Competition Policy for Computer Software Markets

  John Hogan
School of Law,
Trinity College Dublin
johnghogan@hotmail.com
 

I wish to thank Cathryn Costello, of the Law School, and Dr Francis O'Toole, of the Economics Dept, for their invaluable guidance. Suggestions from the JILT editorial team were also of particular assistance. Any errors or shortcomings are of course my own.

Abstract

Against the backdrop of the ongoing Microsoft antitrust litigation, this article examines issues of competition policy for computer software markets. A 'dynamic compatibility regime' is proposed. This approach would recognise that strong intellectual property protection works to provide the incentives necessary to encourage firms to engage in winner-takes-all 'standards races'. However, once a standard is established, the argument for such protection is far less convincing: quality access to the underlying technical information is likely to be indispensable for effective, ongoing competition in related, or 'downstream', markets.

Section One begins by considering the policy goal of 'maximising consumer welfare' within the context of computer software markets. There follows an analysis of the economics of network industries in general and software markets in particular. The nature of competition in software markets is examined; this leads to the conclusion that policy should encourage compatibility.

Section Two discusses the Microsoft case. It is argued that the approach taken by the DOJ and the Circuit Court fails to convince and that 'Raising Rivals' Costs' theory represents a means of analysing the issues that better reflects the nature of competition within the software industry. 'Raising Rivals' Costs' (RRC) involves conduct that raises costs and induces rivals to restrict their output, thereby allowing the dominant firm to exercise monopoly power. Microsoft illustrates that software markets are prone to RRC through technological input foreclosure, e.g. engineered incompatibility, or the denial of quality access to the interface information necessary to produce interoperable products. These 'incompatibility strategies' are examined in Section Three.

Section Three begins with an analysis of the antitrust treatment of dominant firm conduct designed to render previously interoperable products incompatible with an established standard. A focus on the complementary nature of the relationship between a dominant firm and the firms with which it competes in downstream markets is favoured. The final discussion considers whether a dynamic compatibility regime can extend beyond situations where a change in policy has resulted in incompatibility, to imposing a positive obligation on dominant firms to allow access to technical information to the extent necessary for the development of interoperable products. An argument in favour of qualifying both the term and scope of copyright protection for software is constructed.

This is a Refereed article published on 2 July 2001.

Citation: Hogan J, 'Competition Policy for Computer Software Markets', Refereed article, 2001 (2) The Journal of Information, Law and Technology (JILT). <http://elj.warwick.ac.uk/jilt/01-2/hogan.html>. New citation as at 1/1/04: <http://www2.warwick.ac.uk/fac/soc/law/elj/jilt/2001_2/hogan/>.

1. The Nature of Competition in Computer Software Markets

1.1. Introduction

The guiding principle of competition policy is the maximisation of consumer welfare[ 1 . This requires that 'society's resources are allocated so that consumers can satisfy their wants as fully as technological constraints permit'[ 2 . Competition policy regulates the 'means' of the competitive process, not the 'ends' produced by that process: it should not attempt to decide which technology or industry structure is 'right', but instead should attempt to maximise the influence of users in determining outcomes[ 3 . Shapiro argues that the fact that the 'key driver of consumer benefits in information industries is technological progress' requires that the primary goal of antitrust must be 'to promote and protect competition in the introduction of new and improved products and services'[ 4 ].

It has been said that competition in the computer industry is characterised by long 'eras' of stable structures and standards punctuated by 'epochs' of wrenching change, where firms engage in fierce, winner-takes-all 'standards races'.[ 5 ] A theme of this article will be how competition policy can best encourage ongoing competition on the basis of price, quality and innovation within those eras of stable industry structures and standards.

1.2. Network Effects and Standards

Network industries[ 6 ] are characterised by demand-side economies of scale known as 'network effects': the value a user ascribes to a particular product increases as the number of users of that product increases ('positive feedback'). This may include situations where one user's value for a good increases when another user has a compatible but non-competing good, i.e. an interoperable good. Microsoft's Windows operating system is an obvious example. In a 'virtuous cycle', as the number of Windows users increases, software developers face increased incentives (in the form of larger markets) to write Windows-compatible applications. The increased number and variety of compatible applications makes Windows more attractive, enticing further users. And increased use creates an information asset that can guide future development, allowing for added features and improved quality.

The fact that components of the Windows network (i.e. the operating system and the range of compatible applications) are interchangeable and backwardly compatible allows users to stick to the same standard over time. However, the other side of that coin is that users' sunk costs can lock them in: a move to a new standard requires co-ordination among both the users of the existing network and the suppliers of complements, which is extremely unlikely[ 7 ]. So, 'the very demand-side economies of scale that induce the formation of a network in the first place can serve as barriers to competition against the network'[ 8 ]. And, insofar as they facilitate recoupment, the entry barriers that network effects represent make strategies for excluding or weakening rivals more feasible.

Standards are the protocols shared by network participants, necessary for interconnection with users and interoperability with complements: the interface technology underlying the Windows operating system is an example of a standard[ 9 ]. Once a standard exists, software manufacturers will not be able to compete unless their products are compatible with that standard.

1.3. Competition in Software Markets

The previous section demonstrated that software markets are characterised by network effects. Another feature of software markets is that they involve very high 'first-copy' or sunk costs but very low duplication costs, i.e. firms enjoy virtually 'instant scalability'[ 10 ]. This combination of demand-side economies of scale (network effects) and supply-side economies of scale (instant scalability) means that software markets tend to be highly concentrated.

While some commentators have argued that network effects dictate that market power may be transitory, characterised by 'serial monopoly', it seems clear that virtuous cycles are hard to stop and start, but relatively easy to maintain: the costs of advancing a standard are far less than the costs of introducing a new standard[ 11 ]. In his Findings of Fact , Jackson J offers a useful summary of the nature of competition in software markets:

Obviously, introducing a new standard is not just about 'building a better mousetrap': apart from the obvious technical expertise, it also requires marketing and management skills and a range of complementary components. However, network industries involve a particular barrier to entry: the market may 'tip' (to monopoly) in favour of the product that achieves an early lead[ 13 ]. Tipping is driven by consumer expectations, which will be influenced by factors such as a firm's reputation from other markets, its installed base of users and its current products - factors that will tend to favour the incumbent.

Ultimately, however, the network effects theory is ambiguous in its welfare implications: the socially cost-minimising structure may or may not be very concentrated. Melamed has noted that where:

This is so even if consumers individually prefer the new (and 'better') technology:

Network effects make compatibility 'a critical dimension of [software] industry structure and conduct'[ 16 ]. In a network market where products are incompatible, the positive feedback operates at the level of each product individually. It has already been noted that consumers' expectations as to sales will influence durable investment decisions such as software purchases. Under incompatibility, because these expectations are based on the sales of the individual firms in the market, they will favour the dominant firm. Compatibility, however, neutralises the dominant firm's current installed base and expected sales as sources of competitive advantage, i.e. it allows the positive feedback to operate at the level of the market as a whole:

Encouraging effective competition within an established standard requires that antitrust policy compels compatibility, or more particularly, that it enables the development of interoperable, or 'downstream', products. US antitrust enforcement agencies have on occasion mandated compatibility or 'open access' as a condition of approving software mergers[ 18 ]. In the context of Microsoft , which falls for consideration in Section Two, a compatibility regime would require the disclosure of the APIs that hook Internet Explorer (IE) to other parts of the Windows OS, allowing competing browsers the same degree of interoperability with Windows as IE enjoys[ 19 ].

2. The Microsoft Case

2.1. Introduction

The Microsoft antitrust litigation arose out of Microsoft's response to the threat presented by the 'disruptive technology' of the Navigator-Java platform[ 20 ]. The combination of the Navigator browser and the Java programming language was central to the 'thin client' initiative: the 'client' (computer) would feature only basic central processing components, key peripherals, an operating system and a browser; data would be stored and processed at server-level, to be retrieved by the client as needed. APIs would no longer be OS-specific but would instead be exposed by 'middleware' running on top of an OS. This cross-platform compatibility would erode the 'applications barrier to entry' and commoditise the underlying OS[ 21 ]. Consumers would no longer have to take into account the number of applications expected to become available for a specific OS and developers would no longer have to consider which OS would become the standard among consumers.

As a monopolist in the OS market, Microsoft had a clear incentive to prevent the cross-platform compatibility of the Navigator-Java platform. The 'browser wars' were just one element of a broader strategy aimed at preserving the OS monopoly by impeding the adoption of Navigator-Java. So, Microsoft competed head-on with Netscape in the browser market, it made Sun's Java language incompatible with aspects of Windows and it promoted a rival, Windows-specific, version of Java. However, antitrust analysis in the case thus far suffers from a fatal problem of characterisation. Jackson J portrays IE's zero-price as predatory pricing, the bundling of IE and Windows as an illegal tie-in, and the arrangements with OEMs and ISPs as instances of exclusive dealing. As an inevitable result of these characterisations, the arguments advanced under each claim are weak[ 22 ].

2.2. No Such Thing as a Free Browser?

Characterising Microsoft's zero-pricing of IE as predatory pricing is bound to fail, simply because there are plausible economic justifications for a zero browser price. Consider the difficulties Microsoft faced as an entrant to the browser market: Navigator 'already enjoyed a very large installed base and had become nearly synonymous with the Web in public consciousness'[ 23 ]. It could be argued that the zero-price was an instance of penetration pricing necessary to overcome the enormous barriers to entry represented by the network effects working in Netscape's favour. Alternatively, as the marginal cost of producing an extra unit of a software product is zero, it could be argued that a rational firm can be expected to set price at marginal cost.

The Coase Theorem offers a further possible explanation for the zero-price bundling of IE: Windows is a durable good, so Microsoft can be expected to innovate by adding functionality to its OS (in the form of web-browsing capability) so as to maintain demand[ 24 ]. Finally, a zero price for IE could be explained by analogy to the supply of free programming on broadcast TV, in which case the browser's ability to generate advertising revenues and commissions by steering Internet users to particular web-sites makes the marginal cost of distributing another unit negative[ 25 ].

2.3. Did Bill Gates Twist Your Arm?

To establish an illegal tie-in, Jackson J had to identify two separate products. So, he argues that 'the commercial reality is that consumers today perceive operating systems and browsers as 'separate products', for which there is separate demand'[ 26 ]. Arguably, Microsoft's response is more convincing: the application of the 'separate consumer demand' test would 'kill innovation to the detriment of consumers by preventing firms from integrating into their products new functionality previously provided by standalone products - and hence, by definition, subject to separate consumer demand'[ 27 ].

The other element of an illegal tie-in is 'forcing', and the argument here is equally problematic. Consumers were simply not compelled to purchase a product they did not want: they were not charged anything for Internet Explorer and they remained free to install Navigator as their default browser:

Jackson J's argument that hard-drive space is 'scarce and valuable' is difficult to accept[ 29 ], and the fact that IE was zero-priced when it was sold separately casts aspersions on his claim that 'any value to be ascribed to Internet Explorer is built into [the] single [Windows] price'[ 30 ]. The technical integration of the browser and OS in Windows 98[ 31 ] is properly characterised and considered as an instance of raising rivals' costs through contrived incompatibility, but characterising it as a technological tie, Jackson J strains to find the requisite consumer harm:

If browser prices remain competitive (i.e. at MC, or zero) it becomes difficult to identify harm to consumer welfare, and equally difficult to justify intervention[ 33 ].

2.4. Compaq's Self-Twisting Arm

In his Findings of Fact Jackson J devotes considerable space to the contractual restrictions imposed on OEMs preventing them from removing Internet Explorer from Windows. In reality, these restrictions may have been nothing more than the result of bargaining between industry players[ 34 ]. Microsoft imposed certain restrictions on OEMs' ability to reconfigure the desktop and the start-up sequence, but the explanation that this was in order to ensure a common 'Windows experience' for all users seems plausible[ 35 ]. And OEMs were granted discounts off their Windows royalty prices to encourage compliance[ 36 ]. Rather than regarding this as the normal quid pro quo of contractual negotiations, Jackson J seems to see the OEM-restrictions and the discounts as evidence of a predatory intent:

This misses the point completely. OEMs' costs were not raised: they were compensated for the costs of complying with the restrictions - and they could still pre-install Navigator:

The following example from the Findings of Fact clearly demonstrates that Microsoft's arrangements with OEMs involved inducement rather than coercion[ 39 ]. In early 1996, Compaq partnered with Netscape: Netscape seems to have offered Compaq a discount for an exclusive slot on the desktop[ 40 ]. Insofar as the inclusion of Internet Explorer eliminated this ability to sell an exclusive to Netscape, its equilibrium price to OEMs was negative; as Netscape's product was significantly superior at that time, Microsoft would have had to make a very large positive payment (or reduce the OS price) to compensate OEMs such as Compaq. Arguably, it was more economic to engage in a short-term opportunistic contractual 'hold-up', whereby Microsoft enforced the provisions of its OEM licensing contracts and prohibited OEMs from removing any part of the Windows OS[ 41 ]. This prevented Compaq from selling its 'browser slot' to Netscape on an exclusive basis, but while it would have had short-run wealth distribution effects it was not of any competitive significance: there was no harm to consumers, nor was there any anti-competitive exclusionary effect on Netscape - OEMs could still install Navigator and make it the default browser[ 42 ]. In February 1997, Compaq aligned itself with Microsoft; it benefited from lower Windows license fees and a bounty for each Compaq user that signed up for Internet access, i.e. the new licence took account of the impact of the growth of the Internet on the value of the browser slots on the desktop.

Jackson J describes Microsoft's purchase of Compaq's partnership as a 'massive and multifarious' investment, but a more plausible portrayal would be that of a bargain agreed upon by two sophisticated firms. Note that in January 1999, after Netscape agreed to provide it with approximately $700,000 of free advertising, Compaq resumed pre-installation of Navigator on its Presario computers[ 43 ].

In the same way, a critical reading of Microsoft's dealings with ISPs suggests a pattern of inducement, not coercion[ 44 ]. Microsoft's Appellate Brief notes the difficulties faced by the entrant to the browser market:

It is perhaps not surprising that Jackson J's attempt to attribute some anti-competitive quality to Microsoft's dealings with AOL instead suggests that those dealings were nothing more than 'bargained-for exchanges of consideration'[ 46 ]:

And there appears to be a tacit acceptance that the antitrust concern is not the OEM exclusives but rather the issue of technical integration:

This is important: Katz and Shapiro have suggested that competition policy should distinguish between the release of a 'bundled' browser at a low or zero incremental price and instances where a dominant firm imposes incremental costs on the developers or users of rival browsers. Intervention can be justified in the second case[ 49 ]. This approach leads to the following characterisation of the potential antitrust issues in Microsoft:

• Microsoft's OEM/ISP exclusives may have disrupted optimal distribution patterns in the browser market, artificially raising the price Netscape would have to pay to secure distribution through those channels;

• Microsoft may have withheld Windows interface information from rivals;

• Microsoft may have disadvantaged rivals by engineering incompatibility into its OS.

This sits better with the factual background to the case. The imposition of artificially higher distribution costs, the denial of quality access to necessary interface information, and engineering incompatibility with rivals' products are instances of dominant firm conduct best analysed under the theory of 'Raising Rivals' Costs'.

2.5. Raising Rivals' Costs

'Raising Rivals' Costs' (RRC) is the exclusionary exercise of market power to raise or maintain prices above the competitive level. It involves conduct that places rivals at a sufficient cost disadvantage that they are forced to restrict output, allowing the defendant firm to exercise monopoly power by increasing price[ 50 ]. In contrast to predatory pricing (where the predator can be expected to lose money faster than its smaller victim) it may be relatively inexpensive for a dominant firm to substantially raise its rivals' costs. Nor are there any difficulties in recouping the predatory investment: a higher-cost rival will quickly reduce output, allowing the predator to raise price or market share. And as it is always better to compete against high-cost firms rather than low-cost ones, RRC strategies can be profitable without the rival's exit from the market.

Input foreclosure, whereby rivals are denied quality access to necessary inputs, is recognised as an effective means of implementing RRC strategies[ 51 ]. The same is true of contracts with distributors: by disrupting optimal distribution patterns, a dominant firm can impose costs on its rivals. For example, Microsoft's exclusive arrangements with OEMs and ISPs may have raised Netscape's costs of distribution or reduced the size of its installed base of users, thereby raising its marginal costs across the board.

It could be countered that Netscape had equally effective counter-strategies available to it, but once a firm is forced to pay not to be excluded its costs have already been raised. In any case, the fact that a predator outbids its rivals for the purchase of exclusive rights does not mean that the exclusion is economically efficient: the market for exclusionary rights is a market for competition, which is a classic public good. So, even a well-functioning market will fail to yield an efficient outcome[ 52 ]. And the incentives each party faces suggest that this type of RRC strategy may be particularly effective: while the purchaser of the exclusive rights stands to gain increased market power and additional profits, the potentially excluded rivals only gain the more competitive non-exclusion price and profits. If, however, the rivals reduce their output, they gain the benefit of a higher price on their remaining sales: essentially, the purchaser has more to gain than the rivals have to lose.

'Technological' input foreclosure by a dominant firm controlling a software standard such as Windows represents a more interesting RRC strategy than the purchase of exclusionary rights. Rivals could be denied quality access to important inputs such as the interface information necessary to produce interoperable products. It is clear that Microsoft's control of the Windows OS gives it considerable power to raise its rivals' costs in this way[ 53 ]. 'Contrived incompatibility' represents a similar, perhaps more subtle, form of technological input foreclosure, and it seems that efforts were made to make running another browser on Windows 98 a 'jolting experience'[ 54 ]. The passage below illustrates the possible anti-competitive effects of an incompatibility strategy; Section Three will consider how competition policy responds to such conduct.

3. The Antitrust Treatment of Incompatibility Strategies

3.1. Introduction

Dominant-firm incompatibility strategies can be broadly categorised as follows:

• The denial of timely access to interface technologies, such as the Windows APIs, which are necessary to produce interoperable products.

• Contrived incompatibility, i.e. changes made to an existing standard that render previously interoperable products incompatible with the standard.

In practice, however, some overlap can be expected: for example, a technological tie-in may preclude access and interoperability.

3.2. In Defence of Incompatibility

It has been argued that an 'informal' tie-in by innovation or integration can be beneficial in that it secures for the innovator a greater degree of control over the quality of the new product than would otherwise be the case:

It has also been argued that in an environment where the return on investment in innovation is very uncertain a technological tie-in works to protect the innovator, allowing him to extract the exclusive benefit from his new product for some limited time, i.e. until it is reverse engineered[ 57 ]. But all this ignores one crucial possibility: the fact that the new version of the monopolist's product succeeded in the marketplace may simply reflect the absence of any viable choice. For much the same reason, the judicial treatment of dominant firm incompatibility strategies is unsatisfying.

In Foremost Pro Colour v Eastman Kodak [ 58 ] the plaintiff photofinisher alleged that Kodak's development of new products that were incompatible with existing photo-finishing equipment amounted to a technological tie: the new film format could not be processed as before, so it was necessary to purchase a package of film, chemicals and paper. The Court rejected the argument that this was per se unlawful, requiring instead some evidence that the innovation was motivated by a desire to compel the purchase of the entire system:

The claim that contrived incompatibility 'without more' does not foreclose competition is intellectually dishonest: the antitrust issue is that contrived incompatibility raises rivals' costs, allowing the dominant firm to exercise market power.

California Computer Products v IBM ( CalComp )[ 60 ] involved a claim of (direct) contrived incompatibility: the plaintiff argued that IBM's changes to the design of the interface between the central processing unit (CPU) and peripherals amounted to 'technical manipulation'[ 61 ]. The Court seemed to consider price and performance as inseparable, holding that where an innovation provides equivalent function (i.e. no improvement in performance) at a lower price the result is to make the product more attractive to buyers. As such a cost-saving step, the challenged integration could therefore be considered an improvement[ 62 ]. This approach was echoed by the Circuit Court in the litigation arising out of the Microsoft consent decree[ 63 ]. It was held that Windows 95 and Internet Explorer 4.0 constituted a single, integrated product if there were:

Alluding to issues of institutional competence, the Court continued:

To be sure, antitrust challenges to innovations must be handled with care[ 65 ] but it is submitted that a slightly more robust approach would be preferable. Under the EU competition regime, it seems that a software company is free to make design changes 'as long as it does so objectively to improve its own product or service, and not primarily with the effect of making difficulties for its downstream competitors'[ 66 ]. And the proportionality principle could act as a further 'limiting principle': a dominant company is not free to cause substantial inconvenience to its competitors to achieve a minimal improvement in its own product[ 67 ]. In contrast, under the 'sole purpose' standard of the cases considered above, an antitrust plaintiff must establish that the challenged innovation is a device without any 'facially plausible benefits'.

3.3. The Wrong Turn

The Foremost Pro Color line of authority began with Berkey Photo v Eastman Kodak [ 68 ]. There, it was argued that Kodak, as a monopolist in the markets for cameras and photographic film, was in a position to set industry standards and that rivals could not compete without offering similar products to Kodak's. As a result, Kodak was obliged to provide advance information to enable rival camera manufacturers to adapt their products to new film formats it planned to introduce[ 69 ]. This argument was roundly rejected. The Court held that Kodak had no duty to provide advance information to its rivals in the camera market[ 70 ]:

This approach seems largely correct. Kodak had developed an entirely new camera - it had not altered an existing format, nor did it restrict output of existing camera formats so as to boost sales of its new camera and film. Its competitive advantage (the lead time) was a direct result of its innovation in bringing a new product to market - and it was an advantage that would inevitably be eroded as its competitors caught up. Yet in the cases that followed, these distinguishing points were effectively ignored.

Antitrust claims arising out of the introduction of entirely new technology, e.g. Berkey Photo , may lack one crucial element: a pre-existing complementary relationship between the plaintiff and defendant. In the next section, I will demonstrate that a rival producing interoperable products is in a complementary relationship with the dominant firm controlling the existing standard and must therefore be afforded an adequate opportunity to adapt to changes in the standard. This reflects commercial reality within software markets: the initial sponsor(s) of a standard will encourage the development of interoperable products as a means of ensuring rapid diffusion of the new standard. The likely antitrust issue is that the sponsor will subsequently 'close' the standard, either by denying its rivals quality access to necessary interface information or by engineering incompatibilities with their products[ 72 ].

Under the EU regime, the proportionality principle (mentioned above) represents a means of addressing such a situation. In 1984, for example, IBM ended a four-year investigation by the European Commission by agreeing to disclose in good time sufficient interface information to enable its competitors adapt their hardware and software to new IBM products[ 73 ]. IBM's control of the industry standard placed it in a complementary relationship with its rivals, and so it had to have regard to those rivals' interests when making changes to the standard.

3.4. Caring, Sharing Dominant Firms

Returning to the incompatibility strategies mentioned above, it should be noted that both types of unilateral conduct are possible because the disadvantaged firms are in a complementary relationship with the firm controlling the standard, yet also in competition with that firm in the downstream markets in which they operate. For example, developers of word-processing software require access to Windows interface technology if they are to produce Windows-compatible products, yet they also compete with Microsoft in the market for word-processing software[ 74 ]. Two cases support the proposition that a firm with monopoly power violates section 2 of the Sherman Act if it excludes rivals from the monopolised market 'by restricting a complementary or collaborative relationship without an adequate business justification'[ 75 ] - Aspen Ski and Kodak . This principal can be applied to computer software markets.

Between them, the parties in Aspen Skiing v Aspen Highlands Skiing [ 76 ] controlled the four downhill skiing mountains in Aspen, Colorado; Aspen Ski controlled three of the four mountains. The firms were of course rivals, yet for years they had offered skiers a six-day 'all-Aspen' ticket, dividing the revenues according to usage. In 1978, after Highland rejected Aspen Ski's offer of a fixed percentage of revenues considerably below its historical average, the collaborative relationship was terminated. Aspen Ski marketed a multi-area weekly ticket, limited to its three mountains. Highland's share of the market declined steadily over the next four years to about one-half of its previous level. As there was no apparent efficiency justification, Highland succeeded in its antitrust action. The Supreme Court wrote:

Eastman Kodak v Image Technical Services [ 78 ] concerned Kodak practices relating to parts and service for its photocopiers and micrographic equipment: essentially, Kodak discontinued its policy of selling spare parts to independent service operators (ISOs). Like Aspen Ski , while Kodak competed with independent ISOs in the service market, it also supplied them with the necessary parts. The plaintiff ISO successfully argued that this policy change raised its costs and allowed Kodak to monopolise the provision of service: Kodak was ordered to sell parts to ISOs at non-discriminatory prices. Note too, that the Supreme Court, denying summary judgement for Kodak, accepted that significant information and switching costs weakened the linkage between the markets for service and parts and the (competitive) equipment market, allowing Kodak to exercise market power in the downstream markets[ 79 ].

Some points arise from the discussion of Aspen Ski and Kodak :

• There must be a pre-existing relationship between the dominant firm and its competitor(s). Note that Posner J has interpreted Aspen Ski as meaning that a monopolist may violate section 2 'if it refuses to cooperate with a competitor in circumstances where some cooperation is indispensable' to effective competition[ 80 ]. So, the duty arises where effective competition requires some cooperation among competitors.
   

• Each case involved a policy change resulting in harm to the dominant firm's rivals (but not necessarily the exit of its rivals).

The Intel litigation demonstrates the application of these principles. Intel customarily supplied 'strategic' OEMs with advance technical information and samples of prototype central processing units (CPUs) for the purpose of building Intel-compatible computers. Digital, Compaq and Intergraph were three such OEMs, and they each had patents on certain CPU technologies (although only Digital actually competed with Intel in the CPU market). Effectively, they asserted these patents against Intel; Intel responded by cutting off the supply of advance technical information and prototypes in an attempt to force them to licence their patents on favourable terms[ 81 ]. An FTC investigation resulted in a consent decree prohibiting Intel from withholding or threatening to withhold certain advanced technical information from a customer for reasons relating to an intellectual property dispute with that customer[ 82 ]. Note that the OEMs were long-term Intel customers who relied on the advance technical information and product samples to design their products:

3.5. Developing A Compatibility Culture

A 'dynamic compatibility regime' involves strong intellectual property protection while a new standard is developed and advanced, with limits imposed on that protection once the standard has become established in the market[ 84 ]. The foregoing sections considered an aspect of this: the appropriate antitrust response to an anti-competitive change of policy designed to reduce interoperability and thereby disadvantage firms in downstream markets. Certain duties were imposed on the dominant firm as a result of its control of the standard technology, access to which was necessary to produce interoperable products. This section will consider a more far-reaching question: is there a means of compelling access to copyright-protected interface information, thereby encouraging ongoing innovation within a standard and avoiding entirely the possibility of technological input foreclosure?[ 85 ]

In theory at least, the copyright fair use doctrine allows disadvantaged firms to obtain access to necessary interface information: the doctrine permits the reverse engineering of software. In Nintendo [ 86 ] and Sega [ 87 ], the reverse engineering of consoles and cartridges so as to discover interface information necessary to develop compatible games was upheld as fair use. The EC Software Directive permits reverse engineering where it is 'indispensable to obtain the information necessary to achieve the interoperability of an independently created program with other programs'[ 88 ]. This recognition that copyright protection of functional requirements governing compatibility should not confer 'disproportionate leverage' into related markets involves an acceptance that broad copyright protection of software is not necessary to provide the appropriate innovation incentives[ 89 ]. In reality, however, the reverse engineering of an OS such as Windows would not be practicable[ 90 ].

The copyright misuse doctrine represents another means of obtaining access to interface information[ 91 ]. Put simply, this doctrine allows a defence to an infringement action if the copyright is used in a manner contrary to the public policy embodied in the copyright:

The misuse doctrine is broader than fair use in that it can secure access to all that was previously available, not just to the functionality underlying the now-protected material. However, 'it is on such uncertain legal grounds that courts may be reluctant to apply it'[ 93 ]. Ultimately, it has to be recognised that the doctrines of fair use and misuse are only available as defences to infringement actions: they operate to prevent the anti-competitive enforcement of the intellectual property rights in a standard, which will often amount to a change of policy falling under the rule in Aspen Ski and Kodak (discussed above)[ 94 ].

Patterson has suggested a more general approach, arguing that the doctrine of estoppel may imply a 'copyleft'.[ 95 ] Copyleft software is software that users are free to use, modify and distribute on the condition that the source code remains open; a copyleft licence requires the user to agree not to assert copyright in respect of any changes or improvements he makes, to disclose the entire source code for those changes, and to disseminate those changes subject to another copyleft licence. However, insofar as the disclosure is not limited to the information necessary for the development of interoperable (i.e. downstream) products, this approach would fail to preserve sufficient innovation incentives; more practically, it binds the licensee, not the licensor. So, Patterson proposes a modified, or reverse, copyleft: a firm is obliged to keep the specifications open as a condition of its standard being accepted by consumers who have relied on its open-source manifestations. While this approach is preferable to the copyright defences discussed above in that it involves the imposition of a positive obligation on dominant firms, it is again limited to policy changes, which can be adequately addressed under the rule in Aspen Ski and Kodak . Something more far-reaching is required.

3.6. Re-Thinking Copyright

Computer software enjoys copyright protection as a literary work. The argument for reconsidering the appropriate level of copyright protection for software flows from the particular features that distinguish computer programs from other literary works[ 96 ].

The expression in software is 'hidden': executable files depend on object code to operate, not human-readable source code. So software users are not afforded any meaningful access to a literary work as such. Indeed, the utility that a user derives from a computer program comes from its functionality, and not from any appreciation of the protected expression. Stallman has identified another distinguishing feature: the ease and desirability of modifying or customising software, which is arguably 'one of its great advantages over older technology'[ 97 ].

A central element of the 'digital' or 'networked' economy is the decentralisation of the power to manipulate, copy and redistribute information. However, copyright's inherent trade-off between the interests of authors and publishers and those of society was struck long ago, at a time when individuals were not capable of copying. Consider a modern example, the US Copyright Act of 1976: the major new technology at that time was the photocopier (a means of centralised copying) and computers were only owned by large organisations. It has been argued that as a result of technological change, society's freedom to copy, modify and redistribute works is now something of real value:

This would suggest that a re-evaluation of copyright's balance between the public and private interests in literary works is appropriate, but modern legislation has tended to shift the balance further in favour of copyright holders[ 99 ]. However, the particular features of computer programs could support an argument for qualifying the scope of copyright protection for software. For example, Stallman has proposed a three-year copyright term for computer programs, and that protection would be conditional on the deposit of the source code with some designated public body[ 100 ]. Whether a three-year copyright term would sufficiently preserve the innovation incentives is outside the scope of this paper, but the author supports the principal of a significantly shorter term for computer software than currently applies.

Stallman also argues that the nature of software as a functional work requires that users be free to publish modified versions of programs: however, this would practically eliminate innovation incentives. It is submitted that a limited freedom to develop interoperable (downstream) programs would be more appropriate. Copyright protection of software standards would be subject to a positive obligation to allow access to the technical information to the extent necessary for the development by rivals in downstream markets of interoperable products. And on the expiration of the shorter term, the entire source code would be freely available: this would allow the development of competing products, i.e. compatible offerings in the primary (upstream) market[ 101 ].

It will of course be argued that the dynamic compatibility regime envisaged above would fail to preserve sufficient innovation incentives, but despite its superficial appeal this argument should not be unhesitatingly accepted[ 102 ]. In network markets, the natural lead-time that an innovator will enjoy is transformed into a significant competitive advantage, so network effects offer considerable protection to the leading firm. It is at least arguable that a shorter copyright term would encourage drastic innovation over the incremental development of existing standards. And while innovation incentives may be reduced initially, there would be wider dissemination of the technical information necessary for the development of compatible products: this would undoubtedly spur innovation within and around an established standard. Farrell and Katz have cautioned that compatibility may under-reward 'drastic innovation'[ 103 ]. It is submitted that this ignores the fact that drastic innovation will introduce incompatibility, albeit subject to a shorter term of copyright protection (for the primary software product or standard) and an obligation to allow access to rivals to the extent necessary to develop interoperable products.

Ultimately, it should be recognised that the argument that innovation incentives are higher under incompatibility is not an argument against compatibility: rather, it is an unqualified argument in favour of monopoly. A dynamic compatibility regime does not question the link between incompatibility and innovation incentives, but it does introduce some qualifications.

4. Conclusion

This article argued for, and considered the viability of, a 'dynamic compatibility regime' as a model for competition policy within computer software markets. Such an approach recognises that intellectual property rights offer the innovation incentives that are critical to technical progress, the key driver of consumer benefits in software markets. However, once a software standard has succeeded on the market, the economic features of the software industry dictate a shift to a compatibility regime. Quality access to the information necessary to produce interoperable products will be indispensable for effective, ongoing competition.

Appropriate analytical tools exist to address a change in policy that renders previously interoperable products incompatible with the existing standard. But a true compatibility regime would recognise that access to the information necessary for interoperability is a pre-requisite to effective competition in the markets for interoperable software products, i.e. that there is a positive obligation on dominant firms to allow rivals access to interface information to the extent necessary to develop interoperable products. An argument in favour of qualifying the term and scope of copyright protection for software was constructed, and it is to be hoped that further discussion will follow. Ultimately, what is required is legislative intervention on an international level - yet it must be accepted that this is highly unlikely.

Footnotes

1 . Bork, The Antitrust Paradox (1993) at 50-89

2 . Bork, op. cit., at 90

3 . Bork, op. cit., at 91

4 . Shapiro, 'Antitrust in Network Industries', (January 25, 1996), Address before the American Law Institute and American Bar Assn., available at: < http://www.usdoj.gov/atr/public/speeches/shapir.mar.htm >.

5 . Bresnahan, 'New Modes of Competition: Implications for the Future Structure of the Computer Industry', in Eisenach & Lenard ed., Competition, Innovation and the Microsoft Monopoly: Antitrust in the Digital Marketplace (1999) 155, at 156. Bresnahan's paper contains an excellent discussion on software industry structure and its implications for competition.

6 . These are (obviously) industries characterised by network effects. It is important to distinguish natural monopolies, which arise from supply-side economies of scale and which may therefore warrant analysis under the essential facilities doctrine.

7 . See Balto, 'Networks and Exclusivity: Antitrust Analysis to Promote Network Competition', (1999) 7 George Mason Law Rev 523, at 560.

8 . Melamed, 'Network Industries and Antitrust', (April 10, 1999), Address before The Federalist Society, available at: < http://www.usdoj.gov/atr/public/speeches/2428.htm >.
See also, US v Microsoft, Findings of Fact , 84 F. Supp 2d 9 (Dist DC, 1999), at para. 19, 21, 25 (sunk costs) and para. 20 (switching costs).

9 . An operating system (OS) such as Windows controls the allocation of system resources and supports the functions of compatible applications by exposing 'application programming interfaces' (APIs). An application that relies on OS-specific APIs will not function on another OS unless it is 'ported' to the APIs of that OS: Findings of Fact , loc. cit., at para. 2, 4.

10 . Findings of Fact , loc. cit., at para. 38

11 . 'What for Microsoft is a positive feedback loop is for would-be competitors a vicious cycle'. Jackson J, Findings of Fact , loc. cit., at para. 40. See Cass and Hylton, 'Preserving Monopoly: Economic Analysis, Legal Standards and Microsoft ', (1999) 8 George Mason Law Rev 1, at 36-37: 'while network effects imply that the probability of successful entry is lower than it would be otherwise, they also imply that the payoff from successful entry is larger than it would be otherwise'.

12 . Findings of Fact , loc. cit. , at para. 59-60. An 'inflection point' is the same as Bresnahan's 'epoch' (above).

13 . Whether competing standards can ultimately co-exist depends on the strength of network effects. See Liebowitz and Margolis, 'Networks and Standards', Winners, Losers and Microsoft: Competition and Antitrust in High Techonology (1999) 87-115. Also, Rubinfeld, 'Competition, Innovation, and Antitrust Enforcement in Dynamic Network Industries', (March 24, 1998), Address before the Software Publishers Assn., available at: < http://www.usdoj.gov/atr/public/speeches/1611.htm >.

14 . Melamed, loc. cit. Shapiro, loc. cit., comments that 'it is notoriously difficult for new programs to provide sufficiently great improvements in performance to justify the switching costs users would have to incur to adopt them.'

15 . Lemley and McGowan, 'Legal Implications of Network Economic Effects', (1998) 86 California Law Rev 479, at 506

16 . Katz and Shapiro, 'Antitrust in Software Markets', in Eisenach and Lenard ed., op. cit., at 33. See also, Economides, 'Competition, Compatibility, and Vertical Integration in the Computer Industry', in Eisenach and Lenard ed., op. cit., 209, at 210: '[The] benefits of complementarity can be realized through standardization and interoperability among components.' For a discussion on 'the role of network effects and incompatible products in creating a [software] monopoly' see Salop and Romaine, 'Preserving Monopoly: Economic Analysis, Legal Standards and Microsoft ', (1999) 7 George Mason Law Rev 617, at 620-622.

17 . Farrell and Katz, 'The Effects of Antitrust and Intellectual Property law on Compatibility and Innovation', (Fall 1998) The Antitrust Bulletin 609, at 611.

18 . See Katz and Shapiro, in Eisenach and Lenard ed., op. cit., at 45-54.

19 . Arguably, the imposition of a disclosure remedy would be more effective than a break-up of Microsoft. See, for example, Economides, 'US v MS and the Future of the US Computing Industry' (May 5, 2000), available at: < http://www.stern.nyu.edu/networks/usvms.html >, and Stallman, 'The Microsoft Antitrust Trial and Free Software' (March 1999), available at: < http://www.gnu.org/philosophy/microsoft-antitrust.html >. Microsoft discloses the APIs that hook applications to Windows, but not the APIs that tie together the component parts of the Windows OS (which includes IE).

20 . United States v Microsoft Corporation : Findings of Fact, 84 F. Supp. 2d 9 (DDC 1999), available at: < http://www.microsoft.com/presspass/trial/c-fof/default.asp >; Conclusions of Law, 87 F. Supp. 2d 30 (DDC 2000), available at: < http://www.microsoft.com/presspass/trial/col/default.asp >; Final Judgment, 97 F. Supp. 2d 59 (DDC 2000), available at: < http://www.microsoft.com/presspass/trial/remedies/default.asp >; Microsoft's Appellate Brief, available at: < http://www.microsoft.com/presspass/trial/appeals/default.asp >.

21 . Findings of Fact , loc. cit. , at para. 29, 32, 73-76. Note, however, that Jackson J identified 'significant shortcomings' with the thin client model: 'The problems of latency, congestion, asynchrony, and insecurity across a communications network, and contention for limited processing and memory resources at the remote server, can all result in substantial degradation of computing performance' ( Ibid., at para. 26).

22 . On June 28, 2001, the Court of Appeals issued its opinion. It upheld Jackson J's finding that Microsoft employed anticompetitive means to maintain its OS monopoly, reversed the finding that Microsoft attempted to monopolise the browser market, and remanded for reconsideration under the rule of reason the finding that the integration of Windows and IE amounted to an illegal tie. The remedies decree was vacated: it now seems highly unlikely that Microsoft will be broken up. The full text of the opinion is available at: < http://www.microsoft.com/presspass/trial/appeals/06-28opinion.asp >.

23 . Jackson J, ibid. , at para. 137. At that time, Navigator's market share was approximately 80%.

24 . Note that Jackson J appears to acknowledge that a policy of charging for IE would only 'temporarily' enhance consumer demand for Windows ( Ibid., at para. 139).

25 .  Klein, 'Microsoft's Use of Zero-Price Bundling to Fight the 'Browser Wars'', in Eisenach & Lenard ed., op. cit., 217, at 223-225.

26 . Conclusions of Law, loc. cit.  Bork, op. cit., at 379, has said that '[t]here is no way to state the 'inherent' scope of a product. The judge who attempts it either decides according to product dimensions that seem to him natural because he is accustomed to them, or explicitly decides on grounds of efficiency.  An automobile and a can of pears are perceived as single products because it would be too expensive to require the seller to subdivide them further. Economies of scale determine the definition of product'. See also Lopatka and Page, 'Antitrust on Internet Time: Microsoft and the Law and Economics of Exclusion', (1999) 7 Univ Chicago Supreme Court Economic Rev 157, at 164: the 'focus on separate demand does not address the most important efficiency concerns in cases involving [integration]'.

27 . Microsoft's Appellate Brief, loc. cit., at 69.  The Court of Appeals seemed to favour this argument, loc. cit., at 76. It was accepted in Multistate Legal Studies v Harcourt Brace Jovanovich Legal and Professional Publications 63 F. 3d 1540 (1995 10 ^th Cir.) that product improvements may be the cause (or effect) of changes in demand: presumably, this would allow innovation by integration within the 'consumer demand' test where there was subsequent demand for the integrated product.

28 . Lopatka and Page, loc. cit., at 209.

29 . Findings of Fact , loc. cit., at para. 159. See also Lopatka and Page, loc. cit., at 213.

30 . Conclusions of Law, loc. cit.

31 . Findings of Fact , loc. cit. , at para. 173. This is achieved by placing IE-specific code through various OS-specific files: the Windows 98 interface is displayed using HTML; the 'Help' system and the 'Windows Update' feature also depend on IE-specific code.

32 . Ibid., at para. 174.

33 . Note also that by late 1998, Microsoft's IE had a market share of approximately 45-50%; Navigator's share had fallen from approximately 80% in 1996 to the 'mid 50% range' in July 1998 (AOL figures, Ibid , at para. 360).

34 . Ibid., at para. 205-208, 213-220. Noting that the restrictions imposed on OEMs by Microsoft were considerably greater than those imposed by Apple and IBM (rival OS vendors), Jackson J states that the reason for this is that Apple and IBM did not share Microsoft's anti-competitive motivations ( ibid , at para. 229). A more plausible explanation is that the difference merely reflects the relative bargaining strengths of the three OS vendors.

35 . Difficulties in 'finger-pointing' mean that users could associate a slow, over-elaborate start-up sequence with Microsoft, not with the OEM that devised it ( ibid., at para. 223-225). Bork argues: 'If consumers prefer the Netscape platform to the Windows platform, computer manufacturers will configure their machines so that the first screen to appear is that generated by Netscape's browser. If a significant number of computers had the Netscape first screen, software developers would write programs for it, and competition would flourish in the operating systems market. Microsoft's first screen restriction squelches that.' 'The Case Against Microsoft', available at: < http://www.procompetition.org/research/bork.html >. This is nothing short of ridiculous: Bork ignores the fact that OEMs were free to feature Navigator on the Windows desktop and to make it the default browser; he also fails to appreciate that the Navigator-Java platform was not a viable standard.

36 . Findings of Fact , loc. cit , at para. 215, 231.

37 . Ibid , at para. 216.

38 . Ibid , at para. 217.

39 . Klein, loc. cit., at 239-245.

40 . See Findings of Fact , loc. cit. , at para. 232-235, 240. In his prepared testimony before the Senate Committee on the Judiciary Subject: 'Competition, Innovation, and Public Policy', (November 4, 1997), Charles Rule observed that 'it appears that the OEMs who allegedly wanted to remove IE from the desktop wanted to do so in order to give the Navigator icon an 'exclusive' on the desktop'. He continued: 'This is the first time I can recall the Department of Justice taking an enforcement action against one company with a smaller share (in Microsoft's case, less than 40 percent) in order to protect the ability of a dominant competitor (in this case Netscape with more than a 60 percent share) to secure an exclusive'.

41 . Lopatka and Page, loc. cit., at 229 describe the OEM-restrictions as 'a struggle having to do with the distribution of wealth, not efficiency'. Klein, loc. cit., at 241-242, also notes the likely negative precedential effect of allowing an OEM to modify the OS in advance of licence renegotiations: the Windows licences were short-term agreements.

42 . Klein, loc. cit., at 237-239, has argued that Netscape's market share did not decrease significantly when Microsoft began bundling its browser in August 1995: by April 1996 its market share had risen to only 4%. And in August 1996, three months after the introduction of a much-improved version of Internet Explorer, Microsoft's market share was still only 8%. From that point, however, it increased significantly, i.e. Netscape's share substantially declined only after Microsoft had made massive R&D investment and produced a superior product. So, from 1995 to early 1996, the bundling worked to prevent Netscape from foreclosing Microsoft's browser from OEM desktops.

43 . Findings of Fact , loc. cit , at para. 240.

44 . Ibid , at para. 242-310. The Court of Appeals opinion , loc. cit., seems to largely overlook the various forms of compensation accepted by the OEMs and ISPs. Presumably, this conditions the Court's treatment of the related restrictions, i.e. its analysis flows from the assumption that the restrictions were unilaterally imposed by Microsoft. This is unfortunate, because it is at least arguable that if the Court had recognised the OEM-ISP compensation, it would have reached a different conclusion.  Consider the Court's comments regarding the pricing of IE and the IE Access Kit (IEAK): 'we...have no warrant to condemn Microsoft for offering IE or the IEAK free of charge or even at a negative price', loc. cit., at 42.

45 . Appellate Brief, loc. cit., at 32

46 . Lopatka and Page, loc. cit., at 222

47 . Findings of Fact , loc. cit., at para. 294. Whereas Netscape refused to allow its ISP licensees to move Navigator's default home page from Netscape's own portal, Microsoft allowed each ISP to pre-set the default home page ( ibid , at para. 248-249).  More generally, note that with 26 million subscribers, AOL effectively controls one-third of IE users, and has been described as 'the de facto Internet operating system for mass-market America': Fortune , 'AOL's Grand Unified Theory of the Media Cosmos', (January 8, 2001) 32-38, at 37. See also, Lopatka and Page, loc. cit., at 161 (footnotes 9, 10 and associated text). Note that partnership talks between AOL and Microsoft recently broke down. The parties had sought to address issues such as Microsoft's inclusion of AOL software in the Windows XP operating system, AOL's use of IE as its default browser, and AOL's possible support of Microsoft's Windows Media software: 'Microsoft-AOL Talks Collapse in Acrimony; Firms Point Fingers', Wall Street Journal (June 18, 2001), available at: < http://www.msnbc.com/news/588988.asp >.

48 . Findings of Fact , loc. cit , at para. 217.

49 . Katz and Shapiro, 'Antitrust in Software Markets', in Eisenach and Lenard ed., op. cit., at 75.

50 . Krattenmaker and Salop, 'Anticompetitive Exclusion: Raising Rivals' Costs to Achieve Power Over Price', (1986) 96 Yale Law Journal 209, at 214. See also, Salop and Scheffman, 'Raising Rivals' Costs', (1983) 73 American Economic Rev 267.

51 . Krattenmaker and Salop, loc. cit., at 233-253. See also, Krattenmaker and Salop, 'Competition and Cooperation in the Market for Exclusive Rights', (1986) 76 American Economic Rev 109, at 109.

52 . This is so because third parties who are not involved in the bidding, i.e. consumers, will receive many of the benefits from the non-exclusion of rivals.

53 . The Findings of Fact suggest a number of instances of technological input foreclosure. It seems that Microsoft withheld a critical API from Netscape, forcing it to postpone the release of its Windows 95 browser until 'substantially after' the release of Windows 95 ( loc. cit , at para. 90-92). And, in response to IBM aggressively marketing its own software line as alternatives to Microsoft products, it withheld the Windows 95 'golden master' code (needed for product planning and development), offering IBM early access to the source code only if it stopped pre-installing IBM software on its PCs ( ibid , at para. 115-132). Also, application developers were induced to make their Web-based applications reliant on IE-specific technology ( ibid., at para. 337-339) and to adopt Microsoft's proprietary version of Java ( ibid., at para. 386-407) in exchange for early access to necessary interface information.

54 . Internal Microsoft correspondence, Findings of Fact , loc. cit., at para. 160. The Findings of Fact offer other examples. Microsoft made its proprietary version of Java incompatible with Sun's cross-platform version ( ibid , at para. 386-407). It also appears to have required Internet Content Providers (ICPs) to provide 'differentiated content' that would be more attractive when viewed with IE, with 'acceptable degradation' when viewed with other browsers, or which would be available only to IE users ( ibid , at para. 322).

55 . Salop and Romaine, loc. cit., at 634-635. A dominant firm in a network market will favour a strategy of incompatibility: Katz and Shapiro, 'Network Externalities, Competition and Compatibility', (1985) 75 American Economic Rev 424, at 425.

56 . Easterbrook, 'Predatory Strategies and Counterstrategies', (1981) 48 Univ Chicago Law Rev 264, at 310. The tie-in is properly 'subject to erosion as a result of competitors' responses', i.e. once the system has acquired a reputation for quality the need for joint purchases would lapse. Easterbrook comments that all innovations involve some short-term sacrifice (the amount invested in R&D) and all innovators hope to recoup this investment, perhaps even by becoming monopolists if they can. He notes that 'the most desirable innovations would seem to be the most predatory, for R&D costs (the 'sacrifice') and market share (the result) both may be high' ( ibid ., at 309).

57 . Sidak, 'Debunking Predatory Innovation', (1983) 83 Columbia Law Rev 1121, at 1140-1141.

58 . 703 F. 2d 534 (9 ^th Cir. 1983).

59 . loc. cit., at 542.

60 . 613 F. 2d 727 (9 ^th Cir. 1979).

61 . loc. cit., at 743-744.

62 . See also Memorex v IBM 636 F. 2d 1188 (9 ^th Cir. 1980) and Transamerica v IBM 698 F. 2d 1377 (9 ^th Cir. 1983).

63 . 147 F. 3d 935 (D.C. Cir. 1998).

64 . loc. cit., at 948-950.

65 . See Cass and Hylton, loc. cit , at 34: 'Courts adopt bright-line rules because they know that the alternative balancing test is likely to be applied so inaccurately that adopting a bright-line rule, even though it may favour one party, minimizes total error costs'. See also Lopatka and Page, loc. cit , at 198-207.

66 . Temple Lang, 'EC Antitrust Law - Innovation Markets and High-Technology Industries', (1996 Fordham Corporate Law Institute), International Antitrust Law and Policy (1997) 519, at 562.

67 . Ibid , at 563. So, integrating functionality into an operating system probably amounts to technical progress if it makes the OS 'easier to use, or more functional for all users' even if there is no 'net plus' achieved through the integration.

68 . 603 F. 2d 263 (2d. Cir. 1979), at 279/

69 . Kodak had launched a new line of smaller Instamatic cameras which required a new film format; for a period of 18 months after the launch of the new camera, this new film format was made available only with the new camera. The Circuit Court seemed to see the development of the new film format as an integral part of the development of the new camera 'system', loc. cit., at 282.

70 . Ibid , at 281.

71 . Ibid , at 283.

72 . Katz and Shapiro (in Eisenach and Lenard ed., op. cit , at 65) favour limiting the ability of dominant firms to change their policies by shutting down interfaces that had been open, referring to Aspen Ski and the EU Commission's IBM investigation (see below).

73 . It was recognised that disclosure would be inappropriate if it disclosed new (non-interface) features of IBM's designs. For a description of the IBM investigation, see Temple Lang, 'Defining Legitimate Competition: Companies' Duties to Supply Competitors, and Access to Essential Facilities', (1994 Fordham Corporate Law Institute), International Antitrust Law and Policy (1995) 245, at 303, and 'Pssst! Secrets May Get Out: EU Commission Asks Whether Microsoft Needs to Share Data With Rivals', Wall Street Journal (November 20, 2000) 27.

74 . See Wagner, 'The Keepers of the Gates: Intellectual Property, Antitrust and the Regulatory Implications of Systems Technology', (2000) 51 Hastings Law Journal 1073, at 1101: 'The nature of systems technology...enables its designers easily to exclude others from the technology, so that those seeking access can obtain it only if the designers themselves share their intellectual property with them.' Insofar as exclusion will often enhance the designers' return on investment, there is little incentive to allow access.

75 . Baker, 'Promoting Innovation Competition Through the Aspen/Kodak Rule', (November 12 1998), Address before The George Mason University Law Review Antitrust Symposium, available at: < http://www.ftc.gov/speeches/other/mason1098.htm >.

76 . 472 US 585 (1985). Bork has relied on Aspen Ski as authority for the proposition that 'a monopolist is not free to define its product for the purpose and with the effect of excluding a competitor'. (Bork, 'The Case Against Microsoft', loc cit.) It should be clear from the analysis that follows that his interpretation is, at best, strained.

77 . 474 US 585, at 603.

78 . 504 US 451 (1992)

79 . loc. cit , at 473, 476.

80 . Olympia Equipment Leasing v Western Union Telegraph 797 F. 2d 370 (7 ^th Cir. 1986), at 379. Note also that the Court of Appeals in Multistate Legal Studies , loc. cit., rejected as too narrow the defendant's interpretation of Aspen Ski as authority only for the proposition that the duty arose only where the monopolist had accommodated its competitor in the past.

81 . Digital and Compaq settled their claims and licensed their patents to Intel, whereon the cooperative supply practices resumed. Intergraph instituted a private antitrust claim, which failed: Intergraph v Intel 195 F. 3d 1346 (Fed Cir. 1999). It was held that as Intel did not operate downstream in the OEM market, and Intergraph did not operate in the CPU market, there was no evidence that Intel had secured or exploited an unfair advantage for itself, or made either market as a whole generally less competitive. The DOJ's theory of liability in Microsoft could meet a similar fate: see, infra , footnote 22.

82 . See Valentine, 'Abuse of Dominance in Relation to Intellectual Property: U.S. Perspectives and the Intel Case', (November 15 1999), available at: < http://www.ftc.gov/speeches/other/dvisraelin.htm >.
See also: < http://www.ftc.gov/os/1999/9903/d09288intelanalysis.htm >.

83 . Anthony, 'Antitrust and Intellectual Property Law: From Adversaries to Partners' (2000), available at:< http://www.ftc.gov/speeches/other/aipla.htm >. See, however, CSU v Xerox 203 F. 3d 1322 (Fed Cir. 2000). The facts are similar to those of Kodak . Xerox (which manufactures, sells and services high-volume photocopiers) introduced a restrictive parts policy in 1984 and subsequently expanded its scope and tightened its enforcement. The result was that by 1989, ISOs could not purchase Xerox parts directly. However, under a 1994 antitrust settlement, Xerox agreed to suspend the policy. CSU, an ISO, opted out of the settlement and filed suit, alleging that Xerox had violated section 2 by setting prices on its patented products much higher for ISOs than for end-users, thereby raising the costs of its rivals in the service market. The Xerox court rejected the ISO arguments and distinguished Kodak on the grounds that no patents had been asserted in defence of the antitrust claims when that case came to the Supreme Court; it also characterised Kodak as a tying case. This ignores the identity of Xerox and Kodak's collaborative relationships with the ISOs: IP rights may qualify the obligation flowing from that relationship (by giving rise to a presumption in favour of the right-holder) but they should not preclude it.

84 . See O'Rourke, 'Toward a Doctrine of Fair Use in Patent Law', (2000) 100 Columbia Law Rev. 1177, at 1230-1235, and Patterson, 'Copyright Misuse and Modified Copyleft: New Solutions to the Challenges of Internet Standardisation', (2000) 98 Michigan Law Rev 1351, at 1352.

85 . See, infra , footnote 19 and associated text.

86 . Atari Games v Nintendo 975 F. 2d 832 (Fed. Cir. 1992).

87 . Sega Enterprises v Accolade 977 F. 2d 1520 (9 ^th Cir. 1992).

88 . Article 6(1), Directive 91/250/EEC on the Legal Protection of Computer Software.

89 . See Sony Computer Entertainment v Connectix 203 F. 2d 595 (9 ^th Cir. 2000) at 605: 'If Sony wishes to obtain a lawful monopoly on the functional concepts of its software it must satisfy the more stringent standards of the patent law.' It seems that the Directive 91/250/EEC allows for the reverse engineering of both compatible (i.e. complementary) and competing products: Guillou, 'The Reverse Engineering of Computer Software in Europe and the United States: A Comparative Approach', (1998) 22 Columbia VLA Journal of Law and the Arts 533, at 543. In Sony v Connectix the reverse engineering of the PlayStation system was upheld as fair use despite the fact that the end product was a direct competitor of the PlayStation, suggesting an identity of approach under both regimes.

90 . Note also that software developers have begun to acquire patent protection for APIs (Lemley and McGowan, loc. cit , at footnote 212). In response to this shift, O'Rourke has argued for a patent fair use defence, loc. cit , at 1230-1235.

91 . Patterson, loc. cit , at 1371-77. Also, Frischmann and Moylan, 'The Evolving Common Law Doctrine of Copyright Misuse: A Unified Theory and its Application to Software', (2000), 15 Berkeley Technology Law Journal 865.

92 . Patterson, loc. cit , at 1373. See, for example, Alcatel USA v DGI Technologies 166 F. 3d 772 (5 ^th Cir. 1999). Alcatel produced telephone switching equipment which was controlled by its copyrighted operating system software. DGI was formed to produce compatible expansion cards, which would allow Alcatel customers additional capacity. In downloading the OS software in order to test its product, DGI infringed Alcatel's copyright. However, the court found that Alcatel had misused its copyright: its anti-competitive licensing provisions prevented competitors from developing interoperable products. It is also interesting to note that DGI contended that Alcatel inserted software 'patches' designed to render its card inoperable, loc. cit., at 778.

93 . Patterson, loc. cit., at 1377.

94 . Dicta in Bateman v Mnemonics 79 F. 3d 1532 (11 ^th Cir. 1996) could support a more general approach whereby requirements of compatibility would qualify the scope of software copyright protection. 'Whether the [copyright] protection is unavailable because these [external] factors render the expression unoriginal, non-expressive...or whether these factors compel a finding of fair use, copyright estoppel, or misuse, the result is to deny copyright protection to portions of the computer program. Thus, we today join these other circuits in finding that external considerations such as compatibility may negate a finding of infringement .' (Emphasis added.) However, these comments were made in the context of a copyright infringement case, and as such cannot be taken to support anything more than a defence.

95 . Patterson, loc. cit , at 1377-1382.

96 . See Frischmann and Moylan, loc. cit , at 904-919.

97 . Stallman, 'Why Software Should Be Free', in Moore ed., Intellectual Property: Moral, Legal and International Dilemmas (1997) 283, at 289.

98 .  Stallman, 'Copyright versus Community in the Age of Computer Networks' (July 2000), available at: < http://www.carnall.demon.co.uk/stallman/index.html >.

99 . Laddie J, 'Copyright: Over-strength, over-regulated, over-rated?' [1996] 5 EIPR 253, at 259, describes the current copyright system as being built on 'a foundation of accumulated rights, commercial interests, and monetary expectations'. See also Bauchner, 'Technology and Copyright Corruption' (June 13, 2001), available at: < http://news.cnet.com/news/0-1276-210-6269374-1.html >.

100 . Stallman, loc. cit. See also Boudin J, Lotus Development v Borland International 49 F. 3d 807 (1 ^st Cir. 1995): 'Some solutions (e.g., a very short copyright period for menus [i.e. functional aspects of computer programs]) are not options at all for courts but might be for Congress'.

101 . Stallman, 'Why Software Should Be Free', loc. cit. , at 290-295, has identified a number of other advantages that would follow from limiting the scope of copyright protection for software: the elimination of duplicative effort (or, expressed another way, lower development costs), the freedom to customise software which in turn allows for ongoing, evolutionary development, and better education of future developers. See also Wagner, loc. cit at 1109-1111.

102 . Cornish, op. cit., at 368-369: 'To shorten the present copyright period is...unlikely to produce any noticeable effect upon the amount of copyright material which [publishers, record producers or film makers] are prepared to put out for consumption.' He concludes that the argument for greater copyright protection has succeeded, not on the basis of revisions of the appropriate economic incentives, but out of a 'special admiration for aesthetic creativity'. This would support a shorter term for computer software, which is (at least largely) functional.

103 . Farrell and Katz, loc. cit , at 647.