It is now time to compare the legal and technical findings of the paper with a sketched economic model of the legal protection of software works, also taking into account the other economic insights I outlined in the General Introduction to the dissertation. I can anticipate that – if applied as I suggested – the so-called “technology copyright”237 model not only works in a reasonably sensible way, but it is much more consistent
with economic insights than alternative, more protectionist, models. In fact, a higher level of protection potentially offered by an alternative patent-like model would likely raise several problems, without solving major incentive problems. Indeed – as I will try to show – major market failures due to the lack of incentives do not exist even in a more flexible copyright-based model. As I discussed in the General Introduction, I will start from the idea that the typical condition for an efficient market to work is not that producers are able to appropriate the entire social benefit generated by their activity. Such a condition has an interesting property: it ensures that – in a static setting – each and every socially profitable investment is performed. However, it does not consider the huge transaction costs generated by a capillary system of exclusive powers, able to extract the entire surplus generated by a given innovation. What is sufficient to have an efficient market is that investors are able to recoup their costs (taking into account risk), as long as consumers appreciate their products. Obviously, in a market in which there are investments and fixed costs, this condition cannot be respected by the equality among price and marginal costs. Instead, it is verified when “gross profits” are equal to sunk costs.
As long as reverse engineering is not prohibitively difficult and/or hindered by the law (a problem that I will address in the second paper of this dissertation), coupling copyright (preventing literal copying and other form of parasitism) and trade secret (giving limited but significant protection to some categories of innovations) could create a structure of incentives working as a quasi-liability system in the field of software interfaces.238 This system avoids two polar and opposite risks of the patent systems. On the one hand,
rightholders cannot block subsequent innovation for strategic reasons and self-help remains an available solution also in cases in which transaction costs (frequently increased by strategic behaviour and/or asymmetries of information) would have prevented the working of a market. On the other hand, significant incremental innovation, which would not qualify for patent protection, is still protected against easy and cheap appropriation from third parties. A quotation from Prof. Reichman’s seminal 1994 article may be useful in clarifying the role of trade secret as a stimulus for innovation:
“Legal theorists have particularly underestimated the important role of trade secret laws (or equivalent laws of confidentiality) in mediating between formal intellectual property regimes and free competition. These laws do not confer exclusive property rights in the manner of patent and copyright laws, but they do require would-be competitors to extract an innovator's undisclosed know-how by proper methods of reverse engineering.[…]
Second comers who cannot extract valuable undisclosed information by proper means or independently reach similar solutions must acquire the unpatented know-how through licensing agreements with
235 For instance, I am not aware of specific data concerning the proportion of application designed for Windows 95, 98, XP or
Vista that may run under Linux or Mac using Wine, but this could be an interesting illustrative example and my guess and personal experience is that it is fairly easy to run the oldest one without a copy of Windows 95 – but, obviously, this does not pose a significant competitive threat to Microsoft, since these old application can be considered quite obsolete.
236 See the second paper of this dissertation project for a fuller discussion of these issues.
237 I will put a higher emphasis on copyright (instead of patents) because this is the main tool of actual protection of software
related innovation in all legal systems I am aware of. However, I will briefly touch problems related to patent protection of software in § 9.2. May patent law (as currently applied to software) limit interoperability?.
238 See J.H.REICHMAN, Legal Hybrids Between the Patent and Copyright Paradigms, 94 Columbia Law Review, 2432 (1994) and several
innovators. Either way, these legal requirements normally provide those who develop unpatented, noncopyrightable innovation with a period of natural lead time in which to recover their investments while establishing their reputations as producers of quality goods.”239
In fact, I have to stress that Prof. Reichman does not believe (or, at least, did not in 1994) that trade secret is likely to work as described above in the field of software. Indeed, in this and in similar fields, Prof. Reichman thought that innovation is too “near to the surface” of products and risks being easily appropriated, so that the protection offered by secret would be insufficient. Hence, he proposed that the law should reinforce the natural lead-time of innovators, exposed to an excessively cheap reverse engineering:
“In dynamic economies driven by constant technological innovation, competition with respect to the products and processes of routine innovation thus presupposes a degree of natural lead time, which classical trade secret law is presumed to supply. Absent natural lead time, however, investment in innovative but unpatentable applications of science to industry tends to dry up, and competition may languish in the face of a progressive market failure. As will be seen, the likelihood that this type of market failure will occur has greatly increased during the second half of the twentieth century.”240
Despite the fact that this influential article was very likely correct in its theoretical analysis, I suggest that it was slightly less well-grounded in terms of empirical and technical evidence. And what happens in the “real world” in crucial in drawing the consequences of Prof. Reichman’s theoretical work, in fact:
“The term “natural lead time,” though analogous to the economists’ “head start,” […] is chosen to emphasize the extent to which implementation of the governing legal regimes -- trade secret law and the law of confidential information -- depends upon real world events and not legally determined outcomes, such as a fixed term of duration. Natural lead time depends on the individual business decisions and strategies of both innovators and borrowers, not on the dictates of legal rules.”241 [And, I would add, it
crucially depends on the state of technology.]
In fact, I do not think that reverse engineering in the field of software is so cheap that it is likely to create market failures, and I will try to show this in the second paper of this dissertation. Moreover, in that paper I will also show that the law already reinforced significantly trade secret in the software field, hindering reverse engineering. Actually, I will even argue that this reinforcement is excessive and that it could create some market failures similar to the one of a strong patent system; however, should I be wrong, one could see this reinforcement of trade secret in the field of software as an application of Prof. Reichman’s analysis.242
Overall, having different assumptions concerning the cost of software reverse engineering, I share the theoretical analysis of Reichman, but the conclusion I reach about software is almost diametrically opposed to his own. In fact, this difference very probably derives from the fact that Prof. Reichman was analysing innovation in the field of new technologies in general, while I am focusing on a specific kind of technical innovation (interface information), which is much easier to “keep secret” than the average software innovation. In other words (and using extreme examples), if one invents a “word processor” it is evident that this is innovation so “near to the surface” of the product that it can be easily appropriated by other players. The same may be true for innovations concerning user interfaces in general (because these elements are inherently “near the surface” of the software product). However, if one creates an innovative set of APIs the fact of just distributing the humanly not understandable object code, coupled with some appropriate non- disclosure agreements with producers of complementary products, may be very effective in giving years of lead-time to the first comer (easily 5 or 10 years, as some examples will show in the second paper). Thus, at
239 Id., pp. 2438—2439. See also pp. 2440—2441: “On the margins of the pure market economy envisioned by nineteenth-
century liberal economic thought, trade secret laws (and related laws protecting confidential information) thus provide a loosely constructed set of liability rules that reinforce the competitive ethos in subtle and indirect ways. These modified or quasi-liability rules mediate between the potential for overprotection inherent in the statutory grants of exclusive property rights and the potential for underprotection inherent in the competitor's unfettered ability to appropriate the fruits of investment in unpatented incremental innovation.”
240 Id., p. 2442. 241 Id., f.n. 25.
242 Actually, the proposal of Prof. Reichman would be much more sensible, from an economic point of view, of the generalised
reinforcement of trade secret in the field of software that one may observe in modern legal systems. In more practical terms, the quasi-liability system proposed could “provide a maximum blocking period of [two, three or four] years, during which time competitors could not enter the same market with substantially the same unpatented product absent an agreement with innovators” (Id., f.n. 522 and accompanying text.). In the field of APIs this could mean that horizontal interoperability could not be achieved by reverse engineering for a certain period of time (during which – however – study and experimentation would be free).
least in the field of software interfaces, it remains possible to use property rights (and copyright in particular) – with their innate pro-market characteristics – instead of creating special liability rules.
Incidentally – and even if I did not (and will not) discuss the possibility of sui generis protection for software243 – the first part of the present work made clear how precious the general principles and doctrine
underlying copyright law are in order to tackle several complex and cutting edge issues arising in the field of the legal protection of computer programs. An ad hoc model of protection would risk lagging constantly behind technological innovation and would likely provide minor advantages, with respect to copyright, as long as copyright is not excessively expanded (in the direction of a quasi-patent right) and it is interpreted considering the technological reality, as I suggested in the paper at hand.244 For this reason, I share the
conclusion of several authors245 arguing that the desirability of this kind of ad hoc protection (even though
theoretically sustainable as a first best solution, tailored for software) raise several serious questions, mainly concentrated on the lack of experimented and shared legal principles.
In summary, I clearly agree with Prof. Reichman when he argues that “the nineteenth-century vision that subdivided world intellectual property law into discrete and mutually exclusive compartments for industrial and artistic property has irretrievably broken down.”246 However, the breakdown of this divide does not mean
that we should resort to a high number of sui generis paradigms. The patent and copyright paradigms remain perfectly valid and both useful depending on the kind of innovation that they have to protect (at least as long as they are interpreted as I suggested in the General Introduction) and even if copyright is no longer used just as a tool to protect artistic works. In fact, Reichman argues that software and other subject matter protected by “hybrid legal institutions”, such as plant varieties, “violate the negative economic premise that limits copyright protection to cultural goods”.247 As I tried to show in this paper (and as I will discuss further in the second
one), I argue that there is not such an “economic premise”, at most there is an “historic premise” saying that copyright protects just aesthetic creation. Economically, the fact of protecting just expressive form and not the underlying ideas is not necessarily an aesthetic category: in the field of software, this means that the general structure, purposes and methods of a program are not protected, while the actual code is. And writing this code is a very significant line in the budget of a software house (that I would label “development cost”, as opposed – but complementary – to pure “research costs”). To be sure, the “historic premise” according to which copyright concerns artistic creations has some consequences (as the already mentioned and unreasonably long duration of copyright protection), but – in general – copyright remains a good tool to protect all those kind of innovative sectors in which Development costs are significantly higher than Research costs (in the sense I already explained), so that a partial free riding on ideas and principles is not capable of creating significant market failures.248
Finally, I acknowledge that adopting the interpretation of APIs’ legal protection proposed in this paper may seem to largely disregard incentives to realize interfaces. Yet, I already clarified – and the second paper will make even clearer – that, because of the cost of reverse engineering software, there is actually a quite broad scope for the licensing of API specification documents (also under non-disclosure agreements). In any case, I concede that the kind of incentive to create that comes from the possibility of licensing specification documents is not very similar to true property rule. Instead, it is based on something similar to a quasi- liability rule. In other words, licensing API specifications in a world where reverse engineering is allowed and copyright does not protect specifications is equivalent to “selling” (substitutes of) the reverse engineering
243 About sui generis protection of software, see PAMELA SAMUELSON, et al., A Manifesto Concerning the Legal Protection of Computer Program, see id., 2308--2431 and the debate generated by the Manifesto (see, in general, the Columbia Law Review issues of 1994:
JANE C.GINSBURG, Four Reasons and a Paradox: The Manifest Superiority of Copyright over Sui Generis Protection of Computer Software, 94
Columbia Law Review, 2559--2572 (1994); PAUL GOLDSTEIN, Comments on a Manifesto Concerning the Legal Protection of Computer
Programs, 94 Columbia Law Review, 2573 (1994); PETER S. MENELL, The Challenges of Reforming Intellectual Property Protection for
Computer Software, 94 Columbia Law Review, 2644 (1994); ZENTARO KITAGAWA, Comments on 'A Manifesto concerning the Legal Protection
of Computer Programs', 94 Columbia Law Review, 2610--2620 (1994)).
244 See BRETT A.CARLSON, On the Wrong Track: A Response to the Manifesto and a Critique of Sui Generis Software Protection, 37
Jurimetrics J., 187 (1997).
245 See, in particular, GINSBURG, Four Reasons and a Paradox. Several of the authors mentioned in footnote 243 and criticizing the Manifesto share similar conclusions.
246 J.H.REICHMAN, Legal Hybrids Between the Patent and Copyright Paradigms, see id., 2432, 2500. 247 Id., 2502.
248 And – as I already explained – this is even truer in the field of software, given the advantages of first comers (direct and
indirect network effects, learning costs and other switching costs, etc.) and the cost (and imperfections) of software reverse engineering.
activity needed to discover interfaces, not to license the interface specifications themselves.249 Moreover, even
assuming that reverse engineering would be so simple as to destroy direct incentives to innovate (which is an absurd assumption at present), alternative sources of incentives to the realization of good interfaces abound. Actually, the majority of advanced two-sided models concerning software platforms (i.e., the kind of programs that typically expose more interfaces) suggest that controllers try to subsidize developers of complementary software;250 this implies that there is actually a strong incentive, in order to maximize the
value of a platform, to develop and give away for free high quality interfaces. Obviously, problems remain about competing software (i.e. horizontally interoperable software), but here an empirical example may be useful. After all, several open source software are perceived as being of very high quality – in some cases even better than their commercial counterparts – but this does not allow commercial firms to develop their one competing software at no cost (unless they distribute themselves as open source under the GPL or similar licenses). Why so, if knowing the source code of a competitor is enough to free ride on it? Probably simply because that is not enough at all, and well-written software does make the difference, even when the ideas behind it are free to be taken.
8.1. The limits of technology copyright and its natural antibodies
Of course, basing the creation of incentives to write new and good software interfaces on a de facto quasi- liability rule, deriving from the combination of copyright and trade secret (what I call “technology copyright”), is not a panacea and it entails several costs. In particular, Prof. Reichman observed that
On the negative side, this substratum of modified liability rules entails appreciable social costs of its own. It is also singularly prone to yield arbitrary and irrational results whenever the task of reverse engineering unpatented, non-copyrightable innovation proves either too difficult or too easy.251
Differently from what Reichman expected (in 1994), this task normally proves too difficult – and not too easy – in the field of software (and I will discuss more about that in the second paper of this dissertation). Moreover, the actual working of the industry, coupled with copyright, prevents “incremental innovators” from taking someone else’s software as a basis for writing new software. In this way, new bugs are introduced every time a piece of software is developed and it is quite frequent for developers to “reinvent the wheel”. In other words, “technology copyright” is costly, since it slows down the spreading of innovation (and indirectly generates waste through reverse engineering).252 At the same time, even copyright alone would probably not
be a perfect tool to protect software. In fact, it allows developers to learn from others (as long as they can access source code), but then forces them to start from scratch every time they write an application (frequently they do not even use their own past pieces of code, because they are owned by past employees). A specific analysis of these problems has been proposed by Lemley and O’Brien in an article appropriately titled “Encouraging Software Reuse”.253 Also Determann254 highlighted this problem, stressing the faults of
copyright in determining this outcome:
“Since courts first decided to afford copyright protection to computer programs, commercial software development companies have had a strong incentive to avoid reusing existing code owned by others. Independent creation is a defense to copyright infringement, and so software development companies often opt for creating programs from scratch, ideally in a “clean room” environment, so they can prove that their products are not copies of existing programs with similar functionality. Thus, the decision in favor of software copyrightability had a rather dramatic impact on the professional lives and day-to-day activities of programmers: instead of being asked to further develop and improve the “state of the art” and to focus on cutting-edge problems, programmers were asked to spend most of their time reinventing
249 But one has to be aware that it is possible to directly license interface implementations.
250 See J.C.ROCHET & J.TIROLE, Two-Sided Markets: A Progress Report, 37 RAND Journal of Economics, 645--667 (2006) (or J.C.
ROCHET & J.TIROLE, Two-Sided Markets: An Overview, IDEI Toulouse working paper (March, 2004)).
251 REICHMAN, Legal Hybrids, p. 2441. Cf. also LOTHAR DETERMANN, Dangerous Liaisons -- Software Combinations As Derivative Works? Distribution, Installation, And Execution Of Linked Programs Under Copyright Law, Commercial Licenses, And The Gpl, 21 Berkeley
Technology Law Journal, 1421 (2006), p. 1437: “copyright law strikes a delicate balance between access and exclusion rights”, so that “[b]oth under- and over-protection can harm the public interest in creative works.”
252 See RICHARD R.NELSON, Intellectual Property Protection for Cumulative Systems Technology, 94 Columbia Law Review, 2674 (1994),