PROTOCOL NOTES / 12
Scenarios & evidence
Follow complete scenarios and read evidence at the right level.
Android and iOS software implemented · Apple Wi-Fi Aware, Bluetooth LE and mixing tested on hosts and emulators only · Physical qualification pending
Twenty-two views of the same system
Each scenario connects the starting state, ordered events, ownership, failure branches, and evidence. Read them as worked examples alongside the explanatory chapters. Scenarios marked as tested on hosts and emulators only describe the Apple, Bluetooth LE and mixing extension, which has not yet run on physical phones.
- Organizer creates a ride and begins discovery
- Invitation import, discovery and first connection
- NDP through IPv6, TLS, greeting and hello
- Join approval, index assignment and signed roster
- Multiple direct and relayed admissions
- Admission closure, source descriptors and readiness
- Local helmet audio activation and first frame
- Single-hop and multi-hop authenticated media
- Silence, speaking, feedback and voice-activated transmission
- Graceful departure and organizer-local termination
- Network failure, half-open detection and repair
- Abrupt relay disappearance and descendant reconnection
- Partition, temporary root and later merge
- Parent improvement, handoff, abort, rollback and drain
- Duplicates, stale media, hop exhaustion and rejected membership
- Background operation, interruption, suspension and recovery
- Capacity exhaustion and break-before-make fallback
- Shared-NDI ownership and independent teardown
- Apple device joins over Wi-Fi Aware pairing and advert discovery · implemented software; host and emulator verification only
- Apple and Android bridged over BLE with mixing · implemented software; host and emulator verification only
- Android 10 leaf over a single BLE link · implemented software; host and emulator verification only
- Parent change from a BLE edge to Wi-Fi Aware with make-before-break · implemented software; host and emulator verification only
Observe without recording the conversation
RideWeave persists bounded local diagnostics for rides, failed join attempts, and headset checks. A background writer records events, periodic counters, and device measurements. Ordinary records exclude speech and packet content. Shared reports also omit names, ride labels, invitations, keys, and raw network addresses. Bluetooth LE and mixing records add link and mixer counters, never Bluetooth device identifiers or audio.
Correlation pseudonyms support comparison across phones within a ride; they are not an anonymity guarantee against someone who already knows the source identities. Quota eviction can remove detail while retaining summaries. Recorder gaps and missing observations must remain explicit instead of being filled with invented values.
Keep evidence at its actual level
| Evidence | What it can establish |
|---|---|
| Deterministic host | State transitions, authorization and ownership invariants |
| TLS/UDP loopback | Real socket framing and software duplex behavior |
| JVM / Android emulator | Adapter policy, JNI, UI and lifecycle behavior |
| Spike measurements | Bluetooth LE throughput, mixing cost and stack defects on specific phones: design input, not qualification |
| Physical phones and helmets | Wi-Fi Aware and Bluetooth LE range, motion, wind acoustics, coexistence with headset calls, heat and battery: still requires qualification |
A requirement mapped to a test is not itself a passing execution. A scenario tested on hosts and emulators only has no physical evidence. Recorded test results apply only to the revision they were run against. The documentation describes current implementation; cited software evidence does not qualify later radio behavior.
Deliberate boundaries
The Android and iOS apps are implemented. The Apple Wi-Fi Aware path, Bluetooth LE links and bridge mixing are implemented software tested on hosts, emulators and the iOS simulator only; Wi-Fi Aware between Apple and Android phones needs an Android phone with NAN Pairing. Hotspot fallback, remote member eviction, and global authenticated termination are unimplemented. Moving-group range and reliability, Bluetooth LE range, coexistence with headset calls, screen-off survival, helmet acoustic quality, mixed-speech quality, and battery duration are not qualified. Members share source keys, and a bridge re-encodes the voices of its side. Parent coordination rejects known cycles but is not distributed consensus. Voice-activated transmission is on by default and can be turned off; its acoustic quality is unqualified.
Authority: Scenario catalogue; diagnostics; human qualification. Public examples are synthetic. This guide is maintained against the private implementation; it does not imply access to its source.