OPEN SOURCE
WON’T SAVE US.

Free software dates to 1983. Dr. Stallman announced GNU that September, the Free Software Foundation followed in 1985, and the GPL turned the four freedoms into something you could enforce in court. By the time anyone said “open source,” the movement it renamed had fifteen years of code behind it: Emacs, GCC, most of a working operating system.

And the talking mattered as much as the code. In 1991 a Finnish student who’d been steeped in the GNU world wrote a kernel for fun, licensed it under Dr. Stallman’s GPL, and dropped it into the fifteen years of GNU userland already waiting for it — that’s Linus Torvalds, and that’s why Linux became an operating system instead of a hobby tarball. In 1997 Werner Koch heard Dr. Stallman speak in Germany, urging Europeans to build a free replacement for PGP that US crypto export law couldn’t touch; Koch went home and wrote GPG, the tool journalists and dissidents still stake their lives on. No inspiring, no Linux, no GPG. The preacher was load-bearing.

Open source arrived in 1998, and it arrived as a pitch. Part of the community had decided that the word freedom was scaring the customers, and they were candid about the fix:

Hide the ethics, sell the method. It worked better than anyone expected. Millions of people now use and write free software under a different name, and most of them have never once heard the original argument for it. As Dr. Stallman puts it, most discussion of open source “pays no attention to right and wrong, only to popularity and success.”

Free means freedom, not price. Free speech, not free beer. A program is free software when its license respects four freedoms, and every one of them belongs to the person using the program. They read like a bill of rights because that’s what they are — enumerated, unalienable, and infringed just as often:

The GPL was written to guarantee these “for all users of a program,” including someone three distributors downstream from the author. Copyleft is just freedom with an enforcement mechanism attached. Dr. Stallman calls the freedoms essential “not just for the individual users’ sake, but for society as a whole,” and here’s the thing: the other camp never actually disagreed with any of that. They just stopped bringing it up.

The FSF and the OSI both spend their days talking about licenses, which makes them easy to mix up. Follow the money instead. The OSI runs on corporate sponsorship, and its sponsors are some of the largest companies on earth. The FSF runs on membership dues from individuals.

Microsoft and Google do not sponsor things out of sentiment. Money at that scale steers, whatever the org chart says. The FSF is structurally incapable of being bought the same way: there is no corporate membership to buy. A company can donate, within limits, and what the donation purchases is nothing. One of these organizations exists to defend you from corporations. The other one sends corporations a welcome basket.

Here is the structural problem, straight from the essay: almost all the items in the Open Source Definition are “formulated as conditions on the software’s source license rather than on what users are free to do.” Ten criteria about how code may be distributed. Nothing about what you may do with it once you have it.

Try it yourself. The document is short:

The word “use” appears once, in a rule about fields of endeavor: a license may not forbid running the program in, say, a genetics lab or a business. Useful, but that’s a restriction on restrictions, not a right handed to you. Nowhere does the definition actually say you may run the program. The FSF numbered that freedom zero because everything else depends on it. The OSD skips it because the OSD is a document about licenses, written for the people who ship them. Users weren’t the audience.

If the OSD really captured freedom, no license could clear OSI review and still fail the FSF’s. Several have. Everything below is officially, certifiably open source. None of it is free:

These four are the famous exhibits, not the full docket. The FSF’s license list records more OSI-approved licenses it classifies as nonfree, and plenty of freedom-hostile fine print this page never gets to. The names change; the finding doesn’t. None of these documents were written to protect your freedom.

Restrictions also stack on top of licenses, where no definition is looking at all. Dr. Stallman points at the Rust compiler: possibly nonfree in practice, not because of its license but because of trademark terms that forbid selling copies or shipping modified versions under the name. So even a fully free license doesn’t settle the question, because the checkbox never looks past the license.

“Free software” is ambiguous — free as in what? — but one slogan fixes it, and once you’ve heard “free speech, not free beer” you don’t get it wrong again. “Open source” has the worse problem: its plain-English reading is simply incorrect. To most people it means “you can look at the source code,” which describes source-available software, a much weaker thing. No slogan can talk the public out of the obvious reading. Watch it fail in the wild:

And the word kept drifting. Open government, open education, open science, open anything — by now it often means no more than “participatory,” and in Dr. Stallman’s words, “at worst, it has become a vacuous buzzword.” A movement can survive being attacked, but it’s much harder to survive becoming a buzzword.

Say the license is fine and the source really is up there, mirrored and tarballed. The device in your hand still checks cryptographic signatures and refuses to boot anything you compile from it. The industry name for this is tivoization. Dr. Stallman’s is blunter:

This isn’t an edge case. Android is the best-selling computing platform in history, its kernel is GPLv2, and most Android phones ship exactly this way. Open source on the label, tyrant in the box — and by the plain meaning of “open source,” nothing has even gone wrong.

GPLv3 exists specifically to forbid this. The FSF watched the trap close and rebuilt the license. The open source world’s response, for the most part, was to stay on GPLv2 and keep shipping. When the philosophy is whatever works, tyrants are what works.

Jefferson meant tyranny over human minds. Your phone has exactly one mind and a padlock on it. Different century, same design flaw.

The open source case is that openness makes software more powerful and more reliable. That invites a question the framework has no way to process: powerful for whom?

Spying, restrictions, back doors, imposed upgrades. Dr. Stallman lists them as standard proprietary features that “some open source supporters want to implement in open source programs.” Open-source DRM has been earnestly proposed: publish the source of the thing that restricts you, let the community harden it, then ship it in a device you can’t change. Open source, as a methodology, files no objection to any of this. There’s nothing in the framework to object with.

Open-source spyware is still spyware. You just get to read exactly how you’re being watched.

Why did the split happen in the first place? Because ethics made the sales calls awkward. “Raising ethical issues such as freedom… is asking people to think about things they might prefer to ignore,” and executives don’t buy discomfort. So the leaders of open source stopped raising the issues. Here is what that training produces when genuinely excellent proprietary software shows up:

The rebrand delivered users by the millions, and it delivered them unarmed. Open source “brings many people into our community, but does not teach them to defend it.” Sooner or later each of those users gets a compelling offer to go proprietary again — better features, a free tier, a discount — and nothing in the open source worldview supplies a reason to say no. Where loving freedom looks like an eccentricity, the discount wins.

When in the course of computing events it becomes necessary for a community to dissolve the vocabulary which has connected it with a sponsor page, a decent respect to the opinions of mankind requires that it should declare the causes. Better still: prove them.

“Free” and “open” compete for the same slot in your head, and only one of them teaches you anything while it sits there. “Any activity that promotes the word ‘open’ tends to extend the curtain that hides the ideas of the free software movement.” Every sponsored conference and every corporate open-source initiative pulls that curtain a little further across.

The companies funding all of it are not confused. Open is a methodology, a supply chain, a hiring pipeline, a compliance checkbox. All of that is fundable. Freedom is a demand, and nobody funds demands against themselves. Nobody managed to capture the movement’s values, so they captured its vocabulary instead, and that turned out to be enough. Neutral terms won’t undo it either; on FOSS and FLOSS, Dr. Stallman is curt: “if you want to stand up for freedom, using a neutral term isn’t the way.”

So separate. Not out of purity — out of arithmetic. The vocabulary is taken, the definition has sponsors, and the organization holding it answers to them, not to you.

You could keep arguing about this forever. Or you could notice that it stopped being an argument a while ago, because it’s a theorem.

None of this is an assault on open source folks. Most of them are the community: maintaining the packages, reviewing the patches, answering issues at eleven at night for strangers. What’s being called out is a continual pattern of bad practice at the top of the institution, not the people doing the work under its name.

Take the open source program offices at universities. A lot of them do genuinely good work — prying research code out of desk drawers, teaching students to contribute upstream, cleaning up license messes so that public money actually produces public code. But notice what the name is for. “Open Source Program Office” is a phrase built for a provost’s budget spreadsheet. It’s bureaucracy, not philosophy. What the good ones practice day to day looks a lot like free software; the label mostly came stapled to the funding paperwork.

The institution is a different story, and the difference is structural. No organization that lets the world’s largest companies steer it has your best interest at heart. It can’t. Not because its staff are bad people, but because incentives don’t care. The OSI runs on corporate sponsorship, with Microsoft, Google, Amazon and Meta among the names on the checks. The FSF takes no corporate members at all — companies may donate within limits, and the donation buys no vote, no seat, no leverage over the organization. Its money is membership dues from individuals. People like you and me, forty dollars at a time. One of those funding models can be fired by its base. The other one is its base.

Free software came first, in 1983, and it will still be here last, because it’s the better-defined movement: four freedoms you can test any license against in an afternoon, versus a brand whose own advocates spend half their time explaining what it doesn’t mean. A definition with teeth ages well. One with sponsors gets renegotiated whenever the sponsors need it to.