D60 broader computational validation

Date: 7 October 2026. Owner: Moira engine, VED-003. Outcome: broader computational validation of pvr_textbook_linear is complete for the declared mapping and interfaces. Classical continuous-degree attribution remains unestablished. These are separate acceptance questions.

The source standard governs the forward natal-sign sequence, proportional within-part degree, half-open endpoints and separate selected-edition deity names. The earlier receipt and literature adjudication retain their historical scope. This pass adds executed external software corroboration, reproducible numerical evidence and direct interface comparisons.

Independent implementation evidence

Comparator	Pinned identity	Actual execution and limits
PyJHora D60 source	Commit 48e57d29b47a3143519910a24866758116467485, 6 August 2026; commit identifies V4.9.3. charts.py SHA-256 c36c214e59c6a095438e67dacc29c7622dc8917ec25c723389c42655035ca86b.	Execute unchanged shashtyamsa_chart function AST and selected constant/enum AST, with explicit method 1, in an isolated Python child. It returns full sign/degree pairs. No Moira or Swiss import occurs in this oracle process. This is not P.V.R. Rao's closed-source Jagannatha Hora, nor its current recommendation. The software's method label does not establish classical attribution.
Maitreya D60 source	Commit dc468ded92798638f36a318838fc34eaf8623640, 6 March 2026. Varga.cpp SHA-256 f081e3321e8051ddaffe709c5e558151a71b29721b64e335b222de8cb328ee95.	Compile its exact D60 switch branch, sign projection and circular helpers as C++, using installed MSVC 14.50.35717, /fp:precise /Od. The harness represents Rasi by its integer index. Record the original returned sign and instrument its reduced intermediate longitude. The intermediate is not a public Maitreya full-position product. This is source-fragment execution, not a complete application build.

Neither computation was rewritten in Python to imitate outputs. No external implementation enters Moira runtime. scripts/d60_cross_engine_sources.json pins selected source URLs, file hashes and repository license files. scripts/validate_d60_cross_engine.py acquires and executes them in an explicit private workspace. The repository contains Moira-owned orchestration and numerical outputs; external source and binaries remain private. PyJHora's repository includes AGPLv3 licensing and Maitreya GPLv2-or-later licensing; their implementation text is not vendored into Moira.

tests/artifacts/oracle/d60_cross_engine_2026-10-07.json is secondary cross-engine evidence, not primary classical or astronomical authority. Refresh is explicit: reproduce pinned sources, inspect every disagreement, then review an intentional fixture change. Failed comparisons preserve diagnostics and prevent refresh. A new pin, tolerance or interpretation needs renewed evidence adjudication; fixtures are not adjusted to fit an implementation.

Corpus, precision and results

Cohort	Inputs	PyJHora disagreements	Maitreya disagreements
Every half-degree segment, at offsets 0.125, 0.25 and 0.375 degrees	2,160	0	0
Every segment start, exact and adjacent representable floats on both sides, including zodiac wrap	2,160	0	0
Uniform binary64 longitudes, fixed random seed 600017	20,000	0	0
Seven classical planets, eight UTC dates from 1800 through 2400, Lahiri and Raman frames	112	0	0
Total	24,432	0	0

Every input also agrees with independent Fraction evaluation of the conditioned binary64 input: zero Moira/rational disagreements. The sign and segment use rational division; proportional degree is evaluated exactly before binary64 conversion. Moira retains mapped-longitude containment at excluded upper edges. Sign agreement is exact; there is no tolerated sign error or epsilon snap region.

Before comparison, the external numeric budget was declared as 8 * ulp(2160.0) = 3.637978807091713e-12 degrees, based on maximum intermediate arithmetic scale. This is a new secondary-comparator budget, not a widened runtime, source-minute, existing rational or golden tolerance. Maximum observed degree residuals: 1.1368683772161603e-13 degrees for PyJHora, 2.2737367544323206e-13 for Maitreya. All exactly representable interior samples agree exactly. No boundary mismatch was discarded. These are floating-point mapping residuals, not astronomical accuracy claims.

The planetary D1 values originate from Moira's installed DE441 reader. Both comparators independently map those exact sidereal inputs. The sixteen date/frame charts cover all twelve natal signs and both parities. This proves conditioned D1-to-D60 computation and transport, not independent ephemeris, ayanamsa, birth-time or predictive accuracy. Datetime, frame and body are recorded without inferring a publication's hidden seconds or rounding rule.

The offline fixture retains 4,552 cases: all 4,320 deterministic interior and boundary inputs, the first 120 seeded samples (selected by order, not residual), and every planetary input. The full 24,432-input capture, raw outputs, build receipt and disagreement ledger remain under C:\dev\outputs\vedic-d60-cross-engine-2026-10-07.

Interface and regression verification

tests/unit/test_d60_cross_engine.py compares engine positions directly with both frozen implementations, grouped by natal sign and cohort. It checks named/Shodashvarga facade outputs and the D60 sign used by Vimshopaka in both D60-bearing groups. This does not independently validate the entire strength doctrine from position witnesses.

tests/server/test_d60_cross_engine.py compares all seven full-point request shapes with external expected coordinates over twelve natal signs; tests 120-result named/Shodashvarga batches; reproduces all sixteen real-reader chart batches against frozen planetary inputs/outputs; and checks both strength groups' transported D60 signs. Expected D60 coordinates come from external captures, not another call to Moira.

Environment: C:\dev\moira, main, starting HEAD 67b66de14cde0447730b0c9b57e5e14c1edd9935, version 6.9.9; project .venv, Python 3.14.3. Planetary resource resolved through its owning path module to C:\Users\nilad\.moira\kernels\de441.bsp; pytest content-probes DE441. Varga has no admitted native counterpart. This follow-up changes validation tooling/fixtures/tests and documentation. Existing Chara/D60 dirty work is preserved. Runtime/API semantics, defaults, harness, known-issue exemptions, native extension and generated moira.wiki are unchanged.

Documentation reconciliation: the source standard and two research records now link this completed computational evidence; the earlier validation receipt preserves its narrower checkpoint and links forward. The Vedic work register distinguishes completed modern validation from historical source admission. Home adds discovery, and the unreleased changelog records this added coverage.

Pytest uses MOIRA_TEST_MODE=1, MOIRA_STRICT_KNOWN_ISSUES=1, MOIRA_NO_DOWNLOAD=1, and -m 'not external_network'. Server tests admit loopback; frozen evidence tests use no external network. The first new-coverage run passed 205 tests, with 19 successful DE441 resource uses and no skips.

.\.venv\Scripts\python.exe -m pytest tests/unit/test_d60_cross_engine.py tests/server/test_d60_cross_engine.py -q -m 'not external_network' --junitxml=C:/dev/outputs/vedic-d60-cross-engine-2026-10-07/new_tests.xml
.\.venv\Scripts\python.exe -m pytest tests/unit/test_d60_cross_engine.py tests/server/test_d60_cross_engine.py tests/unit/test_d60_full_position.py tests/unit/test_d60_published_evidence.py tests/unit/test_vedic_chara_d60_admission.py tests/server/test_d60_full_position.py tests/server/test_vedic_chara_d60_admission.py tests/unit/test_varga.py tests/unit/test_shodashvarga.py tests/unit/test_d60_deities.py tests/unit/test_vedic_facade.py tests/server/test_server_varga_routes.py -q -m 'not external_network' --junitxml=C:/dev/outputs/vedic-d60-cross-engine-2026-10-07/regression.xml
.\.venv\Scripts\python.exe -m pytest tests/server/test_server_vedic_phase2_routes.py -k vimshopaka -q -m 'not external_network' --junitxml=C:/dev/outputs/vedic-d60-cross-engine-2026-10-07/strength_regression.xml
.\.venv\Scripts\python.exe scripts/validate_d60_cross_engine.py --work-dir C:/dev/outputs/vedic-d60-cross-engine-2026-10-07 --check-fixture tests/artifacts/oracle/d60_cross_engine_2026-10-07.json

The validator reproduced the frozen fixture exactly using checksum-verified cached sources, without acquisition. A fresh private workspace requires --acquire explicitly; kernel downloads remain disabled. Source-fragment reproduction requires installed Windows MSVC. Ordinary offline tests need neither comparator source nor a C++ compiler. Execution/document outcomes are captured in private verification.json. This is focused acceptance, not the whole repository suite or full-application software parity.

Final affected acceptance/regression runs: 1,399 + 3 = 1,402 distinct tests passed, zero failures/errors/skips. They re-execute the earlier 1,197-node checkpoint and the 205 new nodes; the initial 205-test run is not added again. The final runs recorded 25 + 6 successful DE441 resource uses, with no skips or failures and content-probed identity. New script/test Ruff, documentation consistency, generated REST inventory, whitespace and eight-document in-memory owning wiki render/link checks pass. The checked-in fixture reproduces exactly. The lineage audit confirms no external implementation was copied into runtime and no Moira output was used as an external expected D60 coordinate. Link checks found no new missing targets; older changelog file:// links remain pre-existing formatting debt recorded in the private receipt.

Acceptance decision

Question	State
Modern mapping, partition semantics, broad external corroboration and engine/facade/REST/strength projection	Complete under the declared policy, within the corpus and arithmetic budget. No remaining engineering validation blocker identified.
Classical continuous-degree attribution	Unestablished in reviewed sources. Admit no additional classical full-point method without an identified passage and explicit semantics.
Exact source-published planetary D1/full-D60 pairs	Not established by the literature review. Separate from newly captured software outputs: an additional source-evidence frontier, not unfinished conditioned-computation validation.
Current PVR Jagannatha Hora equivalence/recommendation	Not claimed. PyJHora is another project; Maitreya intermediate evidence is expressly scoped.
Predictive effectiveness	Not tested. Arithmetic validation does not adjudicate interpretation.

VED-003 retains bounded historical source admission, while its modern implementation and broader computational validation are locally complete. Secondary software outputs are not relabeled as primary classical evidence. The package remains uncommitted/unpushed; release/deployment are separate.