rippr/docs/BACKLOG.md

Carried forward from the native repo (~/dojo/rippr/docs/v3/BACKLOG.md) when the Flutter port completed. Two items below have changed status since it was written:

• Compose UI tests — no longer a gap. The port has 15 widget tests plus 4 integration tests; see docs/port/PARITY-AUDIT.md.
• Elevation gain accuracy — the algorithm is now proven bit-identical across Kotlin and Dart (tool/parity/run.sh), so any future tuning can be checked against the original rather than guessed at. The instruction below still stands: do not tune it blind.

Everything else carries over unchanged, including the v3 ideas and the "must not regress" list.

---

Backlog — v3 and v4

Everything known-outstanding, with enough context to pick up cold. Nothing here is committed to — it is a menu, roughly ordered by value.

v3 is now broken out into tickets: v3/README.md. This document stays as the reasoning; the tickets are the executable form.

v3 is everything that can be built with no server. v4 is everything that cannot. That split is the most useful thing in this document: it means v3 can proceed indefinitely without anyone deciding to run infrastructure or hold other people's location data.

Before planning anything: run the real-ride checklist in the real-ride checklist. Several items below may turn out to be non-issues, and others may appear that nobody has thought of.

---

1. Carried over from v2 — the honest debt

Elevation gain accuracy · needs real data first

~30 m of phantom gain per ten stationary minutes against synthetic ±8 m uniform noise. Real GPS altitude error is correlated rather than uniform, so the true behaviour is unknown.

The current implementation is a 15-sample moving average plus reversal hysteresis (see ARCHITECTURE.md). A naive version reported 1498 m over a parked bike, so the guard rails matter.

Do not tune this blind. Record a flat ride, check whether the reported gain is plausible, and only then adjust. If it needs work, options are a longer smoothing window, a larger threshold, or using barometric pressure where available (much more accurate than GPS altitude, and most phones have the sensor).

Compose UI tests · the largest coverage gap

Zero UI tests across six screens. Everything was verified by manual screenshot. Worth covering: navigation record→trips→detail→back, rotation/state retention, empty states, NotFound, chart degradation below two points, selection mode enabling Merge only at two.

Unmeasured, and probably should be

• Battery drain over a multi-hour ride — never measured, and it is the thing most likely to make the app unusable in practice
• Map memory across repeated navigation — the osmdroid lifecycle is a known hazard and the wiring was never leak-tested
• GPX import into Strava/Garmin — validated against an XML parser, but schema validity does not guarantee a consumer accepts it

---

2. Live map on the recording screen — decision reversed

v2 deliberately shipped no live map, on the reasoning that the phone rides in a pocket. Dylan has since asked for one — for visual appeal, and because people may mount the phone on the handlebars to watch the route live. Treat the v2 stance as superseded.

The handlebar case changes the premise, not just the feature. v1 and v2 were both built around "start it, pocket it, stop it". A mounted phone is a different product with different constraints, and it is worth deciding explicitly whether that becomes a first-class mode:

• Screen on for the whole ride — battery goes from "a background service" to "a service plus a lit screen plus continuous map rendering". Measure before committing.
• Sunlight legibility — the current dark theme is chosen for glanceability, but daylight behind a visor is a different problem.
• Glove-sized targets — already partly handled (72dp buttons); a map needs the same care.
• Keep-screen-awake handling, and what happens on a call or notification.

What still holds regardless:

• TrackingService must never reference a map. Rendering belongs to the Compose lifecycle of a visible screen, not the service.
• No tile fetch or redraw while backgrounded, even in mounted mode.

---

3. Activity type per ride

Promoted out of the ideas list, because it turned out to be a positioning decision rather than a feature. See LAUNCH.md.

Everything user-facing says motorcycle, but the recording pipeline never did: it records positions, speeds and altitudes, and nothing in it cares what you are sitting on. Dylan already rides both bikes and motorcycles.

Why it matters beyond the feature: it decides the store category, the screenshots and who ever finds the app. Cyclists are a far larger audience than motorcyclists and are already used to paying for ride apps. That question is much cheaper to settle before a store listing exists than after.

Schema

An activity column on Trip, stored as a text enum like TripState:

motorcycle · bicycle · scooter · skateboard · running · walking · other

This is the first real migration the port will ship. The destructive fallback is gone for good, so it needs m.addColumn(trips, trips.activity) with a default of motorcycle for existing rows, schemaVersion bumped to 2, and a migration test that opens a v1 database and asserts the rides survive. Getting that path right once matters more than the feature does — every later schema change depends on it.

Type should drive defaults, not just labels

This is where the real value is, and it is easy to miss:

Setting	Why it differs
Speed noise floor (1.5 km/h)	Fine for a motorcycle; wrong for walking, where real movement lives near it
Speed histogram bucket (10 km/h)	Useless for running — everything lands in one bucket. Wants ~1 km/h.
Accuracy gate (50 m)	A motorcycle at speed can tolerate looser fixes than a walker
Elevation smoothing window	Tuned at 15 samples for 2 Hz road speed; a slower activity covers less ground per sample
Map fit zoom	A 2 km walk and a 200 km ride want different defaults

Treat these as a per-activity profile rather than scattering if (activity == …) through the code.

Do not put a picker in front of Start

The founding premise is press-and-go with gloves on. A modal asking "what are you doing?" before recording begins would undo that.

Better: default to the last activity used, and make it editable on the trip detail screen afterwards, next to rename. Most people do the same thing most days, and the one time they don't, they can fix it after.

Follow-ons, once the column exists

• Filter and group the trips list by activity
• GPX <type> on <trk> — Strava and Garmin read it, so an exported ride imports as the right activity instead of defaulting to something wrong
• Per-activity totals, if a stats screen ever appears

---

4. Route drawing and planning

Drop pins on a map, have the shortest path between them resolved along real roads, and get distance and estimated ride time back before setting off. Save it, ride it later.

This is pre-ride planning: a genuinely new mode alongside recording, not an extension of it. It needs no ride in progress, no server of its own, and could be built entirely standalone — which makes it the largest thing in v3 that carries no dependency on anything else.

It needs a routing engine, and that is the whole decision

Straight lines between pins are easy and reuse haversineMeters. Road-snapped shortest path is not — it needs a routing service over OpenStreetMap data:

Option	Trade-off
Public OSRM demo server	Free, zero setup, not for production use and rate-limited. Fine for prototyping only.
Self-hosted OSRM	Fast, well-understood. Needs a machine and a regional OSM extract (a province is a few GB).
GraphHopper	Self-hostable, good cycling and motorcycle profiles, friendlier ETAs
Valhalla	Best multi-modal profiles, heavier to run
Commercial (Mapbox, Google)	No ops, per-request billing, an API key in the app

Profiles matter here more than usual. A motorcycle route and a bicycle route between the same two pins are genuinely different, and cycling engines avoid motorways while motorcycle riders often want the twisty road rather than the fast one. This is where section 3 pays off — the activity picks the routing profile.

Estimated time is a promise, and an easy one to get wrong

Routing engines return a duration based on posted speed limits. That is not how long you take. Once there is real ride history, a far better estimate comes from the rider's own average moving speed for that activity — which the app already stores on every Trip.

Show the engine's estimate first, and replace it with a personal one when there is enough history to justify it.

Data model

Route + Waypoint, separate from Trip — a plan is not a recording, and conflating them would put unridden kilometres into ride totals. The resolved geometry (the polyline the engine returns) should be cached on the Route so a saved plan opens offline and does not re-bill a routing request every time it is viewed.

Follow-ons

• Follow a planned route on the live map while riding (needs section 2)
• Compare a recorded ride against the plan afterwards — where you deviated, how the real time compared to the estimate
• Export a plan as GPX so it loads into a dedicated sat-nav

---

5. Smaller items

Item	Notes
Trip splitting	Merge exists; split does not. The natural counterpart.
SAF export	Dropped in T16 as unnecessary — share sheet covers it. Add if a real need appears.
Auto-pause	Detect a stop and pause automatically. Rejected in v2 as unreliable in traffic; revisit only with real ride data showing it would help.
Distance units	Metric only, hardcoded. Trivial to add a preference.
Settings screen	None exists. Config has endpoint, deviceId, mapEnabled — the map toggle currently lives on trip detail because one switch did not justify a screen.
Offline tile pre-download	osmdroid caches what it renders; a mountain ride with no signal shows blank tiles. Respect OSM's usage policy — no bulk prefetch of their public servers.
Notification live stats	Show distance/duration in the ongoing notification, readable without unlocking.
Crash reporting	None. A recorder that dies mid-ride currently leaves no trace beyond logcat.

---

6. Ideas

Terse on purpose. Unshaped, to be consolidated later.

• Live map while recording. More visually appealing than a numbers screen. Reverses the v2 decision — see section 2 for the constraints that survive it.
• Pick a real theme. The current look is functional dark + safety orange, chosen to match the icon. Decide on an actual visual identity and push the UI toward something polished rather than merely clean.

Threads running through these

Sign-up, cloud backup and group ride are one programme, not three: they all need a server, identity, and a privacy stance. They have therefore been moved out to v4 below, and should be scoped together or not at all.

Activity type (now section 4) and theming are independent and much cheaper — either could ship alone, and activity type is the natural first v3 task because it forces the migration path to be proven while the stakes are still low.

Live map, handlebar mounting and route following cluster: all three assume a visible screen during the ride, and all three want the same map component. Route planning (section 4) is the odd one out — it needs no ride in progress at all and could be built entirely standalone.

---

---

v4 — everything that needs a server

Split out from v3 deliberately. These three are one programme, not three items: each needs a server, an identity system, and a privacy stance, and none of them is worth building alone. Nothing in v3 depends on any of them.

Live group rides

Invite someone to a ride and see both of you on the map, live.

This was in the v1 brief's goal statement, so it has been the intended destination all along — it was deferred from v2 as "needs real server work", and that is still the honest summary.

Already in place, from v1

• TelemetryUploader — batched POST, retry, offline-safe, and structurally unable to stall recording
• The synced column and backlog semantics
• trip_id / segment_id carried per point, so a server can reconstruct rides and their pauses
• Config.deviceId — a stable per-install id, enough to distinguish riders

Missing

• A server. Nothing exists. This is the actual work.
• UI for the endpoint — still only reachable via Config.setUploadEndpoint()
• Auth, rider identity, group membership, invitations
• Other riders drawn on the map, which also needs the live map from v3 section 2
• Live delivery. The current uploader is a batched backlog drain on a 30 s timer, which is right for archiving and useless for watching someone move. Live positions want a websocket or similar, running alongside the existing uploader rather than replacing it — the batch path is what guarantees no fix is ever lost.

The decisions that are not technical

• Location sharing is consent, not a feature. Who can see you, for how long, and how does it stop? Sharing that outlives the ride is a privacy incident waiting to happen.
• Ride invitations mean handling someone declining, leaving mid-ride, or losing signal for twenty minutes — the map has to say "last seen 8 minutes ago", not silently freeze them in place.
• This is the point where Rippr stops being a local-only app and starts holding other people's location data.

Accounts and sign-up

Prerequisite for everything else here. Note that Apple requires in-app account deletion for any app offering account creation, and that a location app with accounts inherits real obligations — see LAUNCH.md.

Paid cloud backup

Ongoing storage of rides over time, and the only monetization model that justifies recurring money — because it is the only one with recurring costs. Needs accounts first, plus decisions on hosting, pricing, and what happens to someone's history when they stop paying.

---

Things that must not regress · all versions

Hard-won and easy to undo by accident. Each has a comment in the code explaining why.

1. The unbounded Channel + single batched writer. Do not write to the database from the location callback.
2. Points stamped with tripId/segmentId at creation. Refactoring this into a write-time lookup breaks the pause guarantee silently.
3. Recording state derived from the database. Never reintroduce an in-memory flag.
4. fallbackToDestructiveMigration() stays removed. Any schema change ships a Migration against app/schemas/com.rippr.data.AppDatabase/2.json.
5. Decimation is render-only. It must never reach storage or export.
6. RipprTheme's Surface. It sets LocalContentColor; without it, text without an explicit colour renders black-on-black and disappears.
7. The map zoom clamp. zoomToBoundingBox ignores maxZoomLevel; a short ride will render an empty grid without it.
8. Instrumented tests use in-memory databases. One previously wiped the real device database in setUp.

---

Reading order for picking this up cold

1. ../README.md — what the app is and its current state
2. ARCHITECTURE.md — why it is built this way
3. ~/dojo/rippr/docs/v2/PROGRESS.md (native repo) — every bug found during v2 and how
4. the real-ride checklist — especially "What the emulator cannot verify"
5. ~/dojo/rippr/docs/DEVELOPMENT.md (native repo) — when you actually need to build something

The v2 planning approach worked well and is worth repeating: one document per task with goal, context, design, acceptance criteria and risks, written before implementing, plus a running progress log recording what actually went wrong. Several bugs were caught precisely because the risk had been written down first — and one (the osmdroid lifecycle) was written down and then walked into anyway, which is its own lesson.