Canonicalization lab

An executable design model of the non-normative dcp-jcs-v2 candidate, with its corpus generator and runner. It is not an SDK, a verifier or a conformance check: it computes SHA-256 for its synthetic commitments, but it signs nothing and verifies no signature, and dcp-jcs-v1 remains the only normative and default canonicalization profile (the SDKs' candidate APIs are experimental and opt-in, see CI). The design, the rules and what the model does and does not show are in docs/proposals/canonicalization-candidate/README.md.

File	Role
model.py	The design model (imported only by run.py). Its executable behavior is the approved model's; docstrings clarify public paths and the assumed last-duplicate rule
generate_corpus.py	Writes docs/proposals/canonicalization-candidate/corpus.json; expectations are written by hand and it never imports the model
run.py	Integrity, freshness, validity and coverage checks, the 274 cases, and 24 negative controls

Python 3.9 or later on Linux or macOS (Windows is not supported), standard library only. The commands run from the root of a checkout; from any other directory, give the scripts their absolute paths. Outputs go to a new temporary directory; the runner also refuses an existing output path.

out=$(mktemp -d)
python3 -I tools/canonicalization-lab/generate_corpus.py "$out/corpus.json"
cmp "$out/corpus.json" docs/proposals/canonicalization-candidate/corpus.json
python3 tools/canonicalization-lab/run.py --out "$out/run.json"
python3 tools/canonicalization-lab/run.py --controls --out "$out/controls.json"

CI

.github/workflows/canonicalization-lab.yml runs the same commands on Ubuntu with Python 3.9 (the documented minimum) and 3.12, standard library only, on pushes to main and pull requests that change the workflow, this directory or docs/proposals/canonicalization-candidate/. Each job writes to a new temporary directory, regenerates the corpus and compares it byte for byte, runs the 274 cases and the 24 negative controls, and checks both counts; any failure fails the job.

It checks the design model only. It is not SDK conformance: the SDKs' opt-in candidate APIs are tested by their own suites and by the phase 4 matrix; no profile and no default verifier runs the candidate, and dcp-jcs-v1 does not change. A green run says the model, the corpus and its expectations agree; it says nothing about any implementation.

The TypeScript and WASM opt-in APIs are available in @dcp-ai/sdk 2.2.0 and @dcp-ai/wasm 2.1.0; the other SDKs have separate releases.