On April 12, 2002 the AEI-Brookings Joint Center held a conference to discuss a phenomenon that has become a touchstone for controversy in the information technology business––open source software. Will open source software programs like Linux be able to survive against the tough competition of proprietary software such as Windows?  Should the government intervene in an attempt to promote open source? The Joint Center convened an expert panel to offer diverse views on these and other related policy questions.

Robert W. Hahn, Joint Center (moderator):

Open source refers to access to the code, written in a programming language, that constitutes the working software. With open source software––for example, the Linux operating system––users and others can read the code and change it to suit their needs. Access to most proprietary software––for example, Windows––is closely guarded. Thus software creation comes from two very different and sometimes conflicting traditions––which is one of the reasons it is interesting from the standpoint of public policy.

On the one hand, programmers often flourish as part of communities that prize cooperation and openness, and status within the community is largely derived from showing how good one is at furthering the collective effort. Open source can also pay off on the demand side. Some users of software greatly value the option to customize code to their needs or simply to debug it. That’s one of the reasons Linux has been so popular for servers.

On the other hand, widely used programs like the Microsoft Office suite underpin the success of some of the most profitable corporations on the planet. And since the commercial value of these blockbuster products is almost exclusively embodied in the source code, commercial software makers typically have powerful incentives to limit access. Like other businesses relying on intellectual property, they use patents and copyrights to safeguard the value of their property. Unlike other companies however, software makers’ first and strongest line of defense is secrecy. It is relatively easy to copy a CD containing a compiled version of Windows. But the copy contains only the fuzziest clues to that design of the underlying source code––and thus would be of little help in cloning Windows.

The market seems a natural place to resolve the conflict between these diverging models for developing and distributing software. But it isn’t quite that simple. The market is governed by rules. Some open source software is licensed under terms that make it difficult for the software to coexist with commercial software. Indeed, the Free Software Foundation apparently sees the destruction of incentives to produce software under the conventional commercial model as a key objective. Its general public license (GPL) requires that any modification to the software also be licensed as open source and without charge. Thus, while Sun Microsystems was able to create a commercial blockbuster, the Solaris operating system, from a public domain version of the Unix operating system, no one could ever replicate that feat with Linux.

Arguably more important from a policy perspective, some governments have expressly tilted the playing field towards open source software, subsidizing its production and use. Singapore, for example, is offering tax breaks to companies that use the open source Linux operating system instead of commercial alternatives like Windows.

To most economists, including myself, the only justification for such intervention is to correct “market failure”––say, the inability of sellers or buyers to capture the full value of a product. Does such a failure distort the software market, or is the movement to promote open source an ironic throwback to an era in which government was widely seen as the appropriate manager of technological change?

James Bessen, Research on Innovation:

I’m going to take the point of view that (a) open source software will thrive for good solid economic reasons and (b) there is, in fact, a very big role for the government to play for both open source and proprietary software with regard to patents.

Open source vs. proprietary software is often caricatured as a contest between serious business and youthful idealism. Free, open source software is something like free love of the 60’s, and the implication is it will last just about as long. This is inaccurate.

You’re beginning to see serious investment now in open source. You’re also seeing very dynamic open source products, like the Apache web server. I think there are very good reasons to think this dynamism will continue.

There are many reasons why people get involved in open source. Altruism may be one. Then, too, people also want to participate in a community and to improve their reputations as programmers. But I think the profit motive is also at work.

Conventional markets work very well where the commodities are well defined, where the demand and the nature of the customer’s need are well defined, and where the property rights or contractual relationships are well defined. But software is not typically a standardized commodity. Packaged software has never represented as much of a third of all software investment in the U.S.

Open source is thus in some sense a better business model because it’s easier to customize. And under the GPL license, enhancements go back to the public, which enables this process to continue. So, while proprietary products aggregate consumer demand, open source does the reverse: it aggregates a diverse supply of programming resources and makes them widely available. So, it’s really not a case of private incentive vs. free love. It’s really an issue of which private incentives are most important.

Open source has succeeded so far, and I believe will continue to grow without much role of the government. In some areas of the world governments are trying to tilt the playing field, or maybe level it––this country is really biased against open source. But, ironically, open and proprietary software are threatened by the same government intervention––the patent system.

Until the mid-80’s software was protected by trade secrecy and copyright, and was highly dynamic. The only shortcoming of the system was that it didn’t provide much employment for intellectual property attorneys. The courts changed that by extended patent coverage to software.

The number of software patents has soared to about 20,000 a year, and there are about 100,000 software patents out there now. Accordingly, it is difficult for a programmer to do his job without risking infringement, because one idea in software builds on others. Patents are supposed to encourage R&D. What you actually see is that the companies that obtain these large patents reduce their R&D relative to sales.

David Evans, National Economic Research Associations:

Proposals to favor open source are coming out of the woodwork. Now, the reasons vary. The Florentine legislature is worried about “the computer science subjugation of the Italian state to Microsoft,” while the German Bundestag wants to figure out the best software for its back office. But whether you look to hot-blooded ideologues or cold-blooded technocrats, you’ll see claims that open source is more innovative, more secure, better for local industry and more likely to tame the beast from Redmond.

If an open source product like Linux gives the best performance, that’s what the government ought to use. And if all governments were doing was picking the best technology, there really wouldn’t be a whole lot to discuss today. But much of the intervention is based on the proposition that open source software is a solution to a problem that needs fixing.

As Bob Hahn noted, economists have a framework for addressing this kind of issue––a framework that has been used to address issues as diverse as global warming, utility regulation and the protection of intellectual property. We ask two questions. “Is there a market failure?,” and “Is there a government cure that is better than the disease?”

Economists tend to be skeptical about the latter: Government intervention costs money and can have unintended consequences. As governments around the world scurry away from failed efforts to manage the economy, I think anyone who claims that the government ought to be picking software technologies really has a heavy burden to bear.

But start with the question of market failure. Is there any evidence that the software industry isn’t working well? Now, some of you might say that there’s a market failure staring us right in the face––the fact that most Intel-compatible desktop computers run Microsoft Windows. But lots of firms in the economy have what some people call monopolies and nobody is proposing that the government adopt a “we don’t buy from monopolies” policy. Second, if the government ever really wanted to back another operating system, why choose Linux and why now? Why not, for example, choose the Macintosh?

Stepping back, it’s hard to see what there is about software that suggests market failure. Quality-adjusted output has risen 20 percent annually in the last 12 years. And over the last four years, quality-adjusted prices have fallen 27 percent.

The industry isn’t especially concentrated: the top 20 software vendors worldwide and only have 61 percent of global revenues. Microsoft happens to be the biggest seller now, but that is actually a recent event. And it only accounts for 11 percent of global software sales. IBM, which until very recently was the top firm, has 8 percent. Moreover, the top firms in this industry change places all the time.

Jim Bessen quite rightly says this is a different kind of industry. Competition is driven by innovation. Firms have high fixed costs and low marginal costs because they’re making intellectual property. There are also network effects in some segments. But this industry just doesn’t just seem to be screaming out for government meddling.

The software industry isn’t perfect, and open source software wouldn’t have made significant inroads if it didn’t offer something of value. Some people claim that the open source software is more innovative than proprietary software. But I’ve never seen any support for that, and to me it seems like a bogus claim. Linux is essentially an attempt to clone Unix, and a great bit of its success results from reversing the fragmentation of Unix.

But, innovative or not, open source has had no problem getting traction where it has advantages. Otherwise we wouldn’t be seeing Linux soaring in servers and edging into embedded devices.

The closest I’ve seen to a market failure is the assertion that the government needs to help open source become an alternative to Microsoft Windows. Microsoft, it’s said, has a monopoly on the desktop as a result of network effects. People want to use an operating system with lots of applications, and software developers want to write applications for popular operating systems.

If the government buys open source software, the argument goes, it’ll help generate network effects for Linux. Linux would get more traction on the desktop, which would be good for competition and for consumers.

I don’t think this argument flies. Politicians aren’t very good at picking technologies. Second, if it made sense for government to promote an alternative to Windows, there’s no apparent reason why that alternative should be Linux in particular, or open source in general. Third, Windows is extremely popular precisely because consumers and developers place a very high value on network effects. Shifting volume away from Windows would reduce those network effects and consumer welfare along with it.

Lawrence Lessig, Stanford University:

We need to distinguish between not two forms of software provision, but three. There’s free software that’s associated with the General Public License (GPL) and has produced the Linux operating system. There’s open source software, which is responsible for such products as Apache. And there’s proprietary software such as Windows.

These three kinds of software are governed by different rules, but the important difference is not the rule of copyright. All three depend upon copyright law, which essentially says, you may not use property without the permission of the owner.

Open source software can impose a range of conditions on users, but GPL is special in an important way: you must make any derivative use of redistributed software subject to the same license.

I want to reemphasize Jim Bessen’s point that this is not a debate between those who think that everybody’s going to work for nothing, and make lots of good software, and those who want to be rich. We’ve always had an important balance between resources held in common and resources that were proprietary, and the balance between open and proprietary resources is a critical feature of innovation. The copyright system itself is grounded on a constitutional provision that says, this form of monopoly protection is only granted for a limited time.

The idea is to assure that resources eventually pass into common use. There is no part of the economy that survives solely on proprietary rights. Once a lawyer files a brief, somebody else can come along and take the ideas. In fact, that’s what good lawyers do. They recombine what other people have done to produce their own product.

The particular concern here is that GPL imposes a restriction that alters the balance between proprietary and nonproprietary producers. Now, that’s an interesting concern because it rightly focuses on the extent to which we facilitate neutral development of resources. The Free Software Foundation, whose founder is with us today, was in an important sense responding to exactly the opposite concern––that software was going to get bottled up. Just as policymakers may be concerned about closing down innovation through GPL, they should be concerned about the extent to which proprietary software is unavailable to nonproprietary systems of development.

Brad Smith, Microsoft Corporation:

It’s helpful in thinking about open source to ask three questions. First, can one see the source code? Second, can one change the source code? Third, can one redistribute that code in either source or object code form, and if one can redistribute the code, what restrictions come with right? In particular, can one redistribute the code in the modified form, or just in the form in which one originally attained it? And can one charge a price when one redistributes the code? A lot of very important nuances turn on those distinctions.

Open source has always been an important part of software development, and there’s no reason to foresee that changing. Attitudes evolve to accommodate reality. For example, 12 years ago when the European community was debating its software directive, the U.S. computing industry (and IBM in particular) was really very strongly united in arguing that one shouldn’t have a right even to de-compile software, to turn it from object code into source code form, because to do so would put commercial software at risk.

A lot has changed since, and Microsoft is among the many companies that have been taking a different view about open source. We have launched a “shared source” initiative, licensing access to source code to some large customers. And we’re moving to provide selective access to PC manufacturers and systems integrators, as well as to universities.

You saw earlier that packaged software represented only one-third of total software. In fact, the software ecosystem is even larger than that. It probably is not an overstatement to say that the software ecosystem starts in Washington: the federal government funds 70 percent of all university research in computer science. The Internet is only one example of a computing initiative with federal origins. Another is the path-breaking work at Berkeley on relational databases in the 1970s.

Because that source code from that project was licensed under an open-source license without GPL-like restrictions, it was possible to commercialize it: think Sybase and Computer Associates. Similar things have been happening with artificial intelligence––and especially in speech recognition, where a whole host of companies from IBM and Microsoft to start-ups are commercializing products that build on that work.

Professor Lessig asked what industry gives back. Every time a company contributes technology to a standards organization, it gives back in a way that enables others to make use of that technology. Microsoft Research publishes hundreds of academic papers a year, as do other major computing companies. Microsoft also contributes about $75 million a year to universities to pursue basic research. And I think there are other ways we can insure that this diverse computing ecosystem works for private firms and sustains the commons.

A principle concern is the role of the GPL. When software is distributed under the GPL, it effectively becomes impossible to incorporate any of the work in a commercial product. While a company can charge for a copy of code that is distributed under the GPL, the rest of the world has a right to make copies and distribute them free of charge––which makes it very hard to build a software business. You can still sell hardware and you can still sell software, but you probably can’t build an independent commercial software company on that kind of business model.

If people are going to license software, especially from the university community under the GPL, it is important to appreciate the ramifications for the diversity of the software ecosystem. We think that diversity is a good thing, but there may be better ways to encourage diversity than through the GPL.

Responses:

Response by James Bessen: A question for David Evans: If you believe software is an industry that doesn’t require more government intervention, I assume you don’t think software patents are necessary?

Response by David Evans: Open source is a tremendous innovation in a lot of ways. It’s a great institutional structure, and GPL certainly has some advantages over the less restrictive BSD open source license. I certainly envision coexistence, and that open source will become more successful in the years ahead.

The debate over open source in D.C. is very different than the one in Brussels and in so many other places in the world. Outside the U.S., it really is a debate about whether government should help open source. Sometimes it’s put in terms of open source’s advantages in security or innovation. But it’s more or less an ideological discussion about promoting a particular technology. The momentum of this movement is through the GPL. A lot of software has been developed under the BSD license, but BSD does not have a lot of traction these days.

Regarding patents, there’s a tradeoff between giving people monopoly rights in intellectual property and making it available to all. There’s always an issue as to where to draw that line. I don’t have a firm view about patents in software. But I do believe that we don’t have a very good patent office. We don’t invest enough in examining patent applications. But this point is true for all patents, not just those in software

Response by Lawrence Lessig: There’s a lot to be said for free software, as distinct from open source––not, perhaps, if you focus on economic issues, but certainly if you consider the other value issues related to software. I mean the way in which the computer infrastructure regulates people’s lives, and the protection afforded by giving them access to the source code.

Within the economic context, I think one important reason why governments think about “subsidizing” open source is that they see an externality implied by those non-economic considerations.

Finally, regarding patents: We have good reason to be skeptical that patents will do any good where innovation is sequential and complementary. We know patents create huge costs, not just through litigation but through the ambiguity they create regarding property rights. Economists should be extraordinarily skeptical about the lawyers’ claims that this monopoly protection is necessary to induce innovation. I’m surprised how little research has been done on the relationship between patents and innovation.

Response by Brad Smith: Free software plays an important role, as does open source, as does non–open source and commercial software, regardless of whether it’s published in an open or not-open manner. To me the question comes back to how one stimulates basic research and assures that the economy as a whole is able to build upon it.

One issue that has not gotten enough discussion in Washington has been the decline in federal support for basic research in computer science. We really are looking today to the federal government as the principle source of support for basic research. Microsoft and IBM, and perhaps one or two other companies do some basic research. But outside of that it really is the Federal Government that is paying for basic research in the United States. In the long term the computing industry in this country will continue to be healthy only if there is both an ongoing commitment to basic research, and if industry is allowed to make use of the results.

Response by James Bessen: It’s important not to overemphasize the differences between open source and free software. Open-source products have worked well because informal norms have managed to achieve what is explicit in the GPL contract. Products like Apache, which is not licensed through the GPL, are still very dynamic. A company could modify Apache and take the version private, but it would be a poor business decision because the company would lose out on tremendous enhancements and innovations coming from Apache’s broad array of developers.

I will admit there has been some initial difficulty creating good business models for GPL software. But this is a very new industry. It took three decades for the automobile industry to create good business models

Conclusion:
 
To conclude the conference, Bob Hahn asked each panelist to clearly state where they stand on open source policy––is there a significant problem here––do we have a significant market failure?

James Bessen: The great bulk of software is tailored to individual needs, and two sorts of economic problems come up in markets for customized software. One is the difficulty of specifying a contract. Anybody whose tried to write a custom programming contract knows basically you can’t specify the contract until you have the software, because the software is the complete description of everything that it will do. The other difficulty of custom programming is asymmetric information. It’s very difficult to know what a custom product is worth to purchasers. Open source manages to deal with these sources of inefficiency.

Regarding government, I would roll back patent protection to where it was in the 1980s. I don’t think procurement is a big issue. But it would be nice to have a level playing field where governments give equal treatment to the procurement of open source products. I can understand why some countries are doing that.

David Evans: With respect to procurement, I would run the government like a business. If GPL produced the best information technology I’d choose it; if it didn’t, I wouldn’t. There may be externalities that would lead me to favor open source, but I’m not sure that any have been identified. With respect to patents, I don’t really understand why we’re talking this much about them. Patents are a problem for proprietary software and they’re a problem for open source software.

Lawrence Lessig: The real concern about closed systems, which are typically proprietary, is that there is a great deal of knowledge that other people can’t get access to and build on in the future. Now how do you solve that? The Free Software Foundation is using quasi-private law like copyright to force opening. Another solution might be not to grant copyright unless the source code was going to pass into the public domain in a relatively short time, or to condition copyright upon the opening of the source code to guarantee that it can be used as a building block.

Patents in agriculture offer a cautionary tale. They have forced tremendous concentration in the industry as a way of dealing with the transactions costs negotiating rights to poorly defined intellectual property. In software this same burden will be felt disproportionately by free software. So here’s a place where government intervention in favor of free software might be justified.

Robert Hahn: So, should patent protection be changed or not?

Lawrence Lessig: Absolutely.

Brad Smith: The government should continue largely as it has, with a policy of non-discrimination on open source. Second, I think Washington should encourage research in a broad sense, both in the university sector and in the private sector––for example by making permanent the R&D tax credit. Third, poor quality patents make everyone worse off; there may be some room for investment in improving the system.