O’Reilly published two books in 1999 that helped explain open source to a wider world. Open Sources: Voices from the Open Source Revolution, edited by Chris DiBona, Sam Ockman, and Mark Stone, came out in January. Its contributors included people building the software, selling services around it, writing licenses, and organizing its institutions.1 Eric Raymond’s The Cathedral and the Bazaar followed in October, collecting essays he had circulated online, including the title essay first presented in 1997. A revised edition appeared in 2001.23
The term “open source” was barely a year old. Christine Peterson places its naming discussions in February 1998.4 These books were arguments by people with a case to win, but they were also reports from work already underway. Apache had led Netcraft’s web-server survey in 1996.5 At TheStreet.com I moved from a commercial Netscape server to Apache myself. That success was evidence available to the writers, rather than a prediction they still had to prove.
The test now is more specific than whether they were optimistic. Which of their explanations still helps us understand how shared software works, and which needs revision for AI?
Freedom belonged to the user
Bruce Perens’s essay in Open Sources starts with familiar frustrations: software that stops working after a hardware change, needs an improvement its vendor won’t make, or becomes unmaintained when the company disappears. His account of the Open Source Definition connects licensing directly to a user’s ability to keep using and changing a program.6
That argument has aged well. The valuable thing isn’t simply a public file. It is the permission and practical ability to continue. The license protects a user’s options even when the original developer loses interest. Perens also discusses companies cooperating and people selling support, so commercial participation is inside the argument from the beginning.
The AI version of his test is straightforward. If a provider withdraws a model, can you keep running the version you have? Can you modify it and pass that version along? A downloadable checkpoint answers part of the question. Its terms and the code needed to run it answer more. Rebuilding its training process is a further capability, which the earlier software examples didn’t require us to separate so carefully.
Business already shaped the platform
Brian Behlendorf’s “Open Source as a Business Strategy” is especially useful beside the previous chapter’s FOCUS story. He addresses companies deciding whether to start an open project and notes that developers’ contributions increasingly belong to their salaried work. He then examines APIs, protocols, standards, and the cost of switching vendors.7
His argument is that a customer needs an affordable alternative, not just a theoretical right to choose one. A shared specification helps, but an open reference implementation can keep the platform available if a proprietary implementation disappears. That is a stronger account of commercial strategy than a simple contrast between generous volunteers and companies selling their work.
The institutions were there too. Richard Stallman’s essay describes the Free Software Foundation’s income, donations, and paid staff, including the developers of the GNU C library and Bash. It also explains copyleft’s legal purpose.8 Today’s foundations are larger and often funded by competing vendors, but budgets, employees, and legal arrangements were already part of the story.
Michael Tiemann’s account of Cygnus and Robert Young’s account of Red Hat examine businesses built around free software.910 Support and distribution worked. Later models added proprietary features, managed operations, and customer relationships that could be harder to move than the code. MongoDB and Elastic changed licenses after disputes about cloud vendors’ use of their software. HashiCorp followed in 2023, citing competing services built without sufficient contributions back.11
The change is in where control accumulates. A customer may have a right to fork a database while its identity, data, billing, and operations remain tied to a service. Behlendorf’s switching-cost question still applies. The list of things that have to move has grown.
Attention was active work
Raymond’s strongest ideas concern how people coordinate. “Release early, release often” asks developers to expose runnable work to users soon enough for their corrections to matter.12 His many-eyes argument depends on people actually testing and investigating what they receive. Availability makes that possible; it doesn’t supply the attention.
Nor does a bazaar require an absence of leadership. Raymond describes a coordinator who recruits contributors, listens, and turns their work into releases. In the revised edition’s discussion of GCC and its EGCS fork, both sides had free software and overlapping contributors; release practice and control of development distinguished them.13 A project with a small maintainer group can still invite broad public development. Counting leaders alone doesn’t tell us which model it follows.
Heartbleed exposed a gap between a component’s importance and the attention paid to it. The flaw sat in OpenSSL for about two years before its April 2014 disclosure.14 The project had one full-time team member, according to the OpenSSL Software Foundation’s Steve Marquess.15 Many users depended on the code; very few supported sustained work on it.
The xz utils backdoor in 2024 exposed a different failure. An attacker spent years winning the trust of an overextended maintainer.16 More public submissions would not necessarily have helped. Someone had to evaluate the contributor, the changes, and the project’s ability to keep reviewing them.
These cases don’t show that public review is useless. They show that its preconditions have to be funded and defended. Jono Bacon’s The Art of Community, drawing on his work at Ubuntu, treats governance, onboarding, and conflict as practical infrastructure.17 Those are part of what turns an available source tree into a functioning project.
What AI changes
Clay Shirky’s Here Comes Everybody (2008) argued that falling coordination costs let groups organize without the machinery previously required.18 Generative AI lowers another cost: producing something that looks like a contribution. A plausible issue, patch, or security report can arrive before its author understands the project.
That puts pressure on the reputation economy Raymond described in “Homesteading the Noosphere.”19 A record of useful work still matters. The first signal of effort, a coherent submission, is now much cheaper to imitate. Maintainers have to discover whether it contains understanding and whether someone will remain responsible for it after review starts.
AI also changes what users need to inspect. A model release can include weights and inference code without its training data information or training procedure. A hosted service adds filters, system instructions, and infrastructure outside that release. Applying Perens’s freedom test or Behlendorf’s switching test now means checking several artifacts and the terms on each.
The comfort trap
Open source succeeded well enough that it is easy to forget the work behind it. I grew up in a generation that had to argue for it when many companies didn’t believe it was valuable. Some people spent years building the software and persuading their employers that sharing it wasn’t a threat to capitalism.
In the 1980s and early 1990s, packaged software was king, and much of it was awful. I won’t name companies, partly because some whose software pushed me toward open source now champion it. We needed an open foundation under the systems we were building, and for the most part we got one. The infrastructure under the web is largely open, even when the products running on it are not.
That success didn’t solve maintenance. Heartbleed prompted the Core Infrastructure Initiative; xz prompted CISA to warn against placing the security burden on an individual maintainer.2021 Tidelift’s 2024 survey still found that 60 percent of participating maintainers were unpaid.22 That survey doesn’t measure the whole funding history, but it shows how much essential work still depends on people finding time for it.
The earlier books gave us useful tests, not a guarantee that the institutions would look after themselves. Public source needs people who can review it. Freedom to move needs a workable alternative. Commercial sponsorship needs rules that let others participate. The contribution bottleneck in the next chapter begins where the first of those conditions stops keeping pace with the work arriving at the door.
Notes
Chris DiBona, Sam Ockman, and Mark Stone, eds., Open Sources: Voices from the Open Source Revolution, 1st ed. (Sebastopol, CA: O’Reilly & Associates, January 1999), ISBN 1-56592-582-3, full text at https://www.oreilly.com/openbook/opensources/book/index.html. The table of contents lists essays by Eric S. Raymond, Marshall Kirk McKusick, Scott Bradner, Richard Stallman, Michael Tiemann, Paul Vixie, Linus Torvalds, Robert Young, Larry Wall, Brian Behlendorf, Bruce Perens, Tim O’Reilly, and Jim Hamerly and Tom Paquin with Susan Walton. ↩
Eric S. Raymond, The Cathedral and the Bazaar: Musings on Linux and Open Source by an Accidental Revolutionary, 1st ed. (Sebastopol, CA: O’Reilly & Associates, October 1999), ISBN 1-56592-724-9; the October 1999 first-edition date is from the printing history in the revised edition, https://monoskop.org/images/e/e0/Raymond_Eric_S_The_Cathedral_and_the_Bazaar_rev_ed.pdf. On the Linux Kongress presentation (Würzburg, May 27, 1997), see “The Cathedral and the Bazaar,” Wikipedia, https://en.wikipedia.org/wiki/The_Cathedral_and_the_Bazaar, and the essay’s revision history at http://www.catb.org/~esr/writings/cathedral-bazaar/cathedral-bazaar, which marks revision 1.16 (21 May 1997) as “First official presentation at the Linux Kongress.” ↩
Eric S. Raymond, The Cathedral and the Bazaar, rev. ed. (Sebastopol, CA: O’Reilly & Associates, January 2001), ISBN 0-596-00108-8, https://www.oreilly.com/library/view/the-cathedral/0596001088; full text at https://monoskop.org/images/e/e0/Raymond_Eric_S_The_Cathedral_and_the_Bazaar_rev_ed.pdf. ↩
Christine Peterson, “How I coined the term ‘open source,’” Opensource.com, February 2018, https://opensource.com/article/18/2/coining-term-open-source-software. Peterson wrote the account in 2006 and places the naming discussions in the first week of February 1998, including a February 5, 1998 meeting at VA Research. ↩
Netcraft, “December 1996 Web Server Survey,” Netcraft, November 30, 1996, https://www.netcraft.com/blog/december-1996-web-server-survey. Apache had 41 percent of surveyed sites, ahead of Netscape and NCSA. ↩
Bruce Perens, “The Open Source Definition,” in Chris DiBona, Sam Ockman, and Mark Stone, eds., Open Sources: Voices from the Open Source Revolution (O’Reilly, January 1999), https://www.oreilly.com/openbook/opensources/book/perens.html. The opening examples concern users’ continued use, modification, and maintenance; the essay also explains commercial redistribution and competing support providers. ↩
Brian Behlendorf, “Open Source as a Business Strategy,” in Open Sources (O’Reilly, January 1999), especially “It’s All About Platforms,” https://www.oreilly.com/openbook/opensources/book/brian.html. The essay discusses salaried contributions, companies starting projects, nonproprietary platforms, switching costs, and reference implementations. ↩
Richard Stallman, “The GNU Operating System and the Free Software Movement,” in Open Sources (O’Reilly, January 1999), sections “The Free Software Foundation” and “Copyleft and the GNU GPL,” https://www.oreilly.com/openbook/opensources/book/stallman.html. ↩
Michael Tiemann, “Future of Cygnus Solutions: An Entrepreneur’s Account,” in DiBona, Ockman, and Stone, Open Sources (O’Reilly, 1999), https://www.oreilly.com/openbook/opensources/book/tiemans.html. ↩
Robert Young, “Giving It Away: How Red Hat Software Stumbled Across a New Economic Model and Helped Improve an Industry,” in Open Sources (O’Reilly, 1999), https://www.oreilly.com/openbook/opensources/book/young.html; see also Tiemann, note 8, and Brian Behlendorf, “Open Source as a Business Strategy,” in the same volume, https://www.oreilly.com/openbook/opensources/book/brian.html. ↩
MongoDB, Inc., “MongoDB Issues New Server Side Public License for MongoDB Community Server,” press release, October 16, 2018, https://www.mongodb.com/company/newsroom/press-releases/mongodb-issues-new-server-side-public-license-for-mongodb-community-server; Shay Banon, “Doubling down on open, Part II,” Elastic blog, 2021 (license change ahead of the 7.11 release), https://www.elastic.co/blog/licensing-change; Armon Dadgar, “HashiCorp adopts Business Source License,” HashiCorp blog, August 10, 2023, https://www.hashicorp.com/blog/hashicorp-adopts-business-source-license. MongoDB’s release blames “cloud vendors who have not developed the software”; Elastic’s post names “cloud service providers” and Amazon Elasticsearch Service. HashiCorp’s post says it partners closely with the cloud providers and aims the change at “other vendors who take advantage of pure OSS models” without “material contributions back.” ↩
Eric S. Raymond, “The Cathedral and the Bazaar,” section “Release Early, Release Often,” catb.org, version 3.0, http://www.catb.org/~esr/writings/cathedral-bazaar/cathedral-bazaar/ar01s04.html. ↩
Eric S. Raymond, The Cathedral and the Bazaar, revised edition (O’Reilly, January 2001), “Necessary Preconditions for the Bazaar Style,” pp.49–50, and the GCC/EGCS endnote, https://monoskop.org/images/e/e0/Raymond_Eric_S_The_Cathedral_and_the_Bazaar_rev_ed.pdf. This is the revised edition’s retrospective account, not a prediction attributed to the January 1999 Open Sources anthology. ↩
National Vulnerability Database, “CVE-2014-0160 Detail,” NIST, https://nvd.nist.gov/vuln/detail/cve-2014-0160; CISA, “OpenSSL ‘Heartbleed’ vulnerability (CVE-2014-0160),” alert, April 8, 2014, https://www.cisa.gov/news-events/alerts/2014/04/08/openssl-heartbleed-vulnerability-cve-2014-0160. The flawed code entered OpenSSL’s repository on December 31, 2011 and shipped in OpenSSL 1.0.1 on March 14, 2012; see “Heartbleed,” Wikipedia, https://en.wikipedia.org/wiki/Heartbleed. ↩
Steve Marquess, “Of Money, Responsibility, and Pride,” Speeds and Feeds (blog), April 12, 2014, http://veridicalsystems.com/blog/of-money-responsibility-and-pride/index.html. Marquess, the OpenSSL Software Foundation’s “money guy,” wrote that the project had one full-time team member and should have “at least a half dozen.” Wikipedia’s Heartbleed article describes OpenSSL at the time as “maintained by a handful of volunteers, only one of whom worked full time,” https://en.wikipedia.org/wiki/Heartbleed. ↩
Jack Cable and Aeva Black, “Lessons from XZ Utils: Achieving a More Sustainable Open Source Ecosystem,” CISA blog, April 2024, https://www.cisa.gov/news-events/news/lessons-xz-utils-achieving-more-sustainable-open-source-ecosystem. The backdoor is tracked as CVE-2024-3094; see National Vulnerability Database, “CVE-2024-3094 Detail,” NIST, published March 29, 2024, https://nvd.nist.gov/vuln/detail/CVE-2024-3094. The Tukaani project describes XZ Utils as an implementation of the .xz format whose components include liblzma, a compression library, and the xz command-line tool, https://tukaani.org/xz/. CISA calls the compromise “a multi-year effort by a malicious threat actor to gain the trust of the package’s maintainer” and ties it to “maintainer burnout.” ↩
Jono Bacon, The Art of Community: Building the New Age of Participation (Sebastopol, CA: O’Reilly, 2009), https://www.jonobacon.com/books/artofcommunity (second edition PDF: https://www.jonobacon.com/wp-content/uploads/2019/01/jonobacon-art_of_community_second_edition.pdf); Jono Bacon, People Powered: How Communities Can Supercharge Your Business, Brand, and Teams (HarperCollins Leadership, November 2019), https://www.jonobacon.com/books/peoplepowered. The front matter of The Art of Community describes it as based on Bacon’s experience as Ubuntu community manager, and its contents include chapters on governance (with an “Ubuntu Governance Example”) and on handling conflict. ↩
Clay Shirky, Here Comes Everybody: The Power of Organizing Without Organizations (New York: Penguin Press, 2008), https://www.penguinrandomhouse.com/books/300615/here-comes-everybody-by-clay-shirky. ↩
Eric S. Raymond, “Homesteading the Noosphere,” catb.org, version 3.0, http://www.catb.org/~esr/writings/homesteading/homesteading; also chapter 3 of The Cathedral and the Bazaar (O’Reilly, 1999; rev. ed. 2001). ↩
The Linux Foundation, “Never Let a Good Crisis Go to Waste: Core Infrastructure Initiative,” Linux Foundation blog, April 29, 2014, https://www.linuxfoundation.org/blog/blog/never-let-a-good-crisis-go-to-waste-core-infrastructure-initiative; “Core Infrastructure Initiative,” Wikipedia, https://en.wikipedia.org/wiki/Core_Infrastructure_Initiative (announced April 24, 2014; later superseded by the OpenSSF). ↩
Cable and Black, “Lessons from XZ Utils,” CISA, April 2024 (note 15). The post says “the burden of security shouldn’t fall on an individual open source maintainer.” ↩
Tidelift, “Top 5 findings from the 2024 maintainer survey,” Tidelift, 2024, https://assets-eu-01.kc-usercontent.com/ef593040-b591-0198-9506-ed88b30bc023/23071823-ff61-4de9-87ca-e8a67c533c56/survey-one-pager-24-.pdf. The survey found 60 percent of maintainers are not paid for their work, unchanged from 2023. ↩
From But Is It Open? by Timothy O’Brien (Discursive, 2026). © 2026 Timothy O’Brien. All rights reserved.