Skip to content

Contributing

contributions welcome

The full text lives in CONTRIBUTING.md in the repository. The short version:

This is not a business, and it is one person’s project. There is no company, no revenue, no support rota and no service level on anything, including replies. An issue that sits unanswered is not a slight, it is one person’s attention being finite.

Scope is the main risk, not competition. The failure mode named at the top of the project is a half-finished platform abandoned at 40 percent. Every feature that gets refused is refused for that reason, not because it is a bad idea. Most of them are good ideas. That is exactly the problem.

  1. A cold start measurement on a cheap VPS. The project’s central claim is written against a machine nobody has measured. File the numbers with the hardware stated.
  2. The Linux queue producer fix. Add extra hosts to ContainerSpec and emit the flag, or resolve the gateway address into the URL at container start.
  3. A CLI verb for snapshots. The daemon half is written and tested and reachable from nothing.

Anything in the out-of-scope list: Kubernetes, clustering, multi-node, multi-tenancy across owners, a hosted tier, billing, metering, end-user auth as a service, realtime subscriptions, edge execution, DNS or secrets management, Terraform providers, Helm charts.

The list is a toll, not a wall. Write docs/decisions/NNNN-your-case.md, open it as a pull request on its own, and let it be argued before you write the implementation. The existing records show the shape.

Two agreements that will get a patch sent back

Section titled “Two agreements that will get a patch sent back”
  • No em-dashes. Anywhere. Docs, comments, commit messages, program output.
  • Ground claims in code. Cite path/to/file.ts and a symbol rather than describing what you believe the code does.

Three of the worst bugs found so far came from running a container and none from a test suite that was passing at the time. If your change touches lifecycle, readiness, wake or sleep, say in the pull request what you ran and what you saw.