especially if the overall number and scope of software patents increases, which is to be expected.
In the figures above, we have set up a patent insurance together with the garantee fund and state financed the management costs of that fund. This is a reasonable solution considering that private insurances do not want at present to insure software patent litigations. We have also included subsidies for SMEs in order to generate a software patent insurance market in the mid term.
Such costs do not include the 30% increase in software development generated by the requirement for an industrial property strategy for software publishers, nor do they include the future cost of patent insurance for service companies. Those costs are the same for everyone and are paid by the end consumer.
|group||item||unit||unit cost||min units||max units||min total||max total|
The interoperability exception should extend to software patents the interoperability principle defined in the European software copyright Law. The interoperability exception should garantee some kind of automatic free licensing for interoperability purposes. It could be decreed for example that a free license is automatically granted for importing / viewing data from another programme, that a free license is automatically granted for exporting data to another programme unless fully interoperable open formats exist and that a free license is automatically granted to communicate to another programme through a network protocol unless fully interoperable open formats exists to communicate to that other programme. This way, owners of patents on formats and protocols can either enforce their property by guaranteing that all implementations include fully open and interoperable import fil or network protocols or let competitors use their property for interoperability.
The free / open source software exception could explicitely stipulate that the open source / free software process is a kind of R&D or even a parallel system for the promotion of innovation and diffusion and can therefore not be considered as a patent infringement. Various degrees of exemption are possible.
Such explicit exemptions produce near equivalent effects to what the guarantee fund provides.
Thus the publication of software could no longer be viewed as a direct infringement. Even if it could still constitute a contributory infringement, this reduction in liablity could significantly contribute to protecting small software publishers from abusive litigations because contributory infringement provisions are much more flexible than direct infringement provisions.
Such an explicit limitation could state for example that software distributed for free on the Internet can never be considered as a contributory patent infringement as long as all patents required to use it commercialy are clearly listed. Such an approach is required in our opinion in order to make the international nature of software distribution on the Internet compatible with the national nature of patent law. On the other hand, matters could be different for software distributed on physical media or to software sold on the Internet for which some kind of national nature exists (ex. credit card number). Physical devices (ex. PABX, missile) and industrial applications of algorithms (e.g. steel cutting) could receive yet another kind of treatment.
Such explicit limitations produce near equivalent effects to what an algorithm database and better examination provide against legal terrorism. Owners of valid software patents are encouraged by this system to reach an agreement with software publishers.
An even more friendly variant would be to restrict enforcability to complete devices with embedded systems, i.e. devices which are not laid out for reprogramming by the user. This method would protect much of the traditional interests of the telecommunications and industrial software patent owners while keeping those spheres free of patents where an ever-increasing number of freelancers and small companies are making significant contributions to new developments.
The first two features explain why much software innovation comes from individual developers and very small software publishers based on the Internet. It is a justification for protection measures in favour those two groups. The last feature is an argument for stronger competition protection measures.
The epistemologic nature of software is equally striking:
These properties deserve a lot of attention from a legal point of view. They require to consider with great care the meaning of infringement and contributory infringement induced by the use of single software or multiple software. Moreover they make it necessary to define a clear limit to the patent system in order not to let crude control mechanisms which were designed for material objects reach out into the sphere of the human mind. Some writers like Dr. Kiesewetter-Köbinger, patent examiner at the German Patent Office, have therefore argued that software patents constitute a radical breach of basic values of our civilisation and an attack on central tenets of our (German) constitution, ranging from Art 1 "The Dignity of the Human Being is Untouchable" and Art 2 (equality principle) to freedom of expression, freedom of contract design and the requirement that the intellectual property, which in the case of software consists in the tissue, be protected rather than expropriated.
Other approaches to regulation exist and provide similar effects: introducing explicit exceptions and limitations in order to enforce interoperability, to protect open source / free software and to protect small publishers against legal terrorism. Such approaches should be seriously considered and compared to regulation approaches.
Economy however is not the only issue in the case of software patents. Law consistency issues should be addressed too. For example, any extension of the patent system to software without justifications of a positive economic impact may be in violation with Article 10 of the European Convention on Human Rights. Up to now, no one has ever seriously argued let alone proven that software patents have any positive economic effects.
Also, the epistemologic nature of software creates unique issues. For example, the ownership of software innovations generated by "invention machines" is yet unknown. Failing to design a consistent and adequate patent law may achieve the opposite of "clarification" and "harmonisation": higher legal uncertainty than ever before.