In November, 1995, Peter Salus, a member of the Free Software Foundation and author of the 1994 book, A Quarter Century of Unix, issued a call for papers to members of the GNU Project’s “system-discuss” mailing list. Salus, the conference’s scheduled chairman, wanted to tip off fellow hackers about the upcoming Conference on Freely Redistributable Software in Cambridge, Massachusetts. Slated for February, 1996 and sponsored by the Free Software Foundation, the event promised to be the first engineering conference solely dedicated to free software and, in a show of unity with other free software programmers, welcomed papers on “any aspect of GNU, Linux, NetBSD, 386BSD, FreeBSD, Perl, Tcl/tk, and other tools for which the code is accessible and redistributable.” Salus wrote:
Over the past 15 years, free and low-cost software has become ubiquitous. This conference will bring together implementers of several different types of freely redistributable software and publishers of such software (on various media). There will be tutorials and refereed papers, as well as keynotes by Linus Torvalds and Richard Stallman.1
One of the first people to receive Salus’ email was conference committee member Eric S. Raymond. Although not the leader of a project or company like the various other members of the list, Raymond had built a tidy reputation within the hacker community as a major contributor to GNU Emacs and as editor of The New Hacker Dictionary, a book version of the hacking community’s decade-old Jargon File.
For Raymond, the 1996 conference was a welcome event. Active in the GNU Project during the 1980s, Raymond had distanced himself from the project in 1992, citing, like many others before him, Stallman’s “micro-management” style. “Richard kicked up a fuss about my making unauthorized modifications when I was cleaning up the Emacs LISP libraries,” Raymond recalls. “It frustrated me so much that I decided I didn’t want to work with him anymore.”
Despite the falling out, Raymond remained active in the free software community. So much so that when Salus suggested a conference pairing Stallman and Torvalds as keynote speakers, Raymond eagerly seconded the idea. With Stallman representing the older, wiser contingent of ITS/Unix hackers and Torvalds representing the younger, more energetic crop of Linux hackers, the pairing indicated a symbolic show of unity that could only be beneficial, especially to ambitious younger (i.e., below 40) hackers such as Raymond. “I sort of had a foot in both camps,” Raymond says.
By the time of the conference, the tension between those two camps had become palpable. Both groups had one thing in common, though: the conference was their first chance to meet the Finnish wunderkind in the flesh. Surprisingly, Torvalds proved himself to be a charming, affable speaker. Possessing only a slight Swedish accent, Torvalds surprised audience members with his quick, self-effacing wit.2 Even more surprising, says Raymond, was Torvalds’ equal willingness to take potshots at other prominent hackers, including the most prominent hacker of all, Richard Stallman. By the end of the conference, Torvalds’ half-hacker, half-slacker manner was winning over older and younger conference-goers alike.
“It was a pivotal moment,” recalls Raymond. “Before 1996, Richard was the only credible claimant to being the ideological leader of the entire culture. People who dissented didn’t do so in public. The person who broke that taboo was Torvalds.”
The ultimate breach of taboo would come near the end of the show. During a discussion on the growing market dominance of Microsoft Windows or some similar topic, Torvalds admitted to being a fan of Microsoft’s PowerPoint slideshow software program. From the perspective of old-line software purists, it was like a Mormon bragging in church about his fondness of whiskey. From the perspective of Torvalds and his growing band of followers, it was simply common sense. Why shun worthy proprietary software programs just to make a point? Being a hacker wasn’t about suffering, it was about getting the job done.
“That was a pretty shocking thing to say,” Raymond remembers. “Then again, he was able to do that, because by 1995 and 1996, he was rapidly acquiring clout.”
Stallman, for his part, doesn’t remember any tension at the 1996 conference, but he does remember later feeling the sting of Torvalds’ celebrated cheekiness. “There was a thing in the Linux documentation which says print out the GNU coding standards and then tear them up,” says Stallman, recalling one example. “OK, so he disagrees with some of our conventions. That’s fine, but he picked a singularly nasty way of saying so. He could have just said ‘Here’s the way I think you should indent your code.’ Fine. There should be no hostility there.”
For Raymond, the warm reception other hackers gave to Torvalds’ comments merely confirmed his suspicions. The dividing line separating Linux developers from GNU/Linux developers was largely generational. Many Linux hackers, like Torvalds, had grown up in a world of proprietary software. Unless a program was clearly inferior, most saw little reason to rail against a program on licensing issues alone. Somewhere in the universe of free software systems lurked a program that hackers might someday turn into a free software alternative to PowerPoint. Until then, why begrudge Microsoft the initiative of developing the program and reserving the rights to it?
As a former GNU Project member, Raymond sensed an added dynamic to the tension between Stallman and Torvalds. In the decade since launching the GNU Project, Stallman had built up a fearsome reputation as a programmer. He had also built up a reputation for intransigence both in terms of software design and people management. Shortly before the 1996 conference, the Free Software Foundation would experience a full-scale staff defection, blamed in large part on Stallman. Brian Youmans, a current FSF staffer hired by Salus in the wake of the resignations, recalls the scene: “At one point, Peter [Salus] was the only staff member working in the office.”
For Raymond, the defection merely confirmed a growing suspicion: recent delays such as the HURD and recent troubles such as the Lucid-Emacs schism reflected problems normally associated with software project management, not software code development. Shortly after the Freely Redistributable Software Conference, Raymond began working on his own pet software project, a popmail utility called “fetchmail.” Taking a cue from Torvalds, Raymond issued his program with a tacked-on promise to update the source code as early and as often as possible. When users began sending in bug reports and feature suggestions, Raymond, at first anticipating a tangled mess, found the resulting software surprisingly sturdy. Analyzing the success of the Torvalds approach, Raymond issued a quick analysis: using the Internet as his “petri dish” and the harsh scrutiny of the hacker community as a form of natural selection, Torvalds had created an evolutionary model free of central planning.
What’s more, Raymond decided, Torvalds had found a way around Brooks’ Law. First articulated by Fred P. Brooks, manager of IBM’s OS/360 project and author of the 1975 book, The Mythical Man-Month, Brooks’ Law held that adding developers to a project only resulted in further project delays. Believing as most hackers that software, like soup, benefits from a limited number of cooks, Raymond sensed something revolutionary at work. In inviting more and more cooks into the kitchen, Torvalds had actually found away to make the resulting software better.3
Raymond put his observations on paper. He crafted them into a speech, which he promptly delivered before a group of friends and neighbors in Chester County, Pennsylvania. Dubbed “The Cathedral and the Bazaar,” the speech contrasted the management styles of the GNU Project with the management style of Torvalds and the kernel hackers. Raymond says the response was enthusiastic, but not nearly as enthusiastic as the one he received during the 1997 Linux Kongress, a gathering of Linux users in Germany the next spring.
“At the Kongress, they gave me a standing ovation at the end of the speech,” Raymond recalls. “I took that as significant for two reasons. For one thing, it meant they were excited by what they were hearing. For another thing, it meant they were excited even after hearing the speech delivered through a language barrier.”
Eventually, Raymond would convert the speech into a paper, also titled “The Cathedral and the Bazaar.” The paper drew its name from Raymond’s central analogy. GNU programs were “cathedrals,” impressive, centrally planned monuments to the hacker ethic, built to stand the test of time. Linux, on the other hand, was more like “a great babbling bazaar,” a software program developed through the loose decentralizing dynamics of the Internet.
Implicit within each analogy was a comparison of Stallman and Torvalds. Where Stallman served as the classic model of the cathedral architect—i.e., a programming “wizard” who could disappear for 18 months and return with something like the GNU C Compiler—Torvalds was more like a genial dinner-party host. In letting others lead the Linux design discussion and stepping in only when the entire table needed a referee, Torvalds had created a development model very much reflective of his own laid-back personality. From the Torvalds’ perspective, the most important managerial task was not imposing control but keeping the ideas flowing.
Summarized Raymond, “I think Linus’s cleverest and most consequential hack was not the construction of the Linux kernel itself, but rather his invention of the Linux development model.”4
In summarizing the secrets of Torvalds’ managerial success, Raymond himself had pulled off a coup. One of the audience members at the Linux Kongress was Tim O’Reilly, publisher of O’Reilly & Associates, a company specializing in software manuals and software-related books (and the publisher of this book). After hearing Raymond’s Kongress speech, O’Reilly promptly invited Raymond to deliver it again at the company’s inaugural Perl Conference later that year in Monterey, California.
Although the conference was supposed to focus on Perl, a scripting language created by Unix hacker Larry Wall, O’Reilly assured Raymond that the conference would address other free software technologies. Given the growing commercial interest in Linux and Apache, a popular free software web server, O’Reilly hoped to use the event to publicize the role of free software in creating the entire infrastructure of the Internet. From web-friendly languages such as Perl and Python to back-room programs such as BIND (the Berkeley Internet Naming Daemon), a software tool that lets users replace arcane IP numbers with the easy-to-remember domain-name addresses (e.g., amazon.com), and sendmail, the most popular mail program on the Internet, free software had become an emergent phenomenon. Like a colony of ants creating a beautiful nest one grain of sand at a time, the only thing missing was the communal self-awareness. O’Reilly saw Raymond’s speech as a good way to inspire that self-awareness, to drive home the point that free software development didn’t start and end with the GNU Project. Programming languages, such as Perl and Python, and Internet software, such as BIND, sendmail, and Apache, demonstrated that free software was already ubiquitous and influential. He also assured Raymond an even warmer reception than the one at Linux Kongress.
O’Reilly was right. “This time, I got the standing ovation before the speech,” says Raymond, laughing.
As predicted, the audience was stocked not only with hackers, but with other people interested in the growing power of the free software movement. One contingent included a group from Netscape, the Mountain View, California startup then nearing the end game of its three-year battle with Microsoft for control of the web-browser market.
Intrigued by Raymond’s speech and anxious to win back lost market share, Netscape executives took the message back to corporate headquarters. A few months later, in January, 1998, the company announced its plan to publish the source code of its flagship Navigator web browser in the hopes of enlisting hacker support in future development.
When Netscape CEO Jim Barksdale cited Raymond’s “Cathedral and the Bazaar” essay as a major influence upon the company’s decision, the company instantly elevated Raymond to the level of hacker celebrity. Determined not to squander the opportunity, Raymond traveled west to deliver interviews, advise Netscape executives, and take part in the eventual party celebrating the publication of Netscape Navigator’s source code. The code name for Navigator’s source code was “Mozilla”: a reference both to the program’s gargantuan size—30 million lines of code—and to its heritage. Developed as a proprietary offshoot of Mosaic, the web browser created by Marc Andreessen at the University of Illinois, Mozilla was proof, yet again, that when it came to building new programs, most programmers preferred to borrow on older, modifiable programs.
While in California, Raymond also managed to squeeze in a visit to VA Research, a Santa Clara–based company selling workstations with the GNU/Linux operating system preinstalled. Convened by Raymond, the meeting was small. The invite list included VA founder Larry Augustin, a few VA employees, and Christine Peterson, president of the Foresight Institute, a Silicon Valley think tank specializing in nanotechnology.
“The meeting’s agenda boiled down to one item: how to take advantage of Netscape’s decision so that other companies might follow suit?” Raymond doesn’t recall the conversation that took place, but he does remember the first complaint addressed. Despite the best efforts of Stallman and other hackers to remind people that the word “free” in free software stood for freedom and not price, the message still wasn’t getting through. Most business executives, upon hearing the term for the first time, interpreted the word as synonymous with “zero cost,” tuning out any follow up messages in short order. Until hackers found a way to get past this cognitive dissonance, the free software movement faced an uphill climb, even after Netscape.
Peterson, whose organization had taken an active interest in advancing the free software cause, offered an alternative: open source.
Looking back, Peterson says she came up with the open source term while discussing Netscape’s decision with a friend in the public relations industry. She doesn’t remember where she came upon the term or if she borrowed it from another field, but she does remember her friend disliking the term.5
At the meeting, Peterson says, the response was dramatically different. “I was hesitant about suggesting it,” Peterson recalls. “I had no standing with the group, so started using it casually, not highlighting it as a new term.” To Peterson’s surprise, the term caught on. By the end of the meeting, most of the attendees, including Raymond, seemed pleased by it.
Raymond says he didn’t publicly use the term “open source” as a substitute for free software until a day or two after the Mozilla launch party, when O’Reilly had scheduled a meeting to talk about free software. Calling his meeting “the Freeware Summit,” O’Reilly says he wanted to direct media and community attention to the other deserving projects that had also encouraged Netscape to release Mozilla. “All these guys had so much in common, and I was surprised they didn’t all know each other,” says O’Reilly. “I also wanted to let the world know just how great an impact the free software culture had already made. People were missing out on a large part of the free software tradition.”
In putting together the invite list, however, O’Reilly made a decision that would have long-term political consequences. He decided to limit the list to west-coast developers such as Wall, Eric Allman, creator of sendmail, and Paul Vixie, creator of BIND. There were exceptions, of course: Pennsylvania-resident Raymond, who was already in town thanks to the Mozilla launch, earned an quick invite. So did Virginia-resident Guido van Rossum, creator of Python. “Frank Willison, my editor in chief and champion of Python within the company, invited him without first checking in with me,” O’Reilly recalls. “I was happy to have him there, but when I started, it really was just a local gathering.”
For some observers, the unwillingness to include Stallman’s name on the list qualified as a snub. “I decided not to go to the event because of it,” says Perens, remembering the summit. Raymond, who did go, says he argued for Stallman’s inclusion to no avail. The snub rumor gained additional strength from the fact that O’Reilly, the event’s host, had feuded publicly with Stallman over the issue of software-manual copyrights. Prior to the meeting, Stallman had argued that free software manuals should be as freely copyable and modifiable as free software programs. O’Reilly, meanwhile, argued that a value-added market for nonfree books increased the utility of free software by making it more accessible to a wider community. The two had also disputed the title of the event, with Stallman insisting on “Free Software” over the less politically laden “Freeware.”
Looking back, O’Reilly doesn’t see the decision to leave Stallman’s name off the invite list as a snub. “At that time, I had never met Richard in person, but in our email interactions, he’d been inflexible and unwilling to engage in dialogue. I wanted to make sure the GNU tradition was represented at the meeting, so I invited John Gilmore and Michael Tiemann, whom I knew personally, and whom I knew were passionate about the value of the GPL but seemed more willing to engage in a frank back-and-forth about the strengths and weaknesses of the various free software projects and traditions. Given all the later brouhaha, I do wish I’d invited Richard as well, but I certainly don’t think that my failure to do so should be interpreted as a lack of respect for the GNU Project or for Richard personally.”
Snub or no snub, both O’Reilly and Raymond say the term “open source” won over just enough summit-goers to qualify as a success. The attendees shared ideas and experiences and brainstormed on how to improve free software’s image. Of key concern was how to point out the successes of free software, particularly in the realm of Internet infrastructure, as opposed to playing up the GNU/Linux challenge to Microsoft Windows. But like the earlier meeting at VA, the discussion soon turned to the problems associated with the term “free software.” O’Reilly, the summit host, remembers a particularly insightful comment from Torvalds, a summit attendee.
“Linus had just moved to Silicon Valley at that point, and he explained how only recently that he had learned that the word ‘free’ had two meanings—free as in ‘libre’ and free as in ‘gratis’—in English.”
Michael Tiemann, founder of Cygnus, proposed an alternative to the troublesome “free software” term: sourceware. “Nobody got too excited about it,” O’Reilly recalls. “That’s when Eric threw out the term ‘open source.’”
Although the term appealed to some, support for a change in official terminology was far from unanimous. At the end of the one-day conference, attendees put the three terms—free software, open source, or sourceware—to a vote. According to O’Reilly, 9 out of the 15 attendees voted for “open source.” Although some still quibbled with the term, all attendees agreed to use it in future discussions with the press. “We wanted to go out with a solidarity message,” O’Reilly says.
The term didn’t take long to enter the national lexicon. Shortly after the summit, O’Reilly shepherded summit attendees to a press conference attended by reporters from the New York Times, the Wall Street Journal, and other prominent publications. Within a few months, Torvalds’ face was appearing on the cover of Forbes magazine, with the faces of Stallman, Perl creator Larry Wall, and Apache team leader Brian Behlendorf featured in the interior spread. Open source was open for business.
For summit attendees such as Tiemann, the solidarity message was the most important thing. Although his company had achieved a fair amount of success selling free software tools and services, he sensed the difficulty other programmers and entrepreneurs faced.
“There’s no question that the use of the word free was confusing in a lot of situations,” Tiemann says. “Open source positioned itself as being business friendly and business sensible. Free software positioned itself as morally righteous. For better or worse we figured it was more advantageous to align with the open source crowd.
For Stallman, the response to the new “open source” term was slow in coming. Raymond says Stallman briefly considered adopting the term, only to discard it. “I know because I had direct personal conversations about it,” Raymond says.
By the end of 1998, Stallman had formulated a position: open source, while helpful in communicating the technical advantages of free software, also encouraged speakers to soft-pedal the issue of software freedom. Given this drawback, Stallman would stick with the term free software.
Summing up his position at the 1999 LinuxWorld Convention and Expo, an event billed by Torvalds himself as a “coming out party” for the Linux community, Stallman implored his fellow hackers to resist the lure of easy compromise.
“Because we’ve shown how much we can do, we don’t have to be desperate to work with companies or compromise our goals,” Stallman said during a panel discussion. “Let them offer and we’ll accept. We don’t have to change what we’re doing to get them to help us. You can take a single step towards a goal, then another and then more and more and you’ll actually reach your goal. Or, you can take a half measure that means you don’t ever take another step and you’ll never get there.”
Even before the LinuxWorld show, however, Stallman was showing an increased willingness to alienate his more conciliatory peers. A few months after the Freeware Summit, O’Reilly hosted its second annual Perl Conference. This time around, Stallman was in attendance. During a panel discussion lauding IBM’s decision to employ the free software Apache web server in its commercial offerings, Stallman, taking advantage of an audience microphone, disrupted the proceedings with a tirade against panelist John Ousterhout, creator of the Tcl scripting language. Stallman branded Ousterhout a “parasite” on the free software community for marketing a proprietary version of Tcl via Ousterhout’s startup company, Scriptics. “I don’t think Scriptics is necessary for the continued existence of Tcl,” Stallman said to hisses from the fellow audience members.5
“It was a pretty ugly scene,” recalls Prime Time Freeware’s Rich Morin. “John’s done some pretty respectable things: Tcl, Tk, Sprite. He’s a real contributor.”
Despite his sympathies for Stallman and Stallman’s position, Morin felt empathy for those troubled by Stallman’s discordant behavior.
Stallman’s Perl Conference outburst would momentarily chase off another potential sympathizer, Bruce Perens. In 1998, Eric Raymond proposed launching the Open Source Initiative, or OSI, an organization that would police the use of the term “open source” and provide a definition for companies interested in making their own programs. Raymond recruited Perens to draft the definition.6
Perens would later resign from the OSI, expressing regret that the organization had set itself up in opposition to Stallman and the FSF. Still, looking back on the need for a free software definition outside the Free Software Foundation’s auspices, Perens understands why other hackers might still feel the need for distance. “I really like and admire Richard,” says Perens. “I do think Richard would do his job better if Richard had more balance. That includes going away from free software for a couple of months.”
Stallman’s monomaniacal energies would do little to counteract the public-relations momentum of open source proponents. In August of 1998, when chip-maker Intel purchased a stake in GNU/Linux vendor Red Hat, an accompanying New York Times article described the company as the product of a movement “known alternatively as free software and open source.”7 Six months later, a John Markoff article on Apple Computer was proclaiming the company’s adoption of the “open source” Apache server in the article headline.8
Such momentum would coincide with the growing momentum of companies that actively embraced the “open source” term. By August of 1999, Red Hat, a company that now eagerly billed itself as “open source,” was selling shares on Nasdaq. In December, VA Linux—formerly VA Research—was floating its own IPO to historical effect. Opening at $30 per share, the company’s stock price exploded past the $300 mark in initial trading only to settle back down to the $239 level. Shareholders lucky enough to get in at the bottom and stay until the end experienced a 698% increase in paper wealth, a Nasdaq record.
Among those lucky shareholders was Eric Raymond, who, as a company board member since the Mozilla launch, had received 150,000 shares of VA Linux stock. Stunned by the realization that his essay contrasting the Stallman-Torvalds managerial styles had netted him $36 million in potential wealth, Raymond penned a follow-up essay. In it, Raymond mused on the relationship between the hacker ethic and monetary wealth:
Reporters often ask me these days if I think the open-source community will be corrupted by the influx of big money. I tell them what I believe, which is this: commercial demand for programmers has been so intense for so long that anyone who can be seriously distracted by money is already gone. Our community has been self-selected for caring about other things—accomplishment, pride, artistic passion, and each other.9
Whether or not such comments allayed suspicions that Raymond and other open source proponents had simply been in it for the money, they drove home the open source community’s ultimate message: all you needed to sell the free software concept is a friendly face and a sensible message. Instead of fighting the marketplace head-on as Stallman had done, Raymond, Torvalds, and other new leaders of the hacker community had adopted a more relaxed approach—ignoring the marketplace in some areas, leveraging it in others. Instead of playing the role of high-school outcasts, they had played the game of celebrity, magnifying their power in the process.
“On his worst days Richard believes that Linus Torvalds and I conspired to hijack his revolution,” Raymond says. “Richard’s rejection of the term open source and his deliberate creation of an ideological fissure in my view comes from an odd mix of idealism and territoriality. There are people out there who think it’s all Richard’s personal ego. I don’t believe that. It’s more that he so personally associates himself with the free software idea that he sees any threat to that as a threat to himself.”
Ironically, the success of open source and open source advocates such as Raymond would not diminish Stallman’s role as a leader. If anything, it gave Stallman new followers to convert. Still, the Raymond territoriality charge is a damning one. There are numerous instances of Stallman sticking to his guns more out of habit than out of principle: his initial dismissal of the Linux kernel, for example, and his current unwillingness as a political figure to venture outside the realm of software issues.
Then again, as the recent debate over open source also shows, in instances when Stallman has stuck to his guns, he’s usually found a way to gain ground because of it. “One of Stallman’s primary character traits is the fact he doesn’t budge,” says Ian Murdock. “He’ll wait up to a decade for people to come around to his point of view if that’s what it takes.”
Murdock, for one, finds that unbudgeable nature both refreshing and valuable. Stallman may no longer be the solitary leader of the free software movement, but he is still the polestar of the free software community. “You always know that he’s going to be consistent in his views,” Murdock says. “Most people aren’t like that. Whether you agree with him or not, you really have to respect that.”
Finland has a significant (about 6%) Swedish-speaking minority population. They call themselves “finlandssvensk” or “finlandssvenskar” and consider themselves Finns; many of their families have lived in Finland for centuries. Swedish is one of Finland’s two official languages.
Since software construction is inherently a systems effort—an exercise in complex interrelationships—communication effort is great, and it quickly dominates the decrease in individual task time brought about by partitioning. Adding more men then lengthens, not shortens, the schedule.
See Fred P. Brooks, The Mythical Man-Month (Addison Wesley Publishing, 1995)