2026-07-21 evening. Spec: scratchpad/roots-v5-change-spec.md. Sources: ../2026-07-21-frontier-round4-partE.md + partF.md.
Applied
- A. Date pins / estimate retirements — 14 estimate-halves retired across 10 people: Gregory Dexter (eb+ed → 1610/1700), Thomas Angell (eb+ed → 1618/1694), Lucy Burt (eb+ed → 1790/1876), Sarah Mary Salisbury (eb+ed → 1835/1924), Abigail Fullerton (ed → d.1706), Abigail Dexter (eb → b.1655), James Angell (ed → d.1711), James Ashton (ed → d.1651), Alice (Ashton) (ed → d.1643), Alice (surname unproven) (ed → d.1695). Sarah Burt Salisbury reverify tag cleared; Richard Burt + Lumpkin stories carry the new day-level pins (dates were already real).
- B. Corrections — Olney–Sayles marriage → Dec 20, 1692 everywhere (ANC 1408/1409, bios catharine-sayles ×2 + william-olney-1663, john-sayles, roger-williams; Arnold checked/silent note replaces the "Arnold is the closer" line incl. the Open-frontier table row). Angell–Ashton sisters-link marked CONTRADICTED on every occurrence: ANC 2820/2821/5633/705/704 + GROUPS William Olney Ⅳ + bios (marie-ashton, james-ashton, alice-ashton, phebe-angell, william-olney-1694, thomas-angell, alice-surname-unproven) — sweeping beyond the spec's three named nodes was required for coherence. Josephine page: cert #923 added, Sec 81 hypothesis downgraded to POSSIBLE, John M. Oakford 1898 probate = decisive next document (nothing was on the viz — confirmed). Thomas Olney Jr 1630-baptism variant logged, 1632 kept.
- C. New people — ANC 182 John Burt (1775–1871), 183 Sarah Olney (E1 est ~1786–~1848 from the 1807 marriage; do-not-conflate + Sylvester-2-vs-FaG-5 notes), 11276 James Williams (d.1621), 11277 Alice Pemberton (d.1634) with the St Albans/"cosen and Godsonne" crown jewel. Epinetus mentioned in Thomas¹/Marie stories + bios only (not an ancestor). POSSIBLE tier kept off viz.
- D. Pages — people-bios.json: 15 existing entries updated, 7 new (james-williams, alice-pemberton, john-burt, sarah-olney, gregory-dexter, thomas-angell, alice-surname-unproven). build-people.mjs: LINE_BY_AHN +4, MARRIAGE 182/1410, FRONTIER_HINTS +4 / stale roger-williams hint removed, SOURCES +8. Regenerated: 116 pages (was 112).
Decisions / deviations
- James Williams + Alice Pemberton are tree-only, no eb (spec's "est via E1?" resolved NO): the E-rules require a real-record anchor; his marriage year and first child's birth are unrecorded, and Roger's c.1603 is itself convention, not a record.
- Birth-less render fallback added: tree nodes and person pages show "?–1621" instead of "dates open" when only death is real (small code change, index.html tree
yrs+ build-people personPage/personLink). Without it the two new Williamses would have displayed "dates open" while holding real death years. - "c.1610" label rendered on Dexter's tree node via the same hardcoded pattern as Thomas Olney's c.1600 (
n===2822); table/tooltip carry plain 1610 with the traditional/POSSIBLE story note — precedent-consistent, no new data field invented. - Re-derivations forced by retirements: Abigail Fullerton eb 1634→1624 (E2 off first recorded child Stephen 1647 — old basis referenced the retired ~1657 estimate); Abigail Dexter ed 1719→1717 (E3 off real b.1655, floored ≥ last-seen 1711).
- No ahn 364/365 duplicate: John Burt's father is Richard Burt Sr, already drawn at 178 as Lucy's father; the pedigree fold is carried in stories + frontier hint instead of duplicating him.
- Out of spec scope, deliberately NOT applied (flag for parent): Harriett Osborn maiden-name POSSIBLE→PROBABLE upgrade (partE, Osborn Calliway cert); William Olney⁴ b. Oct 6 1694 day pin; wider Eva-Brown reroute (one open-frontier line only). Also added a "New tonight (frontier round 4)" table row + Arnold/Waters/Swan/Sylvester to the intro source list.
Verification (static only, per spec — no browser, no deploy)
node --check on the extracted index.html script: PASS. build-people.mjs ×3 clean, 116 pages, no unassigned-line warnings. people-bios.json valid; indent=1 round-trip formatting preserved. Zero null/NaN/undefined artifacts across all generated pages. Simulated timeline invariants: 114 bars, all ranges ≥ Y0/ordered, all regions valid, every eb/ed paired with ebasis (and vice versa), parent links 182/183→91 and 11276/11277→5638 correct. Sample pages grepped: roger-williams (parents linked), sarah-burt-salisbury (parents linked, reverify cleared), catharine-sayles (Dec 20 ×2), james-williams/alice-pemberton (?–year headers), sarah-olney (~est header + estimated aside), gregory-dexter/thomas-angell (no estimate aside).