Chapter 15. The Revenge of the Hackers
Eric S. Raymond
I wrote the first version of “A Brief History of Hackerdom” in 1996 as a web resource. I had been fascinated by hacker culture as a culture for many years, since long before I edited the first edition of The New Hacker’s Dictionary in 1990. By late 1993, many people (including myself) had come to think of me as the hacker culture’s tribal historian and resident ethnographer. I was comfortable in that role.
At that time, I hadn’t the faintest idea that my amateur anthropologizing could itself become a significant catalyst for change. I think nobody was more surprised than I when that happened. But the consequences of that surprise are still reverberating through the hacker culture and the technology and business worlds today.
In this essay, I’ll recapitulate from my personal point of view the events that immediately led up to the January 1998 “shot heard `round the world” of the open-source revolution. I’ll reflect on the remarkable distance we’ve come since. Then I will tentatively offer some projections into the future.
Beyond Brooks’s Law
My first encounter with Linux came in late 1993, via the pioneering Yggdrasil CD-ROM distribution. By that time, I had already been involved in the hacker culture for fifteen years. My earliest experiences had been with the primitive ARPAnet of the late 1970s; I was even briefly a tourist on the ITS machines. I had already been writing free software and posting it to Usenet before the Free Software Foundation was launched in 1984, and was one of the FSF’s first contributors. I had just published the second edition of The New Hacker’s Dictionary. I thought I understood the hacker culture—and its limitations—pretty well.
Encountering Linux was a shock. Even though I had been active in the hacker culture for many years, I still carried in my head the unexamined assumption that hacker amateurs, gifted though they might be, could not possibly muster the resources or skill necessary to produce a usable multitasking operating system. The HURD developers, after all, had been evidently failing at this for a decade.
But where they had failed, Linus Torvalds and his community had succeeded. And they did not merely fulfill the minimum requirements of stability and functioning Unix interfaces. No. They blew right past that criterion with exuberance and flair, providing hundreds of megabytes of programs, documents, and other resources. Full suites of Internet tools, desktop-publishing software, graphics support, editors, games—you name it.
Seeing this feast of wonderful code spread in front of me as a working system was a much more powerful experience than merely knowing, intellectually, that all the bits were probably out there. It was as though for years I’d been sorting through piles of disconnected car parts—only to be suddenly confronted with those same parts assembled into a gleaming red Ferrari, door open, keys swinging from the lock, and engine gently purring with a promise of power. . . .
The hacker tradition I had been observing for two decades seemed suddenly alive in a vibrant new way. In a sense, I had already been made part of this community, for several of my personal free-software projects had been added to the mix. But I wanted to get in deeper, because every delight I saw also deepened my puzzlement. It was too good!
The lore of software engineering is dominated by Brooks’s Law, which predicts that as your N number of programmers rises, work performed scales as N but complexity and vulnerability to bugs rises as N-squared. N-squared is the number of communications paths (and potential code interfaces) between developers’ code bases.
Brooks’s Law predicts that a project with thousands of contributors ought to be a flaky, unstable mess. Somehow the Linux community had beaten the N-squared effect and produced an OS of astonishingly high quality. I was determined to understand how they did it.
It took me three years of participation and close observation to develop a theory, and another year to test it experimentally. And then I sat down and wrote “The Cathedral and the Bazaar” (CatB)[1] to explain what I had seen.
Memes and Mythmaking
What I saw around me was a community which had evolved the most effective software-development method ever and didn’t know it! That is, an effective practice had evolved as a set of customs, transmitted by imitation and example, without the theory or language to explain why the practice worked.
In retrospect, lacking that theory and that language hampered us in two ways. First, we couldn’t think systematically about how to improve our own methods. Second, we couldn’t explain or sell the method to anyone else.
At the time, I was only thinking about the first effect. My only intention in writing the paper was to give the hacker culture an appropriate language to use internally, to explain itself to itself. So I wrote down what I had seen, framed as a narrative and with appropriately vivid metaphors to describe the logic that could be deduced behind the customs.
There was no really fundamental discovery in CatB. I did not invent any of the methods it describes. What is novel in that paper is not the facts but those metaphors and the narrative—a simple, powerful story that encouraged the reader to see the facts in a new way. I was attempting a bit of memetic engineering on the hacker culture’s generative myths.
I first gave the full paper at Linux Kongress, in May 1997 in Bavaria. The fact that it was received with rapt attention and thunderous applause by an audience in which there were very few native speakers of English seemed to confirm that I was onto something. But, as it turned out, the sheer chance that I was seated next to Tim O’Reilly at the Thursday night banquet set in motion a more important train of consequences.
As a long-time admirer of O’Reilly’s institutional style, I had been looking forward to meeting Tim for some years. We had a wide-ranging conversation (much of it exploring our common interest in classic science fiction) which led to an invitation for me to give CatB at Tim’s Perl Conference later in the year.
Once again, the paper was well-received—with cheers and a standing ovation, in fact. I knew from my email that since Bavaria, word about CatB had spread over the Internet like a fire in dry grass. Many in the audience had already read it, and my speech was less a revelation of novelty for them than an opportunity to celebrate the new language and the consciousness that went with it. That standing ovation was not so much for my work as for the hacker culture itself—and rightly so.
Though I didn’t know it, my experiment in memetic engineering was about to light a bigger fire. Some of the people for whom my speech was genuinely novel were from Netscape Communications, Inc. And Netscape was in trouble.
Netscape, a pioneering Internet-technology company and Wall Street highflier, had been targeted for destruction by Microsoft. Microsoft rightly feared that the open Web standards embodied by Netscape’s browser might lead to an erosion of the Redmond giant’s lucrative monopoly on the PC desktop. All the weight of Microsoft’s billions, and shady tactics that would later trigger an antitrust lawsuit, were deployed to crush the Netscape browser.
For Netscape, the issue was less browser-related income (never more than a small fraction of their revenues) than maintaining a safe space for their much more valuable server business. If Microsoft’s Internet Explorer achieved market dominance, Microsoft would be able to bend the Web’s protocols away from open standards and into proprietary channels that only Microsoft’s servers would be able to service.
Within Netscape there was intense debate about how to counter the threat. One of the options proposed early on was to throw the Netscape browser source open—but it was a hard case to argue without strong reasons to believe that doing so would prevent Internet Explorer dominance.
I didn’t know it at the time, but CatB became a major factor in making that case. Through the winter of 1997, as I was working on the material for my next paper, the stage was being set for Netscape to break the rules of the commercial game and offer my tribe an unprecedented opportunity.
The Road to Mountain View
On January 23rd, 1998, Netscape announced that it would release the sources of the Netscape client line to the Internet. Shortly after the news reached me the following day, I learned that CEO Jim Barksdale described my work to national-media reporters as “fundamental inspiration” for the decision.
This was the event that commentators in the computer trade press would later call “the shot heard `round the world”—and Barksdale had cast me as its Thomas Paine, whether I wanted the role or not. For the first time in the history of the hacker culture, a Fortune 500 darling of Wall Street had bet its future on the belief that our way was right. And, more specifically, that my analysis of “our way” was right.
This is a pretty sobering kind of shock to deal with. I had not been very surprised when CatB altered the hacker culture’s image of itself; that was the result I had been trying for, after all. But I was astonished (to say the least) by the news of its success on the outside. So I did some very hard thinking in the first few hours after word reached me. About the state of Linux and the hacker community. About Netscape. And about whether I, personally, had what it would take to make the next step.
It was not difficult to conclude that helping Netscape’s gamble succeed had just become a very high priority for the hacker culture, and thus for me personally. If Netscape’s gamble failed, we hackers would probably find all the opprobrium of that failure piled on our heads. We’d be discredited for another decade. And that would be just too much to take.
By this time I had been in the hacker culture, living through its various phases, for twenty years. Twenty years of repeatedly watching brilliant ideas, promising starts, and superior technologies crushed by slick marketing. Twenty years of watching hackers dream and sweat and build, too often only to watch the likes of the bad old IBM or the bad new Microsoft walk away with the real-world prizes. Twenty years of living in a ghetto—a fairly comfortable ghetto full of interesting friends, but still one walled in by a vast and intangible barrier of prejudice inscribed “ONLY FLAKES LIVE HERE.”
The Netscape announcement cracked that barrier, if only for a moment; the business world had been jolted out of its complacency about what “hackers” are capable of. But lazy mental habits have huge inertia. If Netscape failed, or perhaps even if they succeeded, the experiment might come to be seen as a unique one-off not worth trying to repeat. And then we’d be back in the same ghetto, walls higher than before.
To prevent that, we needed Netscape to succeed. So I considered what I had learned about bazaar-mode development, and called up Netscape, and offered to help with developing their license and in working out the details of the strategy. In early February I flew to Mountain View at their request for seven hours of meetings with various groups at Netscape HQ, and helped them develop the outline of what would become the Mozilla Public License and the Mozilla organization.
While there, I met with several key people in the Silicon Valley and national Linux community (this part of the history is told in more detail on the Open Source web site’s history page[2]). While helping Netscape was clearly a short-term priority, everybody I spoke with had already understood the need for some longer-term strategy to follow up on the Netscape release. It was time to develop one.
The Origins of “Open Source”
It was easy to see the outlines of the strategy. We needed to take the pragmatic arguments I had pioneered in CatB, develop them further, and push them hard, in public. Because Netscape itself had an interest in convincing investors that its strategy was not crazy, we could count on them to help the promotion. We also recruited Tim O’Reilly (and through him, O’Reilly & Associates) very early on.
The real conceptual breakthrough, though, was admitting to ourselves that what we needed to mount was in effect a marketing campaign—and that it would require marketing techniques (spin, image-building, and re-branding) to make it work.
Hence the term “open source,” which the first participants in what would later become the Open Source campaign (and, eventually, the Open Source Initiative organization) invented at a meeting held in Mountain View in the offices of VA Research on February 3.
It seemed clear to us in retrospect that the term “free software” had done our movement tremendous damage over the years. Part of this stemmed from the well-known “free-speech/free-beer” ambiguity. Most of it came from something worse—the strong association of the term “free software” with hostility to intellectual property rights, communism, and other ideas hardly likely to endear themselves to an MIS manager.
It was, and still is, beside the point to argue that the Free Software Foundation is not hostile to all intellectual property and that its position is not exactly communistic. We knew that. What we realized, under the pressure of the Netscape release, was that FSF’s actual position didn’t matter. Only the fact that its evangelism had backfired (associating “free software” with these negative stereotypes in the minds of the trade press and the corporate world) actually mattered.
Our success after Netscape would depend on replacing the negative FSF stereotypes with positive stereotypes of our own—pragmatic tales, sweet to managers’ and investors’ ears, of higher reliability and lower cost and better features.
In conventional marketing terms, our job was to re-brand the product, and build its reputation into one the corporate world would hasten to buy.
Linus Torvalds endorsed the idea the day after that first meeting. We began acting on it within a few days after. Bruce Perens had the opensource.org domain registered and the first version of the Open Source web site[3] up within a week. He also suggested that the Debian Free Software Guidelines become the “Open Source Definition,”[4] and began the process of registering “Open Source” as a certification mark so that we could legally require people to use “Open Source” for products conforming to the OSD.
Even the particular tactics needed to push the strategy seemed pretty clear to me even at this early stage (and were explicitly discussed at the initial meeting). Key themes:
- Forget bottom-up; work on top-down
One of the things that seemed clearest was that the historical Unix strategy of bottom-up evangelism (relying on engineers to persuade their bosses by rational argument) had been a failure. This was naive and easily trumped by Microsoft. Further, the Netscape breakthrough didn’t happen that way. It happened because a strategic decision-maker (Jim Barksdale) got the clue and then imposed that vision on the people below him.
The conclusion was inescapable. Instead of working bottom-up, we should be evangelizing top-down—making a direct effort to capture the CEO/CTO/CIO types.
- Linux is our best demonstration case
Promoting Linux must be our main thrust. Yes, there are other things going on in the open-source world, and the campaign will bow respectfully in their direction—but Linux started with the best name recognition, the broadest software base, and the largest developer community. If Linux can’t consolidate the breakthrough, nothing else will, pragmatically speaking, have a prayer.
- Capture the Fortune 500
There are other market segments that spend more dollars (small-business and home-office being the most obvious examples) but those markets are diffuse and hard to address. The Fortune 500 doesn’t merely have lots of money, it concentrates lots of money where it’s relatively easy to get at. Therefore, the software industry largely does what the Fortune 500 business market tells it to do. And therefore, it is primarily the Fortune 500 we need to convince.
- Co-opt the prestige media that serve the Fortune 500
The choice to target the Fortune 500 implies that we need to capture the media that shape the climate of opinion among top-level decision-makers and investors: very specifically, the New York Times, the Wall Street Journal, the Economist, Forbes, and Barron’s Magazine.
On this view, co-opting the technical trade press is necessary but not sufficient; it’s important essentially as a pre-condition for storming Wall Street itself through the elite mainstream media.
- Educate hackers in guerilla marketing tactics
It was also clear that educating the hacker community itself would be just as important as mainstream outreach. It would be insufficient to have one or a handful of ambassadors speaking effective language if, at the grassroots level, most hackers were making arguments that didn’t work.
- Use the Open Source certification mark to keep things pure
One of the threats we faced was the possibility that the term “open source” would be “embraced and extended” by Microsoft or other large vendors, corrupting it and losing our message. It is for this reason that Bruce Perens and I decided early on to register the term as a certification mark and tie it to the Open Source Definition (a copy of the Debian Free Software Guidelines). This would allow us to scare off potential abusers with the threat of legal action.
The Accidental Revolutionary
Planning this kind of strategy was relatively easy. The hard part (for me, anyway) was accepting what my own role had to be.
One thing I understood from the beginning is that the press almost completely tunes out abstractions. They won’t write about ideas without larger-than-life personalities fronting them. Everything has to be story, drama, conflict, sound bites. Otherwise most reporters will simply go to sleep—and if they don’t, their editors will.
Accordingly, I knew somebody with very particular characteristics would be needed to front the community’s response to the Netscape opportunity. We needed a firebrand, a spin doctor, a propagandist, an ambassador, an evangelist—somebody who could dance and sing and shout from the housetops and seduce reporters and huggermug with CEOs and bang the media machine until its contrary gears ground out the message: The revolution is here!
Unlike most hackers, I have the brain chemistry of an extrovert and had already had extensive experience at dealing with the press. Looking around me, I couldn’t see anyone better qualified to play evangelist. But I didn’t want the job, because I knew it would cost me my life for many months, maybe for years. My privacy would be destroyed. I’d probably end up caricatured as a geek by the mainstream press and (worse) despised as a sell-out or glory-hog by a significant fraction of my own tribe. Worse than all the other bad consequences put together, I probably wouldn’t have time to hack any more!
I had to ask myself: are you fed up enough with watching your tribe lose to do whatever it takes to win? I decided the answer was yes—and having so decided, threw myself into the dirty but necessary job of becoming a public figure and media personality.
I’d learned some basic media chops while editing The New Hacker’s Dictionary. This time I took it much more seriously and developed an entire theory of media manipulation which I then proceeded to apply. This is not the place to describe the theory in detail, but it centers around the use of what I call “attractive dissonance'' to fan an itchy curiosity about the evangelist, and then exploiting that itch for all it’s worth in promoting the ideas.
The combination of the “open source” label and deliberate promotion of myself as an evangelist turned out to have both the good and bad consequences that I expected. The ten months after the Netscape announcement featured a steady exponential increase in media coverage of Linux and the open-source world in general. Throughout this period, approximately a third of these articles quoted me directly; most of the other two-thirds used me as a background source. At the same time, a vociferous minority of hackers declared me an evil egotist. I managed to preserve a sense of humor about both outcomes (though occasionally with some difficulty).
My plan from the beginning was that, eventually, I would hand off the evangelist role to some successor, either an individual or organization. There would come a time when charisma became less effective than broad-based institutional respectability (and, from my own point of view, the sooner the better!). At the time of this writing I am attempting to transfer my personal connections and carefully built-up reputation with the press to the Open Source Initiative, an incorporated nonprofit formed specifically to manage the Open Source trademark. I am currently the president of this organization, but hope and expect not to remain so indefinitely.
Phases of the Campaign
The open-source campaign began with the Mountain View meeting, and rapidly collected an informal network of allies over the Internet (including key people at Netscape and O’Reilly & Associates). Where I write “we” below I’m referring to that network.
From February 3 to around the time of the actual Netscape release on March 31, our primary concern was convincing the hacker community “open source” label and the arguments that went with it represented our best shot at persuading the mainstream. As it turned out, the change was rather easier than we expected. We discovered a lot of pent-up demand for a message less doctrinaire than the Free Software Foundation’s.
When the twenty-odd community leaders at the Free Software Summit on March 7 voted to adopt the term “open source,” they formally ratified a trend that was already clear at the grass roots among developers. By six weeks after the Mountain View meeting, a healthy majority of the community was speaking our language.
In April after the Summit and the actual Netscape release, our main concern shifted to recruiting as many open-source early adopters as possible. The goal was to make Netscape’s move look less singular—and to buy us insurance in case Netscape executed poorly and failed its goals.
This was the most worrying time. On the surface, everything seemed to be coming up roses; Linux was moving technically from strength to strength, the wider open-source phenomenon was enjoying a spectacular explosion in trade press coverage, and we were even beginning to get positive coverage in the mainstream press. Nevertheless, I was uneasily aware that our success was still fragile. After an initial flurry of contributions, community participation in Mozilla was badly slowed down by its requirement of Motif. None of the big independent software vendors had yet committed to Linux ports. Netscape was still looking lonely, and its browser still losing market share to Internet Explorer. Any serious reverse could lead to a nasty backlash in the press and public opinion.
Our first serious post-Netscape breakthrough came on May 7 when Corel Computer announced its Linux-based Netwinder network computer. But that wasn’t enough in itself; to sustain the momentum, we needed commitments not from hungry second-stringers but from industry leaders. Thus, it was the mid-July announcements by Oracle and Informix that really closed out this vulnerable phase.
The database outfits joined the Linux party three months earlier than I expected, but none too soon. We had been wondering how long the positive buzz could last without major Independent Software Vendor (ISV) support and feeling increasingly nervous about where we’d actually find that. After Oracle and Informix announced Linux ports other ISVs began announcing Linux support almost as a matter of routine, and even a failure of Mozilla became survivable.
Mid-July through the beginning of November was a consolidation phase. It was during this time that we started to see fairly steady coverage from the elite media I had originally targeted, led off by articles in The Economist and a cover story in Forbes. Various hardware and software vendors sent out feelers to the open-source community and began to work out strategies for getting an advantage from the new model. And internally, the biggest closed-source vendor of them all was beginning to get seriously worried.
Just how worried became apparent when the now-infamous “Halloween Documents”[5] leaked out of Microsoft.
The Halloween Documents were dynamite. They were a ringing testimonial to the strengths of open-source development from the company with the most to lose from Linux’s success. And they confirmed a lot of peoples’ darkest suspicions about the tactics Microsoft would consider in order to stop it.
The Halloween Documents attracted massive press coverage in the first few weeks of November. They created a new surge of interest in the open-source phenomenon, serendipitously confirming all the points we had been making for months. And they led directly to a request for me to conference with a select group of Merrill Lynch’s major investors on the state of the software industry and the prospects for open source.
Wall Street, finally, came to us.
The Facts on the Ground
While the Open Source campaign’s “air war” in the media was going on, key technical and market facts on the ground were also changing. I’ll review some of them briefly here because they combine interestingly with the trends in press and public perception.
In the ten months following the Netscape release, Linux rapidly continued to grow more capable. The development of solid SMP support and the effective completion of the 64-bit cleanup laid important groundwork for the future.
The roomful of Linux boxes used to render scenes for Titanic threw a healthy scare into builders of expensive graphics engines. Then the Beowulf supercomputer-on-the-cheap project showed that Linux’s Chinese-army sociology could be successfully applied even to bleeding-edge scientific computing.
Nothing dramatic happened to vault Linux’s open-source competitors into the limelight. And proprietary Unixes continued to lose market share; in fact, by mid-year only NT and Linux were actually gaining market share in the Fortune 500, and by late fall Linux was gaining faster.
Apache continued to increase its lead in the web server market. In November, Netscape’s browser reversed its market-share slide and began to make gains against Internet Explorer.
Into the Future
I have rehearsed recent history here only partly to get it into the record. More importantly, it sets a background against which we can understand near-term trends and project some things about the future (I write in mid-December of 1998).
First, safe predictions for the next year:
The open-source developer population will continue to explode, a growth fueled by ever-cheaper PC hardware and Internet connections.
Linux will continue to lead the way, the relative size of its developer community overpowering the higher average skill of the open-source BSD people and the tiny HURD crew.
ISV commitments to support the Linux platform will increase dramatically; the database-vendor commitments were a turning point. Corel’s commitment to ship their entire office suite on Linux points the way.
The Open Source campaign will continue to build on its victories and successfully raise awareness at the CEO/CTO/CIO and investor level. MIS directors will feel increasing pressure to go with open-source products not from below but from above.
Stealth deployments of Samba-over-Linux will replace increasing numbers of NT machines even at shops that have all-Microsoft policies.
The market share of proprietary Unixes will continue to gradually erode. At least one of the weaker competitors (likely DG-UX or HP-UX) will actually fold. But by the time it happens, analysts will attribute it to Linux’s gains rather than Microsoft’s.
Microsoft will not have an enterprise-ready operating system, because Windows 2000 will not ship in a usable form. (At 60 million lines of code and still bloating, its development is out of control.)
Extrapolating these trends certainly suggests some slightly riskier predictions for the medium term (eighteen to thirty-two months out):
Support operations for commercial customers of open-source operating systems will become big business, both feeding off of and fueling the boom in business use.
Open-source operating systems (with Linux leading the way) will capture the ISP and business data-center markets. NT will be unable to resist this change effectively; the combination of low cost, open sources, and 24/7 reliability will prove irresistible.
The proprietary-Unix sector will almost completely collapse. Solaris looks like a safe bet to survive on high-end Sun hardware, but most other players’ proprietary will quickly become legacy systems.
Windows 2000 will be either canceled or dead on arrival. Either way it will turn into a horrendous train wreck, the worst strategic disaster in Microsoft’s history. However, this will barely affect their hold on the desktop market within the next two years.
At first glance, these trends look like a recipe for leaving Linux as the last one standing. But life is not that simple (and Microsoft derives such immense amounts of money and market clout from the desktop market that it can’t safely be counted out even after the Windows 2000 train wreck).
So at two years out the crystal ball gets a bit cloudy. Which of several futures we get depends on questions like: Will the Department of Justice break up Microsoft? Might BeOS or OS/2 or Mac OS/X or some other niche closed-source OS, or some completely new design, find a way to go open and compete effectively with Linux’s 30-year-old base design? Will Y2K-related problems have thrown the world economy into a deep enough depression to throw off everybody’s timetables?
These are all fairly imponderable. But there is one such question that is worth pondering: Will the Linux community actually deliver a good end-user-friendly GUI interface for the whole system?
I think the most likely scenario for two years out has Linux in effective control of servers, data centers, ISPs, and the Internet, while Microsoft maintains its grip on the desktop. Where things go from there depend on whether GNOME, KDE, or some other Linux-based GUI (and the applications built or rebuilt to use it) ever get good enough to challenge Microsoft on its home ground.
If this were primarily a technical problem, the outcome would hardly be in doubt. But it isn’t; it’s a problem in ergonomic design and interface psychology, and hackers have historically been poor at it. That is, while hackers can be very good at designing interfaces for other hackers, they tend to be poor at modeling the thought processes of the other 95% of the population well enough to write interfaces that J. Random End-User and his Aunt Tillie will pay to buy.
Applications were this year’s problem; it’s now clear we’ll swing enough ISVs to get the ones we don’t write ourselves. I believe the problem for the next two years is whether we can grow enough to meet (and exceed!) the interface-design quality standard set by the Macintosh, combining that with the virtues of the traditional Unix way.
We half-joke about “world domination,” but the only way we will get there is by serving the world. That means J. Random End-User and his Aunt Tillie; and that means learning how to think about what we do in a fundamentally new way, and ruthlessly reducing the user-visible complexity of the default environment to an absolute minimum.
Computers are tools for human beings. Ultimately, therefore, the challenges of designing hardware and software must come back to designing for human beings—all human beings.
This path will be long, and it won’t be easy. But we owe it to ourselves and each other to do it right. May the Open Source be with you!
Get Open Sources now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.
