Skip to content

RC2 E5 content decision for independent review — 23 October 2026

Decision requested, not granted. No GitHub tag, GitHub Release or PyPI upload is authorized by this change.

Source-specific technical evidence

The literal 1.2.0rc2 candidate was merged via PR #471 at commit 9891feee6e5daa638a9de5cada295482a5fa4884, following 33/33 exact-head checks. The main commit passed 30/30 exact-main workflow runs, zero failures.

Critically, experimental-workflow-expansion run #40002802440 finished successfully for Python 3.11, 3.12 and 3.13. Each E1–E4 job passed boundary audits, robustness tests, synthetic end-to-end exercise and building/testing an out-of-checkout candidate wheel. Those checks are software/package evidence, not statistical coverage, field validity or API stability evidence.

Decision proposed

Allow the candidate wheel and sdist to include eyetrajectoriespy.workflows.experimental and its 12 E1–E4 importable helpers, on the strict condition that they are presented as unstable experimental prerelease interfaces. This does not promote any of them into the ten supported eyetrajectoriespy.workflows route contracts or into the package root. Maintain the prior frozen root boundary of 455 stable plus three explicitly experimental exports.

These helpers offer opt-in preprocessing, quality preflight, descriptive sensitivity and evidence-based reporting. None is a validated solution to response-dependent/MNAR gaze loss or provides calibrated 90% functional-inference coverage.

The other permissible choice is to exclude the module from the release distribution and redo exact-source qualification of a changed wheel/sdist. A prose-only exclusion is insufficient because the module is importable in the candidate distribution.

Independent review required

A human reviewer must explicitly approve the proposed experimental classification and the ten-route claims boundary before a separate publication-arming PR. Specifically review:

  1. Importability and non-stability of all 12 helper names, with no root or supported-workflow promotion.
  2. Accurate scientific exclusions for open #289 and #296; retain all negative results.
  3. Provenance of the literal RC2 candidate and three independent Python-version E5 jobs.
  4. Explicitly documented version 1.2.0rc2 as a prerelease, not final stable 1.2.0.

The machine-readable E5_RC2_CONTENT_REVIEW.json and scripts/validate_e5_rc2_content_review.py enforce that the evidence is unchanged and every authorization flag remains false. These checks cannot impersonate a human reviewer. Once reviewed, the scope decision must be recorded in a separate governance transition, with fresh exact-main CI and the required pypi environment approval. The actual release.yml manual workflow_dispatch target=production is outside the connected GitHub tool's capabilities.

Review status: PENDING. Current public release: 1.1.0.