PostgreSQL KO Translation Status — Per-Document Phase Board

Contents:

• Why this doc exists
• The phase model lives elsewhere
• How the PostgreSQL KO mirrors were produced
• Current phase: P2 across the board
• Aggregate counts
• Open follow-ups

Why this doc exists

Every EN doc under knowledge/code-analysis/postgres/ has a KO mirror at knowledge/ko/code-analysis/postgres/, authored as a parallel composition (not a translation) per knowledge/CLAUDE.md §“Bilingual layout”. This board records how mature those mirrors are and what the remaining quality gates are, so a future reader knows whether a given KO mirror is publication-ready or still needs a polish/review pass.

The phase model lives elsewhere

The 5-phase maturity model and its rules are defined once, in cubrid-translation-status.md (and the playbook knowledge/methodology/korean-translation.md). Do not redefine them here. In short:

• P0 — literal KO mirror (direct translation).
• P1 — S1 machine-cleanup gate (kb-ai-tell-check zero S1 findings).
• P2 — structural rewrite (parallel-composed, S1-clean, scannable Korean).
• P3 — natural Korean prose (a dedicated polish pass).
• P4 — native-speaker / opus review signoff.

The per-doc phase marker convention (the p<n> tag in each KO mirror’s frontmatter tags: as the SSOT, and the [KO_p<n>](path) link marker used when other docs point at a KO mirror) is also defined there. If you change the model, change it in the cubrid doc, not here.

How the PostgreSQL KO mirrors were produced

The PostgreSQL tree did not follow cubrid’s P0→P1 path (literal mirror, then a multi-round machine-cleanup of translationese). Instead each KO mirror was authored from scratch as a native parallel composition:

• Model: sonnet, single-pass, no reviewer loop (per knowledge/CLAUDE.md rule 9 — KO authoring is decoupled from the EN writer-reviewer loop).
• Style: the §“한국어 작성 스타일” rules — English technical terms kept with a short Korean gloss on first mention, one idea per sentence, explanatory framing (~다는 점이다, ~란 ~을 의미한다, 이는 … 때문이다).
• Gate run per doc: kb-ai-tell-check against the KO file, with A-category findings driven toward S1 = 0.
• Shared structure with EN: same section order, same English code excerpts, same Mermaid skeletons (same node IDs), same position-hint table, same cross-reference doc names, same updated: date.

Because they are S1-clean parallel compositions (already past the P0/P1 literal-then-clean stages by construction) but have not had a separate prose-polish batch (P3) or a native-speaker review (P4), they sit at P2.

Current phase: P2 across the board

All KO mirrors under knowledge/ko/code-analysis/postgres/ are at P2 as of 2026-06-06. This is the project-wide uniform state — there is no per-doc variation yet, so this board records the aggregate rather than 139 identical rows. (Contrast cubrid, whose 103 mirrors are uniformly P3 after dedicated polish batches.)

PostgreSQL-specific note on the P1 sub-gate: cubrid’s “four-phase rewriter” internal sub-phases (Hangul-anchored → boundary → boundary-relaxed → cross-line) were a cleanup of pre-existing literal mirrors and do not apply here — the PostgreSQL mirrors were never literal, so there was nothing to mechanically de-translationese; the single kb-ai-tell-check pass per doc covers the S1 gate.

Aggregate counts

KO mirrors by group (all P2):

Group	KO mirrors
overview + coverage + translation-status	16
storage-engine modules	20
txn-recovery modules	14
query-processing modules	20
server-architecture modules	9
monitoring-stats modules	2
system-catalog modules	6
ddl-schema modules	10
replication-ha modules	9
client-protocol modules	3
extensibility modules	6
base-infra modules	7
i18n-text modules	3
utilities modules	8
evolution docs	7
Total	≈140

(The exact total tracks the EN tree one-for-one; regenerate from ls knowledge/ko/code-analysis/postgres/*.md if a precise count is needed. EN = KO = meta parity is maintained per batch.)

Open follow-ups

1. Stamp per-doc p2 frontmatter tags. The cubrid SSOT convention puts the phase tag in each KO mirror’s tags: array. The PostgreSQL mirrors do not yet carry it; a mechanical pass should add p2 to each so this board becomes derived state rather than a hand-asserted uniform claim.
2. P3 promotion. A dedicated natural-prose polish batch (sonnet), per the methodology, would lift the tree to P3 — matching cubrid.
3. P4 promotion. Per-doc native-speaker / opus review signoff; the only phase that requires human or opus judgement.
4. [KO_p2](path) link markers. When EN docs link to their KO mirrors, use the marker convention from the cubrid SSOT with the current phase value.