@snowfox @whitequark

The problem (which @whitequark framed fairly succinctly as ‘calling the cops on the cops doesn’t work’) is that you are trying to oppose an entrenched power structure (big corporations) by relying on an entrenched power structure (the legal system).

The rule of thumb in companies for enforcing IP laws is to budget about a million dollars per enforcement action. Sometimes it’s cheaper, sometimes it’s more expensive. Generally, it’s much cheaper if you have more money because no one really wants to be in court when the other side is able to spend ten or a hundred times as much as you can to hire the world’s best barristers in this specific field. So simply having a huge pile of money biases the system toward you, and that’s the system that you’re using as your weapon against trillion dollar companies.

In theory, a well-funded foundation can help balance this. In practice, it doesn’t often work. I contributed to a GNU project where we had clear evidence of a startup taking our LGPL’d code and selling it in a proprietary product without releasing the changes. The FSF decided it was too expensive to sue. Eventually the startup was bought and the founders made a pile of money. At the same time, other small companies that wanted to play by the rules avoided us because the LGPLv3 libraries and GPLv3 tools were considered too high risk by their lawyers.

GPLv3 and AGPLv3 especially are increasingly used by corporate open source that wants to retain control over their ecosystem. And they’re chosen because that’s what they are very good at. If I contribute to an MIT-licensed project, I have almost the same rights on my contributed code as I would have if I had written it in isolation. With a copyleft license, I have fewer rights. This means that it’s easy for the centre of mass of a project to exert complete control. It is not a coincidence that the open-source kernel that has the greatest amount of corporate control is GPL’d: the license encourages centralisation of power.

If you look at the top contributions of GCC, they are IBM (under their ‘Red Hat’ brand), Qualcomm, and SUSE. Three companies make up over 50% of contributions to GCC in the last year. In LLVM, the top three corporate contributors are only 34% of the total, the next largest aggregate is individual contributors, and then no other corporation is responsible for more than 5% of the contributions in the last year.

A company that can afford to spend millions on lawyers will find loopholes. AGPLv3 forces you to distribute your code. But you can still provide open extensions that are wrappers around some RPC to a proprietary service. Or, if you’re a company that controls a project, you can require a CLA and provide ‘enterprise-only’ features that are proprietary and sell an open-core system and put all competitors at a disadvantage because they can’t tightly couple their extensions or the core, but you can.

The GPL family is the license family of corporate control. The ‘viral’ nature requires strong coupling to propagate. This encourages the kind of software-development model that benefits corporations.

If you want to avoid corporate dominance, make it easy for end users to extend your programs. Make it easy for end users to share their extensions. Intentionally create ecosystems where the original creators do not control the wider ecosystem. That makes it very hard for a corporation to exert control because what would they even try to control? If they hire the core developers, they get control over a small fraction of the ecosystem and the rest ignores them if they go in a silly direction, because every component is trivial to fork and trivial for someone else to maintain.

Require cooperation not through complex legalese (which adds a barrier to entry for anyone who doesn’t have a legal team on staff) but by creating ecosystems where everything has multiple interchangeable options.

Don’t build the big AGPL’d monoliths that are prime targets for corporate control.