From 617a54d605aa010fbce5b8768c59ecc6a4c36998 Mon Sep 17 00:00:00 2001
From: Ronald Tse <ronald.tse@gmail.com>
Date: Sun, 1 Nov 2026 19:10:58 +0800
Subject: [PATCH] =?UTF-8?q?docs:=20TODO.all-sys=20=E2=80=94=20every=20rema?=
 =?UTF-8?q?ining=20system,=20fully=20specified=20(INDEX=20+=2012=20work=20?=
 =?UTF-8?q?orders)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 TODO.all-sys/01-ala-lc-batch-1.md    | 13 +++++++++++++
 TODO.all-sys/02-ala-lc-batch-2.md    |  8 ++++++++
 TODO.all-sys/03-ala-lc-batch-3.md    |  5 +++++
 TODO.all-sys/04-ala-lc-batch-4.md    |  6 ++++++
 TODO.all-sys/05-ala-lc-batch-5.md    |  7 +++++++
 TODO.all-sys/06-din-columns.md       |  7 +++++++
 TODO.all-sys/07-pd6505-korean.md     |  8 ++++++++
 TODO.all-sys/08-iso-gaps.md          |  8 ++++++++
 TODO.all-sys/09-icao-ungegn-diffs.md |  5 +++++
 TODO.all-sys/10-b3-release.md        |  9 +++++++++
 TODO.all-sys/11-urd-oracle.md        |  7 +++++++
 TODO.all-sys/12-owner-gates.md       |  8 ++++++++
 TODO.all-sys/INDEX.md                | 26 ++++++++++++++++++++++++++
 13 files changed, 117 insertions(+)
 create mode 100644 TODO.all-sys/01-ala-lc-batch-1.md
 create mode 100644 TODO.all-sys/02-ala-lc-batch-2.md
 create mode 100644 TODO.all-sys/03-ala-lc-batch-3.md
 create mode 100644 TODO.all-sys/04-ala-lc-batch-4.md
 create mode 100644 TODO.all-sys/05-ala-lc-batch-5.md
 create mode 100644 TODO.all-sys/06-din-columns.md
 create mode 100644 TODO.all-sys/07-pd6505-korean.md
 create mode 100644 TODO.all-sys/08-iso-gaps.md
 create mode 100644 TODO.all-sys/09-icao-ungegn-diffs.md
 create mode 100644 TODO.all-sys/10-b3-release.md
 create mode 100644 TODO.all-sys/11-urd-oracle.md
 create mode 100644 TODO.all-sys/12-owner-gates.md
 create mode 100644 TODO.all-sys/INDEX.md

diff --git a/TODO.all-sys/01-ala-lc-batch-1.md b/TODO.all-sys/01-ala-lc-batch-1.md
new file mode 100644
index 0000000..664622d
--- /dev/null
+++ b/TODO.all-sys/01-ala-lc-batch-1.md
@@ -0,0 +1,13 @@
+# ALA-LC Batch 1 — Brahmic/SE-Asian (14)
+
+Sources: alalc/{burmese,khmer,lao,thai,shan,javanese,balinese,batak,
+lepcha,limbu,santali,manipuri,cham,tibetan}.pdf (2012-era tables).
+Recipe per system: read the table (text layers are thin — read pages
+visually at 3-5x); if glyphs are Latin lookalikes (batak.pdf), establish
+glyph→Unicode visually against the script's block; encode
+alalc-<lang>-<script>-Latn-<year>.isc with tests from the doc's own
+examples; one PR for the batch; cross-run each map's tests before push.
+Systems: Burmese mya-Mymr, Khmer khm-Khmr, Lao lao-Laoo, Thai tha-Thai,
+Shan shn, Javanese jav, Balinese ban, Batak btk-Batk, Lepcha lep,
+Limbu lif, Santali sat-Olck, Manipuri mni-Mtei, Cham cja/cjm, Tibetan
+bod-Tibt.
diff --git a/TODO.all-sys/02-ala-lc-batch-2.md b/TODO.all-sys/02-ala-lc-batch-2.md
new file mode 100644
index 0000000..6275964
--- /dev/null
+++ b/TODO.all-sys/02-ala-lc-batch-2.md
@@ -0,0 +1,8 @@
+# ALA-LC Batch 2 — Arabic-script (13)
+
+Sources: alalc/{hebrew-yiddish? per-lang}.pdf (arabic, judaeo-arabic,
+jawi, malay, kazakh, kurdish, ladino, sindhi, uighur, urdu, pushto,
+ottoman-turkish, moghol...). Same recipe as 01. IDs:
+alalc-<lang>-Arab-Latn-<year>.isc. Watch: right-to-left tables,
+vowel-mark conventions, the din-ara technique notes in WO89 (sukun
+strip first, mark-order swaps, long-vowel merges before taa-marbuta).
diff --git a/TODO.all-sys/03-ala-lc-batch-3.md b/TODO.all-sys/03-ala-lc-batch-3.md
new file mode 100644
index 0000000..200cd92
--- /dev/null
+++ b/TODO.all-sys/03-ala-lc-batch-3.md
@@ -0,0 +1,5 @@
+# ALA-LC Batch 3 — Cyrillic (4)
+
+Sources: alalc/{church-slavic,rusyn,byelorussian?,serbian?}.pdf +
+non-Slavic Cyrillic leftovers. Same recipe. din/bsi Cyrillic technique
+notes apply (yery pair, i/_j conventions, disambiguating apostrophes).
diff --git a/TODO.all-sys/04-ala-lc-batch-4.md b/TODO.all-sys/04-ala-lc-batch-4.md
new file mode 100644
index 0000000..c4bedad
--- /dev/null
+++ b/TODO.all-sys/04-ala-lc-batch-4.md
@@ -0,0 +1,6 @@
+# ALA-LC Batch 4 — East Asian (2)
+
+Sources: alalc/{chinese,japanese}.pdf. Chinese = Wade-Giles (large
+syllabary — per-syllable table like ISO 7098); Japanese = McCune-
+Reischauer kana + kanji readings are M (Track B) — encode the kana
+table deterministically, note the M remainder.
diff --git a/TODO.all-sys/05-ala-lc-batch-5.md b/TODO.all-sys/05-ala-lc-batch-5.md
new file mode 100644
index 0000000..0dd07de
--- /dev/null
+++ b/TODO.all-sys/05-ala-lc-batch-5.md
@@ -0,0 +1,7 @@
+# ALA-LC Batch 5 — misc (10)
+
+Sources: alalc/{armenian,coptic,deseret,cherokee,manchu,ncʼtut? amharic
+done...}.pdf — Armenian hye-Armn, Coptic cop-Copt, Deseret dsrt-Dsrt,
+Cherokee chr-Cher, Manchu mnc-Mong, Mandekan N'Ko nko-Nkoo, Moroccan
+Tamazight zgh-Tfng, Tamashek tmh-Tfng, Tod-Oirat xal?, Vai vai-Vaii.
+Same recipe.
diff --git a/TODO.all-sys/06-din-columns.md b/TODO.all-sys/06-din-columns.md
new file mode 100644
index 0000000..ce9c39f
--- /dev/null
+++ b/TODO.all-sys/06-din-columns.md
@@ -0,0 +1,7 @@
+# DIN columns + 1460-2 + 32708
+
+Sources: interscript-private-references/din/*.pdf. 1460-2 = Cyrillic
+pt.2 (verify subject at encoding); 31635 = Persian/Kurdish/Urdu columns
+(Arabic column SHIPPED maps #223 — reuse its technique notes); 31636 =
+Yiddish column (Hebrew column SHIPPED #217/#218; yid id-suffix fix
+#226); 32708 = verify subject. One PR per document.
diff --git a/TODO.all-sys/07-pd6505-korean.md b/TODO.all-sys/07-pd6505-korean.md
new file mode 100644
index 0000000..1adf085
--- /dev/null
+++ b/TODO.all-sys/07-pd6505-korean.md
@@ -0,0 +1,8 @@
+# PD 6505 Korean — promote from staging
+
+Draft encoded+probed (maps PR #227, maps-staging/bsi-kor-Kore-Latn-
+6505-1982.isc). REMAINING: deep pass over source sections 7-11.15
+(renders in /tmp/pd6505/ — recreate from the private PDF if reaped);
+expand tests with the doc's 11.3 vowel-word table + 13.2/13.3 N/S
+pairs; revisit tense-consonant placeholders + rare trailing clusters;
+then move the map from maps-staging/ to maps/ and merge.
diff --git a/TODO.all-sys/08-iso-gaps.md b/TODO.all-sys/08-iso-gaps.md
new file mode 100644
index 0000000..281a0ec
--- /dev/null
+++ b/TODO.all-sys/08-iso-gaps.md
@@ -0,0 +1,8 @@
+# ISO gaps
+
+ISO 9:1995 non-Russian Cyrillic — per-letter coverage check vs registry
+(scripts exist: maps/tools + docs/coverage diffs). ISO 259 Hebrew letter
+table (pointed input = Track B companion). ISO 7098 pinyin (large
+syllabary — the din chinese approach). ISO 15919 Indic remainder —
+enumerate the script set vs the 4 existing pli/san/hin? maps and encode
+the missing scripts.
diff --git a/TODO.all-sys/09-icao-ungegn-diffs.md b/TODO.all-sys/09-icao-ungegn-diffs.md
new file mode 100644
index 0000000..65cca0a
--- /dev/null
+++ b/TODO.all-sys/09-icao-ungegn-diffs.md
@@ -0,0 +1,5 @@
+# ICAO + UNGEGN source diffs
+
+Obtain current ICAO 9303 + UNGEGN romanization documents; diff each vs
+the registry's icao-*/un-* maps; encode genuine gaps, note reissues.
+The registry has icao-{rus,bel,bul,mkd,srp,ukr}-9303 + icao-mul-2015.
diff --git a/TODO.all-sys/10-b3-release.md b/TODO.all-sys/10-b3-release.md
new file mode 100644
index 0000000..1087598
--- /dev/null
+++ b/TODO.all-sys/10-b3-release.md
@@ -0,0 +1,9 @@
+# B3 — stage 2 + release
+
+The two relabel arms auto-consolidate (monitor armed; RELABEL_DONE +
+relabel_stats.json land in bucket run-086-oracle-relabel/). THEN:
+launch run-087-oracle-student per task #93 (byt5-base on oracle
+targets; gates RAW multiref WER < 17.3794, stretch <= 10.1262, Sadeed
+DER within +0.1 of 2.2864). On gates green: export IMF artifact +
+parity (>=500 samples single-row greedy) + certify (tools/certify.py);
+the RELEASE VERSION IS THE OWNER'S DECISION — prepare, never pick.
diff --git a/TODO.all-sys/11-urd-oracle.md b/TODO.all-sys/11-urd-oracle.md
new file mode 100644
index 0000000..6f795e1
--- /dev/null
+++ b/TODO.all-sys/11-urd-oracle.md
@@ -0,0 +1,7 @@
+# Urdu WO65 — stage 1 + 2 (staged)
+
+Launch after B3 frees the a100s — full command pair recorded in
+TODO.sota/65 (arms urd-diac-1.0 + urd-diac-base-1.0 over bucket
+rababa-urdu-cc174b96@data-diacrit, D_RUN=run-088-urd-oracle, 200K x
+300B subset). Stage 2 = train_oracle_student_hf.py adapted
+(google/byt5-base). Gate CER < 3.57, target < 3.
diff --git a/TODO.all-sys/12-owner-gates.md b/TODO.all-sys/12-owner-gates.md
new file mode 100644
index 0000000..56fdd54
--- /dev/null
+++ b/TODO.all-sys/12-owner-gates.md
@@ -0,0 +1,8 @@
+# Owner-held gates (RECORD ONLY — never worked around)
+
+- Hebrew vocalization data asks (B2 line)
+- PyPI registration for interscript packages
+- Urdu G2P commercial-license data selection (urd-ipa ship)
+- SEAlang / Chuon Nath full IPA for Khmer CER<15
+- Corpus stale duplicates bgnpcgn-{av,cv,os}- deletion decision
+- Every release version number
diff --git a/TODO.all-sys/INDEX.md b/TODO.all-sys/INDEX.md
new file mode 100644
index 0000000..b623e5e
--- /dev/null
+++ b/TODO.all-sys/INDEX.md
@@ -0,0 +1,26 @@
+# TODO.all-sys — every remaining system, fully specified
+
+Terminal: every identified transliteration system has an ISC map in the
+registry, tested from its own source document, live on the site.
+
+Status line: 318 done / ~45 open. Sources: all local
+(interscript-private-references/alalc/ — 92 docs; din/, bsi/ present).
+
+Rules (binding): visual transcription from scans is sanctioned; every
+map carries tests from its own document's examples; one PR per batch;
+never delete files; owner-held gates are recorded, never worked around;
+versions are owner decisions.
+
+- [ ] 01-ala-lc-batch-1.md — Brahmic/SE-Asian, 14 systems
+- [ ] 02-ala-lc-batch-2.md — Arabic-script, 13 systems
+- [ ] 03-ala-lc-batch-3.md — Cyrillic, 4 systems
+- [ ] 04-ala-lc-batch-4.md — East Asian, 2 systems
+- [ ] 05-ala-lc-batch-5.md — misc, 10 systems
+- [ ] 06-din-columns.md — DIN 1460-2, 31635 fas/kur/urd, 31636 yid, 32708
+- [ ] 07-pd6505-korean.md — promote the draft from staging
+- [ ] 08-iso-gaps.md — ISO 9-Cyrl check, 259, 7098, 15919 remainder
+- [ ] 09-icao-ungegn-diffs.md — source-doc diffs vs registry
+- [ ] 10-b3-release.md — B3 stage 2 (auto after arms) + ship
+- [ ] 11-urd-oracle.md — WO65 stage 1+2 (staged commands)
+- [ ] 12-owner-gates.md — Hebrew data, PyPI, urd-g2p license, corpus
+      duplicates (record; owner decides)