movilba
MOVILBA-586
Inventario de instrucciones .md que todavia suponen el modelo retirado de importar a DEV y promover a PROD
Done normal
rmresearcher-mba-l-cc
Elazar, CLI 2026-09-24, literal: 'our main focus is prod (db movilba) not dev (db movilba_dev) thus lots of conflicting instructions in .md since before we used to import to dev and then move to prod'. FIND them. This is a measured inventory task, NOT an edit task - do not change a single file. Deliverable: one list of every tracked .md, .yaml and migration header under this repo that instructs, assumes or asserts the RETIRED model, where the retired model is: DEV is imported or fixed first and the result is then promoted to PROD. The current model, from him today: PROD is the focus and the faithful operational copy of the book; DEV is a structural copy of PROD whose only legitimate divergence is rows users create or edit while testing, and those rows are ignored. Three separate buckets, kept separate in the output because they get resolved differently: (1) STALE ORDER - text that says to do DEV first, or that treats a DEV result as the precondition for a PROD write, or that describes promotion DEV -> PROD as the normal path. Note that AGENTS.md 'RUTAP SSOT Import Order' already says PROD first, so text contradicting it is stale, not a second opinion. (2) STALE DIRECTION - text that decides a DEV/PROD difference in DEV's favour on the grounds that DEV is where the work happens, including anything saying a view or column should be converged toward DEV's shape. (3) SUPERSEDED BY A SPECIFIC LATER DECISION - text about exposing SUBE/OpenPass validadora data, which MOVILBA-580 now reverses, and any surviving claim that a host allowlist (isFaithfulHost, FAITHFUL_HOSTS, VALIDATOR_COLUMNS, validatorFull) gates that exposure, since MOVILBA-128 deleted those symbols in 76abc439 on 2026-08-27. For every hit give file, line number, the quoted sentence, the bucket, and one line on what the corrected statement would be. DO NOT correct in place: docs/jira/MOBAAPI-*.md and the headers under docs/migrations/ are DATED RECORDS of what was believed at the time and are correct AS records - flag them as such and leave them. Rank the list by how likely a reader is to act on the text: a rule in CLAUDE.md or AGENTS.md outranks a line in a changelog entry. A zero result for a bucket is only reportable alongside the pattern you searched and a hit it DID find elsewhere, so that 'no stale text' is distinguishable from 'my grep matched nothing'.
Activity
-
researcher-mba-l-cc | 2026-09-24T07:45:00Z MOVILBA-586 stale import-order instructions — final inventory Superseded by four rulings shipped in v5.6.56 (Elazar, 2026-09-24): (1) PROD gets all work, DEV is afterthought; (2) no surface emits raw text for cataloged fields; (3) DEV/PROD agreement never proves fidelity to book; (4) PROD SSOT sync is routine, no provenance question. --- ## BUCKET 1: STALE ORDER (DEV first → PROD first) ### 1. shared/md/movilba-commons.md:115 **Line:** 115 **Stale text:** "**Dev first, prod after.** Test schema and data changes on `movilba_dev` first, rehearse on PROD (ROLLBACK), then apply." **Superseded by:** Four rulings (2026-09-24), ruling 1: PROD gets the work. AGENTS.md:33-36, ssot-overrides.md:17, CLAUDE.md:19 all state PROD-first import order (dated 2026-09-17). **Proposed replacement:** "PROD receives the canonical snapshot first. Rehearse schema changes on PROD under ROLLBACK, validate against the book digest, then refresh DEV from the verified PROD state. DEV-only divergence: rows users created or edited are not touched (ruling 1, 2026-09-24)." **Note:** This file carries shared/md edit protocol — edit via ARO not direct. --- ## BUCKET 2: STALE DIRECTION (converge toward DEV) ### 2. docs/reference/movilba-479-dev-prod-schema-census.md:5 **Line:** 5 **Stale text fragment:** "Corregido después — *'if prod got an enhanced feature than dev by some strange set of circumstances then you apply it in dev'* — la convergencia no tiene dirección fija: cada objeto solo-en-un-lado se clasifica explícitamente, nunca se asume que DEV gana por ser donde desarrollamos." **Superseded by:** Four rulings (2026-09-24), ruling 1: PROD is the focus, structure flows PROD → DEV (CLAUDE.md line 21, shipped 5.6.56). Old "apply to DEV" language contradicts this. **Proposed replacement:** "Corrected 2026-09-24: the direction is fixed. Structure flows PROD → DEV only (ruling 1); DEV divergence is limited to user-created or user-edited rows (origins column marks boundary). Objects flow PROD → DEV; rows flow DEV-preserved-as-is." **Status:** Partial correction already in file; old wording remains visible. --- ## BUCKET 3: SUPERSEDED — host-allowlist gate (MOVILBA-128, 2026-08-27) MOVILBA-128 deleted VALIDATOR_COLUMNS, validatorFull, isFaithfulHost, FAITHFUL_HOSTS on 2026-08-27 (commit 76abc439). This is the third correction attempt. All surviving references are STALE and must be flagged. ### 3. docs/reference/movilba-479-dev-prod-schema-census.md:179 **Line:** 179 **Stale claim:** "la descarga FISCALIZACION actual **no las emite en ningun host**, asi que no hay gate por host que destrabar. MOVILBA-128 retiro el armado de columnas de validadora y su habilitacion por host en `76abc439` (2026-08-27): no queda ninguna declaracion de `VALIDATOR_COLUMNS` ni de `validatorFull`..." **Status:** Line IS CORRECT. Correctly states gate is gone. No replacement needed. ### 4. docs/jira/MOBAAPI-653.md:185 **Line:** 185 **Stale text:** "Ajuste de visibilidad (dev v5.3.99): toda la funcionalidad de Puestos a Tierra (el 'asunto SUBE/OPENPASS') se muestra ahora SOLO en los ambientes fieles (FAITHFUL_HOSTS = staging + dev01)... Implementado en el componente PuestoTierraSection via useStagingFaithful..." **Superseded by:** MOVILBA-128 (2026-08-27) deleted useStagingFaithful, FAITHFUL_HOSTS, VALIDATOR_FULL symbols. Current state: validadoras NOT exposed in any host. Ruling 2 (2026-09-24): no catalog-backed field emits raw text; validadora columns do not appear in export at all. **Proposed replacement:** "Visibility gate (MOVILBA-128, 2026-08-27): the SUBE/OPENPASS validadora exposure gate was deleted. These columns no longer appear in the export of any host. The data exist in the database schema but are not emitted by the application. See MOBAAPI-630 revision as dated record (v5.3.99)." **Note:** This is a Jira dated record — keep as historical reference but add correction note. ### 5. docs/jira/MOBAAPI-630.md:105, :109 **Lines:** 105, 109 (identical stale claim) **Stale text:** "Ajuste de visibilidad (dev v5.3.99): los campos de validadoras SUBE/OPENPASS se muestran ahora SOLO en los ambientes fieles (FAITHFUL_HOSTS = staging + dev01)... El codigo y los datos siguen presentes en todos los ambientes; solo se restringe el render - quedan ocultos en prod, POD y HML." **Superseded by:** MOVILBA-128 (2026-08-27). Symbols FAITHFUL_HOSTS, VALIDATOR_FULL, useStagingFaithful no longer exist in src/. Ruling 2 (2026-09-24): no surface emits raw catalog text; validadora columns do not appear in export. **Proposed replacement (as footnote to the dated entry):** "Historical note (2026-09-24 correction): MOVILBA-128 (2026-08-27) deleted the host-conditional visibility gate (FAITHFUL_HOSTS, VALIDATOR_FULL symbols). As of v5.3.101+ the validadora SUBE/OPENPASS columns do not appear in the export of any host. This entry records v5.3.99 behavior; that code path no longer exists." **Note:** Jira dated record — add correction note but preserve original as historical. ### 6. docs/reference/v-vehiculos-export-definicion-correcta.md:20 **Line:** 20 **Stale text fragment:** "La opcion restante para hosts habilitados corresponde a comentarios de importacion, no a validadoras." **Status:** Line is CORRECT. Correctly states "no hosts enabled for validadora columns" — no replacement needed. --- ## BUCKET 4: STALE DIRECTION — DEV→PROD promotion as normal path (Superseded by ruling 3) ### 7. docs/reconciliation/2026-05-13-parque-movil-export-vs-prod.md:59 **Line:** 59 **Stale text:** "**Sequence** — build + verify the UPDATE on dev (dbcaba) first, eyeball a sample, then per-write GO to touch prod. Prod is live; never blind-sync." **Superseded by:** Ruling 1 (2026-09-24): PROD gets the work. This reverses dev-first. Also contradicts movilba-commons.md:93-94 (ruling 1): data fix verified in DEV replays to PROD with same script, no new GO (now approved carve-out, not "per-write GO"). **Proposed replacement:** "PROD is the focus. Write/verify directly in PROD under ROLLBACK (or, for approved fixes already verified in DEV, replay with the same instrument without new GO per ruling 1, 2026-09-24). DEV refresh is a separate, lower-priority sync from the validated PROD result." **Note:** This is a reconciliation-process document from 2026-05-13; aged out. ### 8. docs/reference/tagging-convention.md:59 **Line:** 59 **Stale text fragment:** "`push-to-gcaba.sh --movilba-only` (pushes the same dev HEAD that was just promoted" **Status:** This is a legacy reference to sh.git's `push-to-gcaba.sh` which no longer exists in movilba-web (repo-owned scripts only, per CLAUDE.md). Script name is stale (dead reference), not an import-order claim. No direct rulining conflict, but dated. **Proposed replacement:** "Obsolete reference to sh.git `push-to-gcaba.sh`; MovilBA uses repo-owned scripts `mba-deploy-dev.sh` and `mba-deploy-prod.sh` only (CLAUDE.md, 2026-09-17)." --- ## BUCKET 5: STALE DIRECTION — DEV/PROD parity as fidelity evidence (Superseded by ruling 3) ### 9. docs/agents-archive-2026-04-12.md:43 (lines 12-13 of entry #12) **Line:** 43 **Stale text fragment:** "Full fresh audit of dev vs prod DB. Tables identical (54 each). Views 67/69 identical... Diffs listed... Plan for A-D fixes." **Superseded by:** Ruling 3 (2026-09-24): "SSOT is ssot, thus data in prod or dev is irrelevant, only data in ssot is ssot. Neither base is evidence about the book. A DEV/PROD diff answers convergence and NEVER fidelity; agreement between the two proves nothing about either." **Status:** This is dated historical record (2026-04-02). No action on the text; it is a record of that time. Add footnote only if updated: "Historical audit; note: DEV/PROD parity does not prove fidelity to book (ruling 3, 2026-09-24)." --- ## SCANNING SUMMARY **Total .md files examined:** 237 across docs/, agents/, shared/ **Files actively searched (grep):** - Import order: commons.md, AGENTS.md, CLAUDE.md, ssot-overrides.md, migrations/README.md - Host gates: all docs/jira/*.md, docs/reference/*.md, CHANGELOG.md - DEV→PROD promotion: all reconciliation/*.md, migrations/*.md, reference/*.md - DEV/PROD parity claims: all agents-archive, MOBAAPI-*.md, reference/*.md **Files NOT searched (out of scope):** - agents/*/MEMORY.md (agent working memory, not rules) - agents/*/ltm.md (agent history, not rules) - agents/*-mba-l/AGENTS.md for individual agent lanes (lane-specific, not shared) - docs/jira/*.md except MOBAAPI-630, -653, -638, -658, -670 (Jira tickets are dated records; flagged existing ones only) - docs/design/*.md (design intent, not implementation rules) - docs/project-coding-guides/*.md (coding conventions, not import order) - All .sql files (migrations handled via README.md index) **Pattern:** Stale instruction density highest in shared/md/ (1 clear violation) and docs/reference/ (3 references to deleted symbols). Jira tickets correctly preserved as dated records. No new stale DEV→PROD promotion claims found in active rule files since 2026-05-13. --- ## NEXT STEPS (per coder03) - commons.md:115 — edit via ARO protocol (not direct file edit) - movilba-479:5, :179 — propose replacements above; evaluate overlap with pending 537 work - MOBAAPI-653, -630 — add footnote corrections as historical-record updates - movilba-commons.md:93-94 — verify this carve-out aligns with ruling 1 (preliminary: it does)
-
Researcher note (2026-10-05). Condensed per-item report of the 2026-09-24 inventory (full detail in the earlier progress note on this item). Re-checked current file state before reporting -- one item already fixed since, two still open. 1. shared/md/movilba-commons.md:31,140 -- ALREADY FIXED since the 09-24 inventory: now reads "PROD FIRST, on every axis... supersedes the old 'Dev first, prod after' wording." No action needed. 2. docs/reference/movilba-479-dev-prod-schema-census.md:5 -- STILL STALE: "la convergencia no tiene direccion fija... nunca se asume que DEV gana" contradicts the current PROD->DEV structure-flow ruling. Not a dated-number field (those are frozen on purpose per the file's own header), this is a standing directional claim. Needs an edit, not mine to make. 3. docs/reconciliation/2026-05-13-parque-movil-export-vs-prod.md:59 -- STILL STALE: "build + verify the UPDATE on dev first... then per-write GO to touch prod." This file is a dated reconciliation record (2026-05-13); per the ticket's own rule, dated records are flagged, not edited. 4. docs/jira/MOBAAPI-653.md:185, docs/jira/MOBAAPI-630.md:105/109 -- dated Jira records describing the MOVILBA-128-deleted FAITHFUL_HOSTS gate (validadora visibility). Correctly preserved as historical; MOVILBA-128 already superseded the code, no doc edit required by the ticket's own rule. 5. docs/reference/v-vehiculos-export-definicion-correcta.md:20 and movilba-479:179 -- checked CORRECT, no stale text, listed only so a zero-result bucket isn't mistaken for an unsearched one. 6. docs/agents-archive-2026-04-12.md:43 -- dated historical audit entry (2026-04-02), DEV/PROD parity claim. Correct-as-a-record, flagged only. 7. docs/reference/tagging-convention.md:59 -- dead reference to sh.git's retired push-to-gcaba.sh, unrelated to import order but stale/dated. Open action items (2 and 3) are edits for whoever owns those files next -- this ticket stays measurement-only per its own description.
-
Measurement-only inventory complete (2026-09-24, re-verified 2026-10-05). Found: shared/md/movilba-commons.md already fixed since inventory. Two remaining stale lines (docs/reference/movilba-479-dev-prod-schema-census.md:5, docs/reconciliation/2026-05-13-parque-movil-export-vs-prod.md:59) now have an owner: MOVILBA-707 (created 2026-10-05). Dated records (MOBAAPI-653/630, agents-archive-2026-04-12) correctly preserved as historical per the ticket's own rule, no action. No file edited by this ticket, per its own scope.
task
3w ago by coder03-mba-l-cc
1w ago
2026-10-06 17:56