Design: namespace and standardization path

[lann]

The package is lann:component-test today. The wasi-test prototype squats on the wasi: namespace, which a real project cannot do — entering the WASI phase process (or WebAssembly/component-model adjacent standardization) is the alternative to keeping a vendor namespace permanently.

This decision matters earlier than it looks: every suite bakes the package name and version into its bindings, so a rename after external consumers exist is a breaking migration for all of them. Related decisions that travel with it:

• Versioning policy pre-1.0: while there are no external consumers, shape regrets are fixed in place, not designed around additively; the change that publishes the package for consumption should say so.
• Relationship to lann/wasi-test: fold the prototype in (archive it, point here) or keep it as the upstream-proposal sketch.
• Whether outcome's closed-variant commitment (see README) should be recorded as a formal evolution rule in the package docs.

[lann]

Status: the urgent half resolved itself — the package renamed to polymorph:test (289a8fd, b80c13b) while consumers were still all in-family, exactly the window this issue said mattered, and the pre-1.0 policy is now documented (frozen WIT, pinned-rev-or-nothing consumption). The closed-variant rule lives in the README design commitments. Still open, and now the whole issue: whether a standardization path (WASI phases or component-model-adjacent) is ever pursued versus keeping the vendor namespace permanently, and the disposition of lann/wasi-test (archive-and-point vs upstream-proposal sketch).

[lannbot]

Status update — several of the travelling decisions have since been made and shipped; recording them here so the issue reflects what actually remains.

Settled:

• The package is polymorph:test@0.1.0 (the lann:component-test name in the body is historical); polymorph: is the vendor namespace in use across the family.
• Versioning policy: adopted and published — everything stays 0.x/unstable, but releases are caret-honest (within a minor line backward-compatible; breaking bumps the minor). v0.1.0 was the publishing moment the second bullet asked for, and v0.1.1 already exercised the within-minor contract (JS-only patch; consumers' cargo pins stayed at the v0.1.0 commit). README states the policy.
• Publication for consumption exists: the crates are consumed as cargo git pins at release commits, and the JS runner core is @polymorph/test on JSR (jsr: publish @polymorph/test (OIDC workflow); Apache-2.0 across the repo #96, browser-driver: layout-agnostic self-mount; 0.1.1 #97, retire the npm-git consumption path: the root facade and the setup action's js-locks #98) — so the "no external consumers" fix-in-place latitude is narrowing as designed.

Still open (the remaining substance of this issue):

• The standardization path itself: vendor namespace permanently vs the WASI phase process / component-model-adjacent standardization.
• The lann/wasi-test relationship (fold in vs upstream-proposal sketch).
• Recording outcome's closed-variant commitment as a formal evolution rule in the package docs.