Assumptions and open questions
Separate working assumptions, accepted decisions, and open questions. Do not invent statistics. Prefer experiments, legal advice, or user tests where noted.
Accepted (do not reopen casually)
| # | Decision | Where |
|---|---|---|
| — | Three repos (tunes / tunes-ios / tunes-web) | ADR-003 |
| — | No implied TfL endorsement | Charter |
| — | Docs-first; no product feature code in this execution | Plan |
Working assumptions (provisional)
- Canonical unit = station-to-station section with duration (methodology notes + professional practice).
- Privacy default = local record → review → consent → derived-only upload; raw audio on-device by default.
- Schema = generic railway concepts; London data as instances (London-first portable schema).
- Quality = tiered contributions (A–E working labels); crowd in-service data is survey-grade at best.
- Objective ≠ subjective in the data model; no unexplained blended score.
- Quiet-room check ≠ calibration.
- iPhone-first scientific collector; not browser/PWA for v1 collector unless experiments prove otherwise.
- Prefer dense pilot coverage over shallow network-wide mapping.
Brief questions — Wave 2 disposition
| # | Question | Disposition |
|---|---|---|
| 1 | Underground only vs all London rail? | ADR-001 — pilot Underground-first; launch breadth deferred |
| 2 | Tube-specific vs generic core? | ADR-002 — generic core |
| 3 | Monorepo vs multi-repo? | ADR-003 — three repos |
| 4 | Raw audio leave device? | ADR-004 — default no |
| 5 | Derived-only? | ADR-005 — yes |
| 6 | Grade distinction? | ADR-006 — tiers A–E |
| 7 | Watch / AirPods? | ADR-007 — experiment backlog |
| 8 | Voice Memos? | ADR-008 — Tier D if at all |
| 9 | Manual metadata burden? | ADR-009 — minimise |
| 10 | Subjective timing? | ADR-010 — after journey |
| 11 | Claim language? | ADR-011 — strict |
| 12 | Licences? | ADR-012 — provisional; legal before publish |
| 13 | Controller and entity path? | ADR-013 — named controller; CLG by launch |
Additional open questions
- Exact retention periods and final lawful-basis wording (counsel) after ADR-013 entity path
- Whether ethics review is required before public beta campaigns (see R11 / R12; DPIA required before public recruitment)
- Duplicate / fraud detection thresholds without punishing honest edge cases
- Whether GPS is used at all underground (likely topology snap, not raw trust)
Recommendation
Questions 1–13 now have ADR dispositions. Treat those ADRs as binding and keep working assumptions 1–8 unless experiments falsify them; use the additional-open-questions list for unresolved work.
Confidence: High for the separation of concerns; Low for licence and medical/claim language until legal/acoustic review.
Depends on experiment/legal/user-test? Yes — see table.
Links: 02-research-map.md; governance/decision-log.md; H18 assumptions.