Codename legend

Reference material, synced from the compiler repo’s spec/. The guide is the gentler path in.

The spec and code comments carry a few families of short internal tags from Hale’s development history. Most are navigation breadcrumbs, not part of the language — you can ignore them and read the surrounding prose as current behavior. This page says which is which.

F.1, F.22, F.32-1γ-v2, … Each names one locked design commitment: what a construct commits the design to, what was considered and rejected, and whether it’s shipped, superseded, or still a sketch. The full records live in the decision log, decisions.md. A reference like “§ F.31” is a link into that log — not something you need to know to use the language.

m26, m82, m105, … Internal build milestones from the development history (roughly: a unit of implementation work). They appear in code comments and occasional spec asides as provenance — “shipped in m27,” “the m105 adapter-inbound path.” They are not language surface: the behavior each describes is specified in the relevant spec file, and the ship order is in ../CHANGELOG.md. Safe to skip. (A few carry an inline gloss where it aids reading — e.g. m26/m27 cooperative-then-pinned schedulers, m82 “locus all the way down,” m96 the tree-sitter grammar.)

WS1, WS3, v1.x-FORM-4, v1.x-WINDOWED, … Names for larger threads of work. Same status as milestones: history/provenance, not language surface. Ignorable.

Some short symbols look like codes but are real, current parts of the language, defined normatively in the spec — keep them in mind, don’t skip them:

• k_max and the accept-budget formula k_max = B / [(1 − φ)·c + φ·σ] — the bound on how many children a locus may accept. Defined in types.md; the ASCII source spellings B, c, sigma, phi are reserved identifiers (see tokens.md).
• bulk / harmonic / resolution — the three execution modes (real keywords, not codenames).

Rule of thumb: an F./m/WS/v1.x- tag is history you can ignore; everything else in the spec is the current contract.