Cathedrals and Bazaars
Raymond received substantial backing (which, in the open-source world, means moral support) from the business people of the fledgling Linux industry, mostly graduates of the free-software movement who had braved the waters of the Real World by commercializing, distributing, and servicing Linux and related software.
From their perspective, Raymond’s explanation was more than just plausible—it was evocative. The business community would instinctively understand both “cathedrals” and “bazaars.” Cathedrals are designed top down. An architect draws a set of plans and explains them to the builder. The builder parcels the work among foremen, who then delegate it to workers. The organizational structure is hierarchical, and someone always maintains an overarching view of where the project is heading and what it will look like when it’s completed. Discoveries and complications along the way may lead to some design changes, but even those changes must work their way down the hierarchy.This basic management style proved to be popular far beyond the somewhat narrow confines of the cathedral construction community. When Fred Brooks revisited The Mythical Man-Month in the mid-1990s to update the lessons that he had first learned about software engineering in the 1960s, he concluded that:
Today I am more convinced than ever. Conceptual integrity is central to product quality. Having a system architect is the most important single step toward conceptual integrity. These principles are by no means limited to software systems, but to the design of any complex construct, whether a computer, an airplane, a Strategic Defense Initiative, a Global Positioning System.17
Three months after Brooks penned those words in March 1995, the Netscape IPO put the Web on our collective radar screens. The PC revolution gave way to the Internet revolution.
Software development clicked into hyperdrive. Linux grew from a hobbyist’s project into a powerful product. And leading software engineers began to rethink development. Roughly two years after Brooks concluded that a systems architect was more important than ever, Raymond first presented the bazaar model of software development.To appreciate the radical nature of the bazaar model, picture the “organized chaos” of an Arab souk. Shouting abounds, prices are negotiable, sales pitches are everywhere, and all deals emerge from intense haggling and bargaining. The experience can be a bit disorienting; it’s hard to get a holistic view of the souk.
A complex software system based on similar principles seems unlikely to work. But “conservative business people” know well that situations like the souk truly are capitalism stripped bare—and as such, they produce aggregate rational behavior. With one little metaphor and a slight change in terminology, we’ve moved from the quasi socialistsounding ideals underlying free software to the rawest known form of capitalist markets to describe open-source software—and we haven’t changed a damn thing!
Remember the unrealistic textbook model in which “market forces” drive infinite numbers of buyers and sellers toward aggregate rationality, set the prices in the right place, align supply with demand, and allocate resources in the best possible manner? Well, we just found it. The bazaar model is the basis of modern microeconomic price theory. Rationality is the property that emerges from the apparent chaos of the bazaar. And it turns out that the notion of rationality as an “emergent” property is not unique unto bazaars; it’s reflected in the “tipping to a standard” that network scientists have detected in physics, chemistry, biology, and elsewhere.18 And as we’ve already seen, these ideas had worked their way far beyond academia by the mid-1990s. Why not into the world of software development?
Raymond’s bazaar model couldn’t dispense with the architect entirely.
But it could cut the architect’s role in half and rename it the “maintainer.”19 Whereas an architect disseminates a holistic view of the completed project and then watches its pieces coalesce, the maintainer is only in charge of coalescence. The maintainer, of course, was often the one who came up with the original idea for the system and go things started—but in the world of open source, even that wasn’t necessary. The community developed protocols for handing off maintenance responsibilities, or even for claiming the role of the maintainer on a project that had somehow fallen into disrepair.20 Maintainers collect submissions, decide which to accept and which to reject, monitor bugs, release patches as needed, interact with other projects, and announce new “official” releases. Maintainers are usually individuals—Torvalds, as noted, maintains Linux—though they can also be committees or even companies. The software system is what emerges from the apparently chaotic submissions to the maintainer.If it’s immediately obvious to you why someone would want to become a maintainer, you may grok hacker culture yet.
grok vt. [from the novel Stranger in a Strange Land, by Robert A. Heinlein, where it is a Martian word meaning literally “to drink” and metaphorically “to be one with”] The emphatic form is grok in fullness. 1. To understand, usually in a global sense. Connotes intimate and exhaustive knowledge.21
After all, becoming a maintainer is the best way to become a demigod. If not, you might accept one of the other theories as to what motivates hackers. These theories include reputation within the community, resume enhancement beyond the community, and participation in what anthropologists call a “gift culture.” Suffice it to say, though, that while maintainers will never be as wealthy as Bill Gates, they are likely to find ample, rewarding, and high-compensation opportunities to customize software and to consult.
In other words, the open-source movement may not be as offbeat as it sounded at first.
The bazaar model embodies plausible approaches to both product development and commercialization. And with a particularly delicious twist of irony, the downgrading of the architect to a mere maintainer puts the capitalist shoe on the other foot. While Gates may deride Stallman’s free-software movement as socialist (or at the very least, as anticapitalist), the open-source community may now respond by echoing Scott McNealy’s charge that Microsoft’s cathedral-style design is central planning.And so there you have it. When the Internet expanded the open-source community to the entire world, it made the souk big enough for its emergent properties to possess real power and value. This expansion suggests that the information sector’s next few chapters may be shaped less by competition among companies than by competition among philosophies, approaches, product-development strategies, and business models. It also gives us a number of different ways to characterize this competition. At the obvious level, it pits secret source code against open source. At the metaphoric level, it’s cathedrals vs. bazaars. At the pejorative level, socialism squares off against central planning. Perhaps the most useful characterization, though, is that the established “product-oriented” vision of the software industry faces a challenge from a newer “service- oriented” vision of the industry.
The long and the short of it is that a valid business model does exist behind open source. The first ones to see the model were those, like Tiemann, who already lived inside the community. Open-source hackers founded companies like Cygnus Solutions, Red Hat, and VA Software, so it’s not surprising that they adopted its implicit business model. The real question was whether anyone outside the community would ever come on board. The answer came less than a year after Raymond first released The Cathedral and the Bazaar, courtesy of Netscape.
On January 23, 1998, Netscape made two announcements.
The first, as reported by C∣Net: “In an unprecedented move, Netscape Communications will give away its Navigator browser, confirming rumors over the last several weeks.”The second: “It also will give away the source code for the next generations of its Communicator suite.”
The decision to give away the browser came as no surprise, but the release of the source code stunned the industry. It hit the pages of newspapers around the world, and even the Open Source community was surprised at the move. Never before had a major software company opened up its proprietary code. What was Netscape up to now?
We had decided to change the playing field, and not for the first time. Always known for thinking outside the box, this time Netscape was taking the commitment to building a better Internet to a new level
As fast and surprising as the announcement seemed to both insiders and outsiders, it reflected several converging tracks of thought. Netscape executives were discussing a whitepaper... advocat[ing] that Netscape free its source... citing Eric Raymond’s paper, “The Cathedral and the Bazaar.”...
In the engineering pit, there was a similar view. Many Netscape employees had experience working with Open Source.22
In other words, Netscape’s management, not its hacker base, decided to open the source code. Management cited Raymond’s work as the dispositive influence and hired him as a consultant. He helped them understand not only how to set up an open-source development project, but also what rights they would need to reserve in order to keep both their existing business commitments and those that they anticipated making in the future. Thus was born Mozilla, the open-source version of Netscape’s browser.23
How did the Real World view Netscape’s bold move? Well, Cusumano and Yoffie, whose chronicle of the browser wars gave them inside access to both Microsoft and Netscape, seemed to think that it was the right strategic move at the right time.
They, of course, didn’t speak of cathedrals and bazaars—they weren’t hackers. They employed their own set of management metaphors to report the browser wars in terms of “sumo strategy” and “judo strategy”; the former pits strength against strength, while the latter attempts to turn an opponent’s own strength against it. They saw Netscape’s decision to open its source code as “classic judo. Netscape management recognized that flexibility was critical: Without the resources to fight Microsoft directly, it had to find a creative way to compete It also needed huge external resources to offsetMicrosoft’s size and financial strength. Once again, it hoped to find these resources on the Net.”24 In other words, the hacker community allowed Netscape to “hire” a huge numbers of new developers by giving them something other than cash. Netscape gave them innovative technology with which to play. The plan worked—to a point. Mozilla.org (the organizational maintainer of the open-source Mozilla browser) began to receive new and imaginative contributions within six weeks of opening its source.
But we already know how the story ended. While Netscape was trying to leverage the hacker community, Microsoft found more effective ways to leverage its own strength in Windows—and succeeded in blocking Netscape from virtually all reasonable distribution outlets. Shorn of its ability to distribute its product easily, and faced with Microsoft’s premature integration of browsing capabilities into the platform, Netscape conceded defeat before 1998 was out. Its decision to open its source code to developers couldn’t prevent Microsoft from blocking its path to consumers. The open-source community had lost its first great opportunity to demonstrate the superiority of its development and business models. At the same time, though, Netscape did provide the community with high-profile exposure, and may have made a number of people consider Linux and Apache more seriously than they might have otherwise.
Furthermore, no matter how things may have looked to the Real World, the computer industry understood what had happened. Other computer companies were well aware of Microsoft’s business tactics, and while they recognized that open-source was not, in and of itself, a strong enough model to overcome monopoly leveraging, that hardly made it a bad strategy. They began to explore its ramifications themselves, and began to contemplate ways to make money in the growing Linux and Apache markets—a trend that accelerated with the introduction of the GNOME and KDE desktops that provide a graphical interface to Linux in a role similar to the one that early versions of Windows played for DOS. Some, like Sun, developed open-source versions of important products, such as its office suite, StarOffice.25 IBM opened the source to some of its products, and sells a sizable collection of both secret and opensource applications that run on Linux. Hewlett-Packard (HP) opened the source to some of its printer drivers, and established an entire Linux Systems Division to coordinate its open-source strategy. Finally, when Merrill Lynch identified itself as part of the open-source community, the notion that open-source development was antithetical to capitalism began to ring more than a bit hollow.
With important tech companies like Sun, IBM, and HP incorporating open-source into their strategic plans, the battle no longer appears to be Richard Stallman against the world. While much of the world has yet to adopt Stallman’s more radical proposals—and may view even some of Raymond’s “pragmatism” as a bit extreme—large parts of the commercial software world clearly are drifting toward open-source and may even leave Microsoft standing out in the crowd. David Stutz, for example, generated a fair amount of controversy when he retired from Microsoft in February 2003 and immediately posted an essay on his Web site contending that Microsoft’s refusal to embrace open-source was eroding the value of its products.26 The looming battle appears to pit Bill Gates against the world. And though Stallman was easy to marginalize, the antagonists in this new battle may be evenly matched.