Skip to content

Governance

Fid is early, and its governance is deliberately simple. This page says how decisions get made today, honestly, rather than describing a foundation that doesn’t exist yet.

Who decides

Fid is designed and built at Pledge & Grow. It has one maintainer today, Mehdi Berel. Language changes (syntax, semantics, the safety systems) are decided by the maintainer, in the open: design debates, rejected directions, and known limitations are documented in the compiler repository’s docs, not hidden.

As the community grows, governance will formalize. The commitment now is transparency: what’s decided, what’s still open, and why.

Proposing changes

There is no heavyweight RFC process yet, and pretending otherwise would be theater. The path today:

  1. Open a GitHub Discussion describing the problem before the solution: what can’t you express, what surprised you, what broke?
  2. Small fixes (docs, error messages, tests) can go straight to an issue or pull request.
  3. Language-level proposals get discussed in the open; accepted directions land on the roadmap with a phase and status.

Proposals are weighed against these design principles: safe by default, no hidden control flow, no macros, one keyword per concept, zero-cost guarantees, and honesty about what’s implemented versus planned.

Versioning

The Fid v1.0.0 stable release is planned for December. The self-hosted compiler is the primary development path, and the v1.0 syntax is frozen. Going forward:

  • Breaking changes are announced in the changelog with what changed and how to move.
  • The feature matrix stays truthful. Feature Status and the repository’s features.md distinguish working, partial, and planned: when marketing and reality disagree, reality wins.
  • Syntax migrations are mechanical and documented: the v1.0 keyword migration was published the same day the change landed.

Releases and platform support

  • Channels. Tagged releases (v1.x.y) on GitHub Releases are the supported channel; main is the development branch and may break between tags. There is no nightly channel yet.
  • Cadence. Minor releases follow the release roadmap (1.1 semantic depth, 1.2 ergonomics, 1.3 hardening, 2.0 dogfooded); patch releases ship when a wrong-code or security fix lands. Each release gets a changelog entry.
  • Compatibility. The v1.0 syntax is frozen for the 1.x series. A program that compiles on 1.0 is meant to compile on every 1.x; where the 1.1 semantic fixes change the meaning of code that was previously silently wrong, the changelog says so explicitly.
  • Platform tiers. Tier 1 (supported, the whole suite runs on it): Windows x64 (interpreter, C, native) and Linux x64 (interpreter, C). Tier 2 (builds, partly tested): native Linux executables for x86-64, ARM64 and RISC-V, Linux ARM64 and RISC-V through a C cross-compiler, macOS through C (built, not run on a Mac), WebAssembly (WASI) through C. Tier 3 (present, experimental, not supported): native macOS and WebAssembly output, bare metal. A platform moves up a tier when its tests pass in CI. Only Windows has a prebuilt installer today.
  • Security. Fixes go to main and the latest release; see the Security Policy.

The stability promise that exists today

One thing is already stable: published URLs and published claims. Changelog entries are never rewritten to look better in hindsight, and documented limitations stay documented until they’re actually fixed.