bali_S20.md

world/regions/bali/page/bali_S20.md

Section 20 — Session Planning and Source Guide

CANON SUBORDINATION — this document is PROPOSAL-TIER: it serves canon and never outranks it.

The canon: the CVD · the T1 foundation docs · the T0 registries (registries/) · the spine

(docs/spine/CH_*.md) · the region pages (_source/02_Tier_2_Region_Pages/). Authority order: docs/DOC_MAP.md § 0.

Canon served (scanned from this document's own citations — widen it by hand where it is thin):

T0_Creature_Roster · T0_Source_Material_Registry

READ THAT CANON FIRST — open it and derive from it before you build anything from this document.

If this document disagrees with canon, CANON WINS and this document is the defect — fix the

document, never the canon. Nothing here is applied until it is ratified into canon.

This section is the build-gating record for the region: which processed source feeds which authoring

session, what state the retrieval layer is in, and what is honestly still closed. Pass-2 changed the

arithmetic. The page was cut against two processed sources; the bali shelf now carries FOUR, all four

deep-read and all four audited, and the two new ones are natural-science records that land in

different sections than the ethnography and the engineering dossier do. The Region Identity header's

"Sources processed this session: SRC_00157, SRC_00251" line predates that; the roster below is the

current state, and the header line wants refreshing to it at the assemble pass.

The processed source roster

Chunk counts are the T0_Source_Material_Registry declared values; all four rows carry ocr_method

native_text, so none of this shelf rides an OCR-recovery path and no chunk is quarantine-flagged.

Across the four records 105 typed rows exist outside the civilizational-mark class: 85 advance, 11

are HELD by critic MAJORs, 9 were cut at the audit. A further 16 civilizational_marks carry no status

field at all (8 from SRC_00157, 7 from SRC_00251, 1 from SRC_00482; SRC_00496 honestly carries none),

and a mark cut by the audit lives only in that record's care_findings — so a Section-4/7/8/9 session

reads the findings before it consumes any mark.

Per-section population plan

The plan is per-session, not per-source: a session opens one section, pulls the named source's rows,

and stops at the section boundary rather than harvesting the whole record.

• Sections 4 and 5 (people, culture, spiritual traditions) — SRC_00157 throughout, the region

keystone and the only processed source with participant-observer depth on lived practice: the

balian's craft [SRC: SRC_00157, p. 147], the calonarang and the widow-reading of Rangda

[SRC: SRC_00157, p. 296], the rwa bineda polarity with its third balancing centre

[SRC: SRC_00157, p. 2]. SRC_00251 joins this session for one thing the ethnography does not carry —

a working deity roster of the water landscape: Dewi Danu at the crater-rim lake temple and its

45-deity pantheon [SRC: SRC_00251, p. II-19], Dewi Sri at the rice sacrifice [SRC: SRC_00251,

p. II-16], the forest deity of the western mountain [SRC: SRC_00251, p. II-43], the deity of the

weir [SRC: SRC_00251, p. II-45], the sea god of seasonal pestilence [SRC: SRC_00251, p. II-53].

Deity rows route to T0_Creature_Roster under species_class deity, never to a boss registry on the

strength of a quest-giver or rescuer portrayal.

• Sections 1 and 7 (terrain, architecture, engineering) — SRC_00251 is the anchor and carries the

concrete engineering the page's subak claim needs: the proportional flow-dividers legible at a

glance [SRC: SRC_00251, p. II-5], the ridge tunnels emerging a kilometre downslope

[SRC: SRC_00251, p. II-3] and their named engineer corps in the Bebetin inscription

[SRC: SRC_00251, p. II-54], the large earthen storage dams [SRC: SRC_00251, p. II-73], the rock-cut

candi carved into living basalt [SRC: SRC_00251, p. II-32]. SRC_00157 supplies the architectural

grammar of the thresholds — the kori agung, the bhoma head, the candi bentar [SRC: SRC_00157,

p. 8]. The dividers session drops the vernacular term the extraction supplied for them: it is

attested nowhere in the dossier and an unsourced vernacular word inside living-tradition material

is exactly the authenticity failure the localization glossary in Section 17 would then propagate.

• Sections 6, 9, 10 (blade, resources, music) — SRC_00157, with a datable boundary now attached to

the keris material: the source dates the earliest firm kris records to 13th-century relief carvings

at Candi Panataran in East Java [SRC: SRC_00157, p. 152], which sits at or just past this page's

~1200 CE present frame. That is not an anachronism finding — keris-class daggers plainly predate the

earliest dated relief — but four rows lean on it, so the Section-6 session records the boundary

rather than losing it. The metallurgy itself (crucible reduction, lamination for pamor pattern) is

at [SRC: SRC_00157, p. 152-153]; the gamelan's paired-tuning shimmer at [SRC: SRC_00157,

p. 316-317] with the bronze casting at [SRC: SRC_00157, p. 319-320]; the four panasar

translator-clowns at [SRC: SRC_00157, p. 307]. The SRC_00251 weapon row did not advance, so

Section 6 draws the ethnography alone and the region's legendary seat stays what Section 6 already

reserves — an unnamed, lineage-carried pusaka line.

• Section 8 (politics) — the thinnest section against the processed corpus, and only partly improved.

SRC_00251 carries a dated royal-inscription series (Sukawana AD 882, Bebetin AD 896, Batuan AD 1022,

Pandak Bandung 1071) [SRC: SRC_00251, p. II-54], which is real Old Balinese royal-era political

texture — the two earliest predate this page's own Blanjong 914 CE literacy anchor, the two later

ones sit inside the Warmadewa era — but it arrives as a civilizational_mark with no status field and

therefore gates on the care_findings read before use. Nothing in the four records covers court

structure, succession, or the kepeng currency layer. Section 8 and the currency half of Section 9

stay deep-research gaps.

• Sections 13 and 14 (quests, trades) — the seed layer. 38 proposal-tier seeds sit in

docs/proposals/quest_seed_queue/bali.md, and that file is a regenerable merge view that carries the

HELD rows alongside the advanced ones. It is not the work-list. The consumption manifest is, and a

Section-13 session that reads the queue without filtering it through the manifest will consume rows

a critic is holding. Trade routing is unchanged: Trade 3 from SRC_00157, Trade 8 from SRC_00251,

Trade 4 from the usada materia-medica material [SRC: SRC_00157, p. 130-131].

• Section 15 (cross-regional) — the weave lane is the most heavily cut class on this shelf: of 17

weave links across the four records, 5 advance, 4 are held and 8 were cut, most for failing the

requirement that both ends be named AND cited. Section 15 draws only the advanced set and takes no

fauna link at all (below).

• Section 16 — queued to gain two citation rows for SRC_00482 and SRC_00496, both use_class

SUBSTRATE.

The two fauna records: Section 3 and Section 19

SRC_00482 and SRC_00496 are the new material, and they are complementary rather than redundant —

range, habitat and the historical extinction drivers on one side, phylogeny and restoration genetics

on the other. They split cleanly across two sections, and the split is an era rule, not a preference.

Section 3 (era-present tracking and fauna) takes what a person standing on the island in ~1200 CE

could observe or act on:

• The animal itself. Both records name Panthera tigris balica and Panthera tigris sondaica at

[SRC: SRC_00482, p. 3]; the Bali tiger's island-wide forest range at [SRC: SRC_00482, p. 5]; a

smaller overall build than the Javan tiger at [SRC: SRC_00496, p. 252]; roughly five-year generation

time and dependence on habitat carrying ungulate prey at [SRC: SRC_00496, p. 255]. This is what the

CR_0017 OBSERVE loop is built from, and OBSERVE stays its primary verb with combat defensive-only.

• Habitat wording is a correction the Section-3 session must apply. The phrase "volcanic-slope

monsoon forest" reads as sourced but is spliced: the volcanic-slope sentence at

[SRC: SRC_00482, p. 5] is about irrigation, and the monsoon-forest phrase at

[SRC: SRC_00482, p. 6] is about east Java. The honest anchor for Bali tiger habitat is that source's

own "large blocks of forested land" at [SRC: SRC_00482, p. 5]. The Agung-watershed monsoon-forest

setting stays legitimate because Section 1 is its terrain authority; it simply is not sourced here.

• Taxonomy is governed by SRC_00496, not by the 2006 bulletin. The bulletin uses the classical

multi-subspecies scheme; the genetics places the Sunda tigers in one monophyletic clade differing by

1-2 nucleotides [SRC: SRC_00496, p. 247]. "Bali tiger" as the island's own apex and balica as the

historically-used binomial are both fine for the frame; presenting separate-subspecies status as

current science is not.

• The tracking minigame is not yet groundable. The field-sign loop — pugmarks, scrapes, cached kills,

scent-marks — is attested by neither record; the genetics paper supports body size and prey

dependence and nothing about tracking behaviour. The loop advances on its own evidence, but the

Section-3 build spec waits on a tiger-ecology source, or the rewards-intelligence pillar is being

served by unsourced detail in the one place the game promises real science.

• The Zone Catalog owes this loop a home. The highland forest of the Agung watershed has no zone row;

the catalog's seven zones are the temple, ward, springs, terraces, performance ground, reef and

sea-cliff. Routing a tiger-tracking loop into besakih_mother_temple would put it inside a protected

living sacred precinct, which is worse than leaving the field null. Either a highland-forest zone

row or an explicit region-ambient routing closes it.

Section 19 (the modern-revisit strata) takes everything the ~1200 CE player cannot reach:

• The extinction itself, with both dates carried rather than one silently correcting the other: last

confirmed record 1937, in the 1930s per [SRC: SRC_00496, p. 255]; gone from the wild by the end of

the 1940s per [SRC: SRC_00482, p. 5], with [SRC: SRC_00496, p. 247] treating the subspecies as

extinct in the 1940s. Both are true of different events, and the page states both.

• The drivers: the Dutch colonial sport-hunting regime plus agricultural and road fragmentation

[SRC: SRC_00482, p. 5]. This is documented history and is rendered with weight, named as the record

names it, and never re-attributed to the living Balinese community.

• The restoration question: the Sumatran tiger as the closest living relative of both extinct

subspecies and the proposed stock for any managed restoration [SRC: SRC_00496, p. 247, p. 255].

This is authorial and design-layer substrate, not player-investigable content in any frame the arc

stages.

• The museum trail, restated to what Table 1 actually shows: two voucher Bali specimens at Stuttgart,

plus four BAL-haplotype specimens of uncertain Indonesian origin at Amsterdam and Leiden

[SRC: SRC_00496, p. 249]. The Shanghai, St Petersburg and New York holdings in that table carry

Javan and Sumatran haplotypes and are not part of the Bali claim.

• The deep-time material — Pleistocene sea-level fall repeatedly joining Bali, Java and Sumatra across

the exposed Sunda Shelf [SRC: SRC_00496, p. 248] — is L3 authorial substrate. It grounds a

volcanic-catastrophe register by resonance only; it never becomes an assertion that Agung is

Toba-class or that a Sumatran event is a Balinese one.

Two hard constraints bind every session that touches this material. First, a modern conservation-

genetics result is not L2 depth "available to the player who investigates it" in a pre-colonial

frame — a player of that era cannot investigate an inbreeding coefficient, so the finding belongs to

the modern strata or to authorial notes. Second, Bali sits on the Sunda Shelf, west of the Wallace

Line, and is not a Wallacea island; the line runs between Bali and Lombok, and the source says so

directly in listing "the Sunda Islands (Sumatra, Java, Bali and Borneo)" [SRC: SRC_00482, p. 7]. One

extraction asserted the opposite and the audit caught it. That correction is also why no fauna weave

link reaches Section 15: both fauna links are held by critic MAJORs — one on the biogeographic error

just corrected, one on a mis-typed relation, since the Bali, Javan and Sumatran tigers are one recent

lineage rather than an independent parallel [SRC: SRC_00496, p. 247]. Section 15 takes no fauna link

until the link lane rules on the held rows, and it loses nothing by waiting: Section 3's fauna

register and Section 15's existing Bali-to-Sumatra-Java lines already carry the content honestly

without a graph row.

Processing and query-readiness

• All four sources are processed and query_ready in the book_substrate collection with region_tags

bali, and the region-to-slug map resolves bali one-to-one, so region-filtered retrieval reaches this

shelf and only this shelf: python harness/rag/query.py "<question>" --collection book_substrate

--region bali. Region filtering is not optional here — semantic-only retrieval over a multi-region

corpus ranks by topic and not by place, which is the failure the filter exists to stop.

• A substrate mis-shelving repair ran against this corpus recently, so retrieval provenance is

current. Six sources had been mis-shelved onto a neighbouring region tag; the registry rows were

reshelved and 253 chunk region_tags were repaired in the store, verified 253 before, 253 changed,

0 remaining, with the legitimately-tagged chunks untouched and the embeddings and vector table not

rewritten. No bali-shelf source was in the mis-shelved set, and the store is gitignored, so this

note plus the STATUS.md entry are the provenance trail for a change that leaves no diff.

• Two provenance-hygiene items sit on the registry rather than on the page. The registry titles for

SRC_00251 and SRC_00496 are filename stubs rather than bibliographic titles, so a title lookup

against the registry will not identify either source — Section 16 and this section carry the correct

citations. And the four registry rows still read extraction_status deep_read_extracted while all

four EXR records read deep_read_audited; the registry lags the audit, and any consumer that gates on

the registry field will under-read this shelf's actual state.

• Anchor conventions differ per source and are not interchangeable: SRC_00157 uses the book's print

pages (print page equals PDF page minus 13); SRC_00251 uses the dossier's own "II-N" chapter-page

markers; SRC_00482 uses the bulletin's printed page numbers, where an unrepaired "p. 2" anchor

returns the cover rather than the article; SRC_00496 uses journal print pages 247-257. Anchors were

verified against the sources at the audit, not assumed, and several were repaired by one to two

pages — the repaired anchors are the ones this page cites.

Reading the harvest without consuming what is held

The consumption manifest at docs/proposals/region_pass2/bali_CONSUMPTION_MANIFEST.md is script-emitted

from the records' typed status fields and is the only work-list. The derived views are convenient and

none of them are filtered: the seed queue carries HELD seeds inline, and the per-registry merge streams

under docs/proposals/category_extractions/ still show every entry at proposed because they were

generated before the audit advanced anything. A session that opens a derived view first will read rows

that a critic is holding, so the manifest is opened first and the view second. Regenerate, never

hand-edit, either one.

Two other holds shape what a session can finish. No ratified quest_type vocabulary exists and the mint

decision is pending a ruling, so 19 seeds across this shelf carry quest_type mint flags and none of

them may be force-fit into an attested value to close a section. And four Zone Catalog additions are

proposed against material these records document at places the catalog cannot currently name — a Batur

crater-rim water-temple zone and a Pakerisan-gorge zone from the dossier, a pande-forge zone and a

cremation-ground / Pura Dalem zone from the ethnography, with the highland-forest zone above making

five shelf-wide. Until those land, the honest value on a seed's zone hint is null, and staging one

community's documented sacred history at a neighbouring community's temple to fill the field is a

misattribution, not a workaround.

Build-ready gating

The machine layer is build-ready now against the v3.1 spine: the Zone Catalog rows copy the CH_04 site

rows verbatim and add only page-side geometry, and the Derived Indexes resolve as live joins into the

typed layer. Nothing in the four processed records touches the antagonist apparatus — they are a

religion ethnography, an engineering dossier and two natural-science papers — so the chapter's

institutional-proxy strand is spine-authored throughout and needs no source session at all.

Two gates stay open, and one of them has changed shape:

• The Rangda-ambivalence keystone. Belo remains the §17.1-critical anchor for the never-villainized

guardrail, and the processed corpus cannot carry it — the comparative far ends of the keystone

ethnography's own links are the author relaying general Indological knowledge outside his fieldwork,

so those readings should not harden before a specialist source lands. What has changed is that this

is now an INGEST gate rather than an acquisition gate: the owned-inventory ledger records Belo's

Trance in Bali as already owned, alongside Spies and de Zoete's Dance and Drama in Bali, so the

blocker is chunk-and-embed into the bali shelf, not purchase. Two items still want resolving at that

ingest: the companion Rangda and Barong monograph is not in the owned inventory, and the ledger's

Malay-magic row lists An Analysis of Malay Magic under Skeat rather than Endicott, which is the

source Section 16 names for the pengiwa/penengen craft distinction. Lansing, Perfect Order remains

genuinely unacquired and appears nowhere in the source registry.

• The Warmadewa political-detail deep-research. Section 8's court structure and Section 9's ~1200 CE

kepeng-currency layer have no processed anchor; the inscription series above is the only new

purchase on the era's politics, and it is a status-less mark. Default path is deep-research

synthesis rather than acquisition.

Session sequencing follows the cost curve. The two fauna records are 54 chunks between them and their

sections are narrow, so Section 3 and the Section-19 strata close first. The dossier session (Sections

1, 7, and the Section-8 partial) closes second — 114 chunks and the highest concentration of

concrete engineering marks. The keystone ethnography drives the largest block, Sections 4, 5, 6, 9, 10

and the seed layer, and is best run as several passes rather than one. The depth pass on Sections 5

and 6 waits behind the Belo ingest, and the two deep-research gaps run independently of all of it. The

v1.0 book-cost fields are superseded by the closed acquisition ledger.

Generated by harness/site/structure_site.py — the URL path is the repo path. review root