TCS Daily

Open Source and Its Enemies

By Julian Sanchez - December 10, 2002 12:00 AM

It should surprise nobody to learn that Bill Gates is not a fan of Linux or open source software [OSS], for roughly the same reason that light bulb manufacturers might resent sunny days. What ought to be surprising is that many libertarian writers have recently launched a series of vehement attacks against OSS, and especially against proposals to preference open software in government procurement. The most hysterical go so far as to characterize the Open Source movement as socialist, though it is difficult to think of a less apt description of the dynamic, decentralized, and voluntary process by which OSS is developed.

There are good reasons not to require state agencies to use OSS under any and all circumstances. Government, after all, is not supposed to be a software company, and it would be frivolous to waste tax dollars writing an open source program for which a reasonably priced proprietary alternative exists already. What, though, is the problem with a policy of "affirmative action" for OSS, as we might call it -- the sort of policy being considered in much of Europe? James V. DeLong of the Competitive Enterprise Institute says that governments "should keep their hands off" of software markets, while the Pacific Research Institute's Sonia Arrison writes that procurement decisions must be made exclusively on the basis of cost and efficiency, rather than for "purely political reasons." These are sound principles on face, but it's not clear precisely what a "hands off" policy means in a context where government is necessarily acting as a player in the market, nor how efficiency and cost are best factored into procurement decisions, nor, indeed, whether "political reasons" of a very specific kind shouldn't shape the decision making process.

Let's begin with strictly technical considerations. DeLong argues that OSS is unsuitable for government use, because only in proprietary markets, where "good old reliable greed creates an incentive for developers" will we see a sufficient variety of applications developed. But as free software guru Richard Stallman has put it: "We do develop a lot of free software. If a theory says we can't, you have to look for the flaws in the theory." NYU law professor Yochai Benkler has suggested that traditional economic models fail to properly account for the distributed, modular, and granular nature of peer production processes. OSS is coded and debugged, not by a small stable of employees, but by a dispersed network of hundreds or thousands of self-selecting coders, each working on a small piece of the puzzle. Depending on the size of the chunk tackled by each, they may be motivated by such diverse incentives as "good old reliable greed," the intrinsic pleasure of the task, their own need (either as a private user or a tech support guru) for a bug-fix, and the desire for a reputation boost. But of course, should these prove insufficient, government may well opt for a proprietary solution as a last resort.

What technical reasons, then, support a systemic bias in favor of OSS? Some of the familiar advantages of open software, such as its typically lower cost or greater stability, can be at least roughly gauged on an ad hoc, case-by-case basis. But many others are harder to capture in any static analysis. Because OSS code is available to a large user community, unforeseen glitches in the software tend to be spotted and, more importantly, fixed far more quickly than in the case of proprietary software, where only an elect cadre of high priests can dig into the program's innards. The proprietary model is also more likely to be hobbled by the "innovator's dilemma," made famous by the Harvard Business School's Clayton Christensen. Christensen's thesis, in a nutshell, is that market incumbents may quite rationally resist revolutionary or "disruptive" technologies inconsistent with their business models, leading in them in the long run to fall to smaller, less risk-averse competitors. In software markets, the network benefits accruing to software characterized by "bandwagon effects" (about which more later) coupled with the exclusive power of incumbent firms (or their licensees) to use and build upon their own dominant code make it easier to forestall disruptive innovations for far longer than would be possible with OSS.

In truth, however, the best reasons to prefer OSS are political reasons. Both liberal principles of neutrality and public choice considerations weigh strongly in favor of adopting OSS when that's feasible.

In a better world, procurement decisions would be isolated from the lobbying and political maneuvering that give us such grand contributions to the general welfare as mohair subsidies and steel tariffs. We know, alas, that this is not the case. Mere months ago, Oracle made headlines by securing a suspiciously favorable contract with California, no competitive bidding required. Oracle had, just by coincidence, made ample contributions to the reelection campaigns of Gov. Gray Davis and state Attorney General Bill Lockyer. Companies like juicy government contracts, however obtained; there's probably not much to be done about that. A policy of open procurement, however, at least reduces the incentive for any one firm to exert political pressure.

Proprietary software makers know that client data is locked up in a format they own. This places them in a unique position to provide upgrades, fixes, and other forms of technical support - especially when dealing with inertia bound bureaucracies less subject to the pressures that might make a private firm switch platforms more readily. They also know that software, like a VCR or fax machine, is often a "network good" characterized by "bandwagon effects," which make its value a function, not only of its intrinsic characteristics, but also of the number of other people using the same product. These facts taken together mean that firms can parlay government use of a proprietary format - PowerPoint, say - into sales of the client software to read it that format. Even when the client software is given away without charge, as with Adobe Acrobat Reader, firms know that if more users need to download their proprietary client in order to communicate with the government, that larger user base expands the market for their authoring software. Since companies can't expect to similarly capture those network benefits when producing open source software - and, perhaps more importantly, needn't fear being locked out by a competitor who does - the stakes are far lower for any one contract. With proprietary software, government's potentially standard-setting procurement choices give it the role of market kingmaker.

The Initiative for Software Choice, a lobbying group formed by software industry incumbents to oppose the threat posed by OSS to their market share, claims to be all for interoperability and "open" standards. But the devil is in the details, and ISC's proposals are in actuality a perfect illustration of rent seeking through procurement. Government should ensure interoperability, they suggest, via standards "available to all through reasonable and non-discriminatory licensing." In other words, the state should adopt standards that embed royalty-generating patents, such that software makers must license the right to build compatible programs using that technology. The effect is to ensure that commercial software, produced by a centralized firm that can pay those licensing fees, squeezes out free software.

Even if we didn't have to worry about untoward corporate influence on procurement decisions, many of the considerations sketched above would constitute freestanding reasons to shy away from proprietary procurement. The "kingmaker" power of government in software markets, a result not only of bandwagon effects in those markets, but of the state's unique status as an entity with which citizens must deal in various ways, makes it doubly important that procurement be constrained by principles of liberal neutrality. Government is not just one more market actor: it may not practice the sort of nepotism that might be acceptable for a small family-owned business, it may not hold mandatory prayer services for its employees, and it may not engage in hostile takeovers of other firms. Private companies don't have to think about whether they're exerting undue influence on the larger market, or weigh the benefit to the general public that's created when they commission the development of OSS that any citizen can then take and use. Governments do. If we think a general "separation of economy and state" is desirable, then we should want the state to be as scrupulously neutral between firms as between faiths. James DeLong is right that public policy should be tailored to minimize the extent to which government decisions determine the shape of the market. He is wrong to think that a policy of open source procurement conflicts with that principle. In the civic space carved out by procurement, a preference for open source is a sign of appropriately "secular" restraint.

TCS Daily Archives