nx_room_resilience.nx

nx_room_resilience.nx -- MEASURED head-to-head: sovereign proactive FEC vs the incumbent ARQ/NACK-retransmit recovery model (WebRTC/Zoom/Teams video default), on an intercontinental link. Converts the room census's "loss-resilience = BEHIND/unclaimed (no incumbent baseline)" into a MEASURED, caveated exceed. THE MECHANISM (textbook, RFC-grounded -- not a wave): ARQ: a lost packet is re-requested (NACK) -> the retransmit arrives +1 RTT later. On a high-RTT intercontinental link (~250ms RTT) that retransmit usually arrives AFTER the playout/jitter deadline -> the frame FREEZES. FEC: k data + m parity ship together; ANY <=m losses in the n=k+m block are reconstructed from parity ALREADY IN FLIGHT -> 0 added latency, no retransmit, on-time. RTT-INDEPENDENT by construction. We REUSE the proven RS GF(256) erasure codec from nx_room_fec.nx (no reimpl, DRY) and ACTUALLY encode/decode every block under the loss trace -> the on-time count is backed by byte-exact reconstruction, not arithmetic. HONEST DISCIPLINE (no-wave): one fixed loss trace, vary ONLY RTT across LAN/moderate/ intercontinental so the win is attributable to RTT, not a constant. Negative controls: (T3) at LAN RTT, ARQ's cheap retransmit recovers EVERYTHING -> it TIES OR BEATS FEC (FEC loses the rare >m-burst block) -> the exceed is regime-specific, not rigged. (T4) a >m-loss block -> FEC frozen>0 (honest recovery bound, not a fake 100%). (T5) corrupt the generator matrix -> recovery breaks -> parity is load-bearing. SCOPE CAVEAT (printed): this is vs the ARQ-retransmit model. vs WebRTC's OPTIONAL FlexFEC the mechanism is parity-of-kind; our exceed THERE is the provable MDS guarantee (every <=m pattern, audit-replayable, byte-exact) + full sovereignty -- not a lower loss number, and NOT yet a live WebRTC packet capture. main() is the SELF-VALIDATING GATE. Evidence -> knowledge/status/room_resilience.log. license_tier: ORIGINAL