Stop claiming weaver-spec.dev schemas are live until canonical hosting actually resolves

[dgenio]

Problem

ContextWeaver's README / docs/weaver_spec_mapping.md currently say that the schemas fetched from raw.githubusercontent.com/dgenio/weaver-spec/main/contracts/json/ are also published at:

https://weaver-spec.dev/contracts/v0/

But the canonical weaver-spec hosting policy currently says the domain is not serving the schemas at a stable HTTPS endpoint, and a fresh DNS lookup from the adoption audit could not resolve weaver-spec.dev.

This makes the public/PyPI adoption documentation overclaim infrastructure that is not currently available.

Upstream tracking:

• P0: Serve canonical JSON Schema $id URLs at immutable HTTPS endpoints weaver-spec#213 — implement canonical live $id hosting
• dgenio/weaver-spec docs/SCHEMA_HOSTING.md — current authoritative status

Progress — 2026-08-11

PR #852 corrects the smaller canonical mapping document immediately. It now states:

• CI currently fetches schemas from mutable weaver-spec/main source;
• weaver-spec.dev is the reserved canonical $id namespace, not yet a live endpoint;
• immutable conformance input is tracked by Pin the gating weaver-spec conformance check to a tagged schema version #469;
• current implementation-generated conformance evidence is tracked by Publish current weaver-spec conformance bundle at .well-known/conformance.json #846;
• live hosting is tracked upstream in weaver-spec#213.

The same stale live-host wording still exists in the very large README / generated publication surfaces. Those should be updated atomically from their canonical source/generation path rather than by a risky partial whole-file connector rewrite.

A second independent drift in docs/weaver_spec_mapping.md around ExecutionCandidate / ExecutionFeedback is tracked separately in #851; do not mix that contract-shape decision into the hosting-status cleanup.

Goal

Keep ContextWeaver docs truthful during the transition.

Preferred resolution

If upstream weaver-spec#213 lands and the URLs become live + hash-verified before the README cleanup merges, update/verify the claims against the real endpoint and close this issue.

Otherwise, temporarily change README/PyPI wording to something like:

CI currently pins/fetches schemas from the reviewed GitHub source. weaver-spec reserves https://weaver-spec.dev/contracts/v0/... as the canonical $id namespace; live hosting is tracked upstream and must not be assumed until available.

Update all generated/public surfaces derived from the source docs, including llms-full.txt / PyPI README as applicable.

Acceptance criteria

• PR docs: correct canonical weaver-spec schema-host status #852 is green/merged for the mapping document.
• README does not claim the domain is live while it is unresolved.
• Generated llms-full.txt and other derived docs are regenerated from corrected sources.
• PyPI-facing README wording is correct on the next release.
• Once upstream [routing] Namespace-aware tokenizer split for dotted tool IDs (with measured impact) #213 is live, add an automated liveness/hash check before restoring the statement that canonical schemas are published there.

Non-goals

• Changing the canonical $id namespace.
• Switching gating CI from immutable/pinned schema inputs to mutable remote main.
• Duplicating upstream hosting implementation.