Skip to content

fix(conformance): reconcile 3 vectors with normative spec — resolve 3/8 xfails#210

Merged
dp-web4 merged 1 commit into
mainfrom
worker/web4-20260518-120015
May 18, 2026
Merged

fix(conformance): reconcile 3 vectors with normative spec — resolve 3/8 xfails#210
dp-web4 merged 1 commit into
mainfrom
worker/web4-20260518-120015

Conversation

@dp-web4
Copy link
Copy Markdown
Owner

@dp-web4 dp-web4 commented May 18, 2026

Summary

  • t3-002: Corrected T3 composite aggregate from unweighted mean (0.8167) to weighted composite (0.825) per spec §10.2 protocol-invariant weights (talent=0.4, training=0.3, temperament=0.3)
  • t3-006: Corrected talent decay expectation — talent MUST NOT decay (spec §2.3 normative invariant, §10.2 protocol-invariant). Also corrected temperament from "move toward 0.5" to "recovery" (+0.01/month per §2.3)
  • r7-rep-001: Corrected V3 behavioral dimensions from [valuation, veracity, validity] to [veracity, validity] — valuation is economic (ATP settlement, spec §3.3), not behavioral

Conformance suite: 34 passed, 5 xfailed (was 31/8). The 3 resolved xfails were vector errors where vectors contradicted normative spec language strengthened by Sprints 44/47/48. The 5 remaining xfails are genuine architectural gaps (constraint checking → PolicyGate, role authorization → governance layer, federation API shape, sub-dimension rollup, success flag semantics) requiring operator decisions.

0 new files. 0 SDK code changes. 3 files modified (2 conformance vectors, 1 test file).

Test plan

  • pytest tests/test_conformance.py -v — 34 passed, 5 xfailed (was 31/8)
  • Full SDK suite — 2749 passed, 5 xfailed
  • mypy --strict web4/ — 0 errors
  • JSON validation — both modified vector files parse correctly
  • GitNexus detect_changes — LOW risk, 0 code symbols affected

🤖 Generated with Claude Code

… of 8 xfails

Three conformance vectors contradicted the normative spec (strengthened by
Sprints 44/47/48, classified as protocol-invariant in §10.2):

- t3-002: aggregate was unweighted mean (0.8167); corrected to weighted
  composite (0.825) per §10.2 protocol-invariant weights (0.4/0.3/0.3)
- t3-006: expected talent to decay; corrected to talent_unchanged per §2.3
  normative invariant ("Talent MUST NOT decay through inactivity"). Also
  corrected temperament assertion from "move toward 0.5" to "recovery"
  (+0.01/month per §2.3)
- r7-rep-001: expected all 3 V3 dims from behavioral quality; corrected to
  veracity+validity only — valuation is economic (ATP settlement, §3.3)

Conformance suite: 34 passed, 5 xfailed (was 31 passed, 8 xfailed).
Remaining 5 xfails are genuine architectural gaps requiring operator
decisions, not vector errors. 0 new files, 0 SDK code changes.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@dp-web4
Copy link
Copy Markdown
Owner Author

dp-web4 commented May 18, 2026

APPROVED: Conformance-vector reconciliation to normative spec (web4 development phase — reusable conformance integrity, design-aligned). Verified locally on the PR branch: pytest tests/test_conformance.py → 34 passed, 5 xfailed (exactly as claimed; was 31/8). Direction is correct — vectors corrected to match spec (§2.3 talent-invariant, §10.2 weighted composite, §3.3 valuation-is-economic), NOT SDK bent to pass tests. Behavioral properties preserved; 5 remaining xfails are genuine architectural gaps. 3 files modified, 0 new files, 0 SDK changes. No drift signals.

@dp-web4 dp-web4 merged commit 3f12462 into main May 18, 2026
0 of 6 checks passed
@dp-web4 dp-web4 deleted the worker/web4-20260518-120015 branch May 18, 2026 23:10
dp-web4 added a commit that referenced this pull request May 19, 2026
Version bump and documentation update consolidating post-Sprint-53 work:
- Cross-society MCP types (PR #195/#199): 7 new exports (369->376),
  CrossSocietyContext, ReputationEnvelope, MCPContextResource, 3 enums,
  CrossSocietyError class (7th error category, 30 total codes)
- Conformance xfail resolution (PR #210): 3 of 8 xfails resolved,
  reconciled t3-002, t3-006, r7-rep-001 with normative spec (now 34/5)
- Bug fix: _pkg_version("web4") -> _pkg_version("web4-sdk") to match
  the PyPI rename from v0.27.0 (stale egg-info was returning old version)

0 new files. 9 files modified. 2749 tests (2744 passed, 5 xfailed).
mypy --strict clean. ruff lint clean.

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
dp-web4 added a commit that referenced this pull request May 19, 2026
…tion)

PR #210 (3f12462) revealed 3 of 8 documented xfails were vector-spec
staleness, not architectural gaps — vectors authored before Sprints
44/47/48 strengthened invariants. Suite now 34 passed / 5 xfailed.

Executive Summary line 41 corrected from "8 xfailed gaps documented for
the next operator architectural-decision pass" to "5 xfailed gaps awaiting
operator architectural decisions" — sharpens framing alongside the count
fix to disambiguate vector errors from operator-decision-bound gaps.

PUBLISHER_CONTEXT.md updated with 2026-05-19 entry covering #208/#209/#210/#211.
md + docs/whitepaper-web artifacts rebuilt; PDF/web not refreshed (prose
edit only, no structure change).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant