Is Open Source a Closed Network?

In “Architectures of Participation” (Issues, Summer, 2022), Gerald Berk and Annalee Saxenian present a compelling question: “Given the complexity and divergent trajectories of today’s innovation systems, how should public policy foster innovation and openness, and support the process of making data more accessible?” In the 2000s, the authors note, open-source ecosystems, or “networks of networks,” propelled Silicon Valley’s first wave of technological advancements in internet innovations. Since the 2010s, however, as computation capabilities and software systems became larger and more complex, a handful of dominant platforms—Amazon, Google, Microsoft—appear to have abandoned that openness of the previous era.

These giants have done so in a number of ways, including by restricting access to their application programming interfaces (or APIs, which simplify software development and innovation by enabling third parties to exchange data and functionality easily and securely), by acquiring start-ups, and by developing their own proprietary (closed) systems. For example, in order to maintain and enhance their market share, these firms’ cloud platforms impose “anti-forking” requirements that restrict access to data and block software developers from building out new applications on their platforms. Also, as Berk and Saxenian write, critics say Amazon Web Services has improperly “copied the open-source code for a pioneering search engine named Elasticsearch and integrated it into its proprietary cloud services offerings,” thus making it harder for smaller companies to use the search engine to market their products. In response, the authors note, “at least eight open-source database companies, including Elastic, modified their licenses, making them so restrictive that they are no longer considered open-source by the community.”

Berk and Saxenian offer several prescriptions for policymakers, regulators, and jurists to consider. Antitrust law, or competition policy, offers one tool to restore and foster an open-source collaborative ecosystem. In opposition to today’s centralizing tendencies, the authors argue that interoperability and “the democratization of the use of data” are key to high-quality, fast-paced innovation. As they highlight, Google Cloud already collaborates with the Linux Foundation and shares revenue with its smaller partners. But, of course, the worry—at least in my mind—is that Google will not maintain the partnership when it sees an opportunity to build out proprietary products or services, particularly in areas complementary to its existing dominant position in internet search or mobile phone operating system. As a Department of Justice complaint against Google argues, the firm appears to have used restrictive licensing and distribution agreements with hardware producers for its Android operating software, requiring producers to place Google Search along with a bundle of other Google applications in prominent places on the screen. These cannot be deleted by users, and these contracts contained anti-forking requirements as well.

The DOJ has a strong case against Google, but what remedy should the agency seek? Here, Berk and Saxenian cut to the heart of the matter. Neither breaking up Google nor simply prohibiting the distribution agreements will restore the dynamic, innovative engine of open-source collaboration. We need to think beyond the dichotomy of market competition versus monopolistic or oligopolistic firms in order identify creative, forward-looking solutions. Here, we might draw from recent examples, such as the Microsoft 2001 consent decree that required interoperability for competing browsers on Microsoft’s operating system. Moving forward, courts, regulators, and legislators should consider the procompetitive benefits of interoperability and information pooling, which may answer these authors’ call for both fostering collaborative competition and maintaining economies of scale.

Laura Phillips-Sawyer

Gerald Berk and Annalee Saxenian focus on the information technology industry in Silicon Valley and the relationship between open-source and proprietary-platform companies. They base their article on interviews with a range of software developers and managers, but significantly all from inside the industry. They argue that the open-source segment of the industry is especially innovative, that the proprietary segment can and does work with open-source companies, a partnership that further strengthens the innovative capacity of the industry as a whole, and for that reason ought to be promoted by public policy. They focus on antitrust policy, but the logic would apply to other domains of policy, most particularly the immigration of high-skilled workers from abroad.

I find their argument unpersuasive and ultimately disturbing. My main concern is the way it hinges on “innovation” as if that were the singular goal of public policy and could be pursued independently of other policy issues and debates. In effect, the authors seem to be arguing that society cannot have too much innovation.

Two specific issues at the forefront of current policy debates call into question the value of unlimited innovation. One is the debate about the impact of artificial intelligence and robotics on jobs, and the fear that workers are being displaced more rapidly than they can be absorbed into other parts of the economy. The other centers on the distribution of income and social mobility; in particular, Silicon Valley is a bifurcated economy, with one sector with highly paid software developers and other professionals and the other with low-paid service sector workers catering to the professional class. I would be much more persuaded of the value of open source if the authors could show that it was not only more open to ideas but also more open to a demographically diverse workforce.

The failure to recognize these issues reflects in part the limits of the analytical lens through which the authors are working. That lens is, as the authors recognize, the network structure of the industry. Network theory in the social sciences has basically been concerned with the interaction of members within a defined network. It has not been especially concerned with the boundaries of the network, how its members are recruited, trained, and socialized to the norms and standards that govern their relationships. Indeed, it does not generally recognize that social networks are typically closed to outsiders.

A limited ethnographic literature looking at what software developers actually do on the job suggests that a better lens through which to view the industry is probably craft work, where skill and proficiency are acquired through experience on the job and close interactions with more senior developers. However, because experienced workers in such a system have to work closely with “apprentices” in order to train them, they tend to resist admission of new members from backgrounds very different from their own. The classic example is the skilled trades in the construction industry, which are notorious for resisting government pressure to admit workers from underrepresented demographic groups, particularly women and ethnic and racial minorities.

An important difference between the authors’ focus on innovation and the craft analogy emerges in the debate about expanding visas for highly skilled foreign workers. In the authors’ perspective, expansion is promoted as an impetus to innovation. In the perspective that recognizes other social goals, such expansion is also way of avoiding pressures for the upward mobility of low-skilled immigrants and their children.

Michael J. Piore