Everybody wants Julia to be successful and for its developers to be successful, and I’m sure the JC team are well-intended and capable technologists. But for me, my concerns are mainly not a matter of trusting those individuals. When investing in a foundational technology, I don’t want to rely on a single company. Incentives stemming from Julia Computing’s $28.6M investor obligations are only a small part of the risk—any single company would have some incentive structure that is not perfectly aligned with the broader community’s.
It is useful to consider two kinds of projects: those that are principally controlled by a single company, and those with distributed control.
Python is managed by the Python Steering Council, chosen by election from the core developers, and the nonprofit Python Software Foundation (PSF). It has been led by a distributed group of core contributors including van Rossum for a long time. “Python” and “PyCon” are trademarks of the PSF.
PostgreSQL has an “unwritten rule that the Postgres core team should not have over half of its members from a single company” and quickly expanded the Core Team after an acquisition merged large employers in order to resolve the issue.
Rust Core Team member Steve Klabnik wrote recently in a thread beginning “I refuse to let Amazon define Rust”:
In the beginning, Rust did have one sole patron: Mozilla. Everyone was uncomfortable with that arrangement, including Mozilla.
We spent years trying to get away from this situation. It had tons of negative effects.
Go and Swift each are essentially controlled by a single company. The company’s incentives drive the direction of the language and its ecosystem. Swift is regarded as an excellent language but has little adoption outside Apple-product developers.
Looking at Julia’s GitHub contributors of the last two years, I count 6 of the top 10 are employees of Julia Computing. The ecosystem packages are also largely supported by Julia Computing. The “Julia” name and logo are owned by Julia Computing and Stefan, respectively. In 2017, there was discussion about transferring the IP to a foundation but I haven’t seen movement on that. On the other hand, the JuliaCon proceedings committee isn’t a Julia Computing team.
I know Julia Computing’s contributions to the language and its ecosystem are massive, and the largest contributors perhaps unavoidably have the largest influence on the language. I don’t know in what ways the concentration of control and funding risk harming the language ecosystem as a whole, or how these risks can be avoided. But it’s worth looking for ways to mitigate such problems before they arise. I hope these discussions continue until that risk fades away.