Skip to content

TODO.md

Latest commit

 

History

History
275 lines (226 loc) · 55.2 KB

TODO.md

File metadata and controls

275 lines (226 loc) · 55.2 KB
Raw

xng — Implementation TODO

Derived from COMPARISON_RESEARCH.md (audited 2026-06). This is the actionable backlog; the research doc is the rationale/sourcing for each item. Check items off (- [x]) as they land. Progress so far (on feat/per-decoder-airframes-feeding): the FEED per-decoder Airframes feeding feature, and a per-mode decode batch across ADS-B, AIS, VDL2, HFDL, STD-C, Aero, Iridium and ACARS — see the checked items below.

How to use this file

  • Checkbox — - [ ] open · - [x] done. Check a parent only when all its subtasks are done.
  • Item ID — CATEGORY-N (top-level) and CATEGORY-N.M / CATEGORY-N.M.K (subtasks, tree). IDs are stable — never renumber; if an item is dropped, mark it ~~struck~~ (dropped: reason) rather than deleting.
  • Tags — ★ quick win (high value / low, clean-room effort) · ◆ big bet (large, high-leverage) · ⚑ correctness (possible live bug, verify first) · (A1)/(B)/(C)/(D) = source section in the research doc.
  • Categories: FEED feeding feature · XM cross-mode foundations · ACARS VDL2 HFDL AERO STDC IRID AIS ADSB per-mode · ECO ecosystem/tooling · NEW new modes · VERIFY research/correctness checks.

Suggested sequencing (recommended order, not binding)

Feeding follows decode. A mode's Airframes feed (FEED-2.x, and enabling that mode under [outputs.airframes]) must land after the mode's decode is complete enough to populate the ingest's native wire format — feeding partial/wrong records to Airframes is worse than not feeding. The only feeding work with no decode dependency is the mode-independent scaffolding FEED-1 (config schema, resolver, validation), which may be built early.

  1. Foundations — XM-1 (shared signal-quality), XM-2 (entity model + one position→SBS adapter); optionally FEED-1 config scaffolding (no decode dependency).
  2. Per-mode decode quick wins — ACARS-2, ADSB-1/ADSB-2, AIS-1/AIS-3, AIS-4, STDC-1, HFDL-1, IRID-8/VERIFY-1 (⚑).
  3. Big bets — VDL2-1+VDL2-2 (ASN.1/PER core + CLNP/COTP → ATN-B2 ADS-C), ECO-1/ECO-4, XM-4 (MLAT/TDOA).
  4. Per-mode Airframes feeds (only after the mode's decode is solid) — FEED-2.1 after VDL2 decode, FEED-2.2 after HFDL, FEED-2.3 after AIS, FEED-2.4/ACARS-5.1 after ACARS text decode.
  5. New modes — by priority: NEW-P0-* → NEW-P1-* → NEW-P2-* (each new mode also gets its feed only after its own decode lands).

FEED — Per-decoder Airframes feeding feature (C)

The explicitly-requested feature: per-mode ingests, per-mode native wire format, per-decoder disable, per-mode/per-session station-id override. asf-2.0 is the exempt multiplexing path. Verified ingest table: ACARS UDP :5550 (acarsdec JSON), VDL2 UDP :5552 / TCP :5553 (dumpvdl2 JSON), HFDL UDP :5556 (dumphfdl JSON), AIS HTTP :5599 (AIS-Catcher PROTOCOL AIRFRAMES). IMSL/IRDM/STD-C/Aero-C/ADS-B → asf-2.0 only.

Ordering: FEED-1 (config scaffolding) has no decode dependency and may be done early. FEED-2.x per-mode serializers must come AFTER the corresponding mode's decode is complete enough to fill that ingest's native format — see the ⏳ after … notes below.

  • FEED-1 Config schema + plumbing for per-mode feeding (station.rs, main.rs, runtime.rs) — done on feat/per-decoder-airframes-feeding
    • FEED-1.1 Add [outputs.airframes] block (enabled, station-id, auto-suffix) + per-mode sub-tables acars/vdl2/hfdl/ais → AirframesToml / AirframesModeToml{enabled,station_id,host,port}
    • FEED-1.2 Add SessionToml.feed: Option<bool> (per-decoder kill-switch) and airframes_station_id (per-session id override)
    • FEED-1.3 Per-session AirframesTarget resolver with documented precedence (session feed=false → mode-block enabled → id: session override ⟶ mode-block id ⟶ auto-suffix(base,mode) ⟶ base) — station::airframes_router
    • FEED-1.4 auto-suffix(base, mode) station-id derivation (strip-then-reappend a known mode suffix) ★ quick win
    • FEED-1.5 Validation (deny_unknown_fields rejects adsb/imsl/irdm/stdc/aeroc sub-tables; feed=false precedence; duplicate id+endpoint warning; base id always resolvable since top-level station-id is mandatory)
    • FEED-1.6 OutputConfig.airframes: Option<AirframesRouter> replaces the shared udp Airframes coupling; generic udp kept for other sinks
    • FEED-1.7 --feed-airframes back-compat preserved (ACARS-only, id verbatim — live station unchanged); --feed-airframes-modes flag deferred
  • FEED-2 Per-mode native serializers (each returns None for non-matching bodies) — ⏳ sequence each AFTER its mode's decode is complete enough to fill the native format — ✅ ACARS (2.4) + VDL2 (2.1, dumpvdl2-verified) serializers done; HFDL (2.2)/AIS-Catcher (2.3) deferred (see subitems)
    • FEED-2.1 dumpvdl2_json.rs — dumpvdl2 decoded:json emitter → VDL2 UDP :5552 — ✅ 2026-06-18: src/outputs/dumpvdl2_json.rs emits the nested {vdl2:{…avlc:{…acars:{…}}}} for VDL2 ACARS-over-AVLC; to_message carries the AVLC wrapper via core.app["_vdl2_link"]; wired in the airframes router (has_serializer + serialize_datagram). Verified field-for-field against real dumpvdl2 2.6.0 captured on the vendored off-air fixture (golden test, HB-IJW downlink). Non-ACARS AVLC/XID frames + the TCP :5553 variant remain.
    • FEED-2.2 dumphfdl_json.rs — dumphfdl decoded:json (nested) emitter → HFDL UDP :5556 — ⏳ after HFDL-1 (HFNPDU records) + HFDL-4 (positions) — audited 2026-06-18: partially oracle-gated. dumphfdl 1.7.0 IS installed; the vendored 8s fixture decodes to one SPDU squitter, whose scalar schema I captured authoritatively (src/spdu_version/rls/iso/change_note/frame_index 2397/frame_offset/min_priority/systable_version 52 + gs_status). The scalar squitter fields map faithfully to xng's parse_spdu details, BUT (a) xng's SPDU decode captures only local-GS freqs + neighbor2 freqs + neighbor3 id (no neighbor utc_sync / neighbor3 freqs), so the full gs_status array can't be reproduced byte-faithfully; (b) the valuable feed paths (positions / enveloped ACARS) have NO authoritative captured schema here — the 8s fixture has no aircraft traffic and the 48k bench capture didn't decode under dumphfdl in-session. Squitter-only feeding is low value, so deferred until a richer dumphfdl capture (with HFNPDU positions + LPDU ACARS) is vendored to verify the nested hfnpdu/lpdu/acars schema. Router has the Mode::Hfdl => :5556 target ready; serialize_datagram returns None for HFDL pending that capture. — ⏸ DEFERRED: dumphfdl JSON verifiable only on the squitter (the vendored fixture is squitter-only); the valuable HFDL ACARS/positions need a traffic-bearing capture vendored to tests/data — deferred (recon-confirmed 2026-06-18)
    • FEED-2.3 AIS-Catcher PROTOCOL AIRFRAMES HTTP-POST task (batched interval) → HTTP :5599 (proto=http path, distinct from datagram path) — ⏳ after AIS-1/AIS-3 field decode — ⏸ DEFERRED: AIS-Catcher not installed + the PROTOCOL AIRFRAMES :5599 JSON unverifiable without it — deferred (skip-don't-fake)
    • FEED-2.4 Wire existing acarsdec_json.rs (ACARS :5550) under the new router via format_acarsdec_with_station (station-id stamped at format time; provenance untouched). Richer fields ⏳ after ACARS-2/ACARS-4.1 → ACARS-5.1
  • FEED-3 asf-2.0 (exempt) handling
    • FEED-3.1 asf-2.0 stays on its own path (canonical station_ident), never routed through the per-mode resolver; runs alongside per-port legacy feeds
    • FEED-3.2 Modes with no public per-port ingest (IMSL/IRDM/STD-C/Aero-C/ADS-B) are skipped by the router (fed via asf-2.0); documented in airframes.rs + example config
  • FEED-4 Open items to confirm (see VERIFY-13) — ✅ 4.1 confirmed+implemented (dumpvdl2 decoded:json = FEED-2.1); 4.2 external (see subitem)
    • FEED-4.1 Confirm VDL2 ingest accepts dumpvdl2 decoded:json (emit that, not vdlm2dec flat JSON) — ✅ confirmed + implemented: VDL2 ingest format = dumpvdl2 decoded:json (FEED-2.1, verified vs dumpvdl2 2.6.0)
    • FEED-4.2 Track whether Airframes exposes public IMSL/IRDM/STD-C ports + settled Iridium feeding mechanism — ⏸ DEFERRED: external: whether Airframes exposes public IMSL/IRDM/STD-C ports + the settled Iridium feeding mechanism — not answerable from the repo

XM — Cross-mode structural foundations (B — cross-mode bets)

Highest-leverage: each closes a gap flagged separately in many modes.

  • XM-1 Shared per-burst SignalQuality{signal dBFS, noise dBFS, SNR, CFO-ppm, fec_corrected} schema populated by every demod ◆ big bet (closes output gap in ACARS/VDL2/HFDL/STDC/Iridium/Aero/AIS) — ✅ ADS-B slice DONE 2026-06-18: AdsbFrame.noise_dbfs stamped from the demod's power EMA after scanning, to_message fills noise_db + snr_db (= rssi − noise). Oracle decision: a dump1090-rssi numeric oracle was rejected (no calibrated floor reference) in favour of an internal-consistency + monotonic-SNR-vs-added-noise test; benchmark unchanged (adsb_modes1 = 323, demod only reads the EMA). HFDL already populates snr/freq_skew/fec. ❌ remaining: envelope-only modes (acars/vdl2/ais/stdc/iridium/aero/uat/…) need a genuine noise-floor estimator before they can honestly fill noise_db — skip-don't-fake until calibrated captures exist; CFO-ppm where a tracker exists (VDL2-7).
  • XM-2 Unified cross-mode entity model ◆ big bet — ✅ 2.2 position→SBS/Beast/map adapter done; 2.1 TrackStore deferred (see subitem)
    • XM-2.1 Entity{kind: aircraft|vessel|sat|beacon, id: ICAO|MMSI|IMEI|HexID, positions[], identities[], source_modes[]} track store — ⏸ DEFERRED: unified Entity TrackStore is a large refactor; the per-type maps + the XM-2.2 adapter cover current needs — deferred
    • XM-2.2 One mode-agnostic position → SBS/Beast/map adapter keyed on the entity (replaces the 4 separate per-mode wirings; unlocks HFDL/Iridium/Aero positions to SBS/tar1090/VRS) — ✅ map: dashboard merges ModeS/UAT/HFDL/ACARS aircraft by ICAO. ✅ SBS + ✅ aircraft.json (ECO-4) + ✅ Beast (2026-06-18): the shared outputs::aircraft::AircraftFix extractor feeds all three. Mode S wraps its raw frame; UAT 978 / HFDL positions are SYNTHESIZED into DF17 ES frames (xng-mode-adsb::synth: CPR position pair + Q=1 altitude + callsign + TC19 ground-velocity encoders, each round-trip-verified through the crate's own decoder; 2026-06-18 added velocity) so raw-Beast consumers (tar1090/readsb) plot them with a velocity vector too. ❌ remaining (out of XM-2.2 scope): ACARS/Aero ADS-C positions key on flight/reg (no ICAO hex); Iridium positions are satellites/terminals (not aircraft).
  • XM-3 Shared ICAO/registration resolver (tail↔ICAO↔reg↔operator↔dbFlags mil/PIA/LADD) serving ACARS/VDL2/HFDL/Aero/Iridium/ADS-B — ⏸ DEFERRED: ICAO/registration resolver needs a bundled aircraft DB (data dependency); partial dbinfo exists — deferred
  • XM-4 Multi-receiver geolocation primitive over the asf-2.0 fan-in (Iridium TDOA + Doppler self-position + ADS-B MLAT as one engine keyed on the entity) — ⏸ DEFERRED: multi-RX geolocation/MLAT/TDOA needs multi-receiver infra + synchronized captures — deferred
  • XM-5 Cross-mode dedup keyed on (entity_id, content_hash, time-window) (covers ADS-B multi-RX, AIS unique on, ecosystem fan-in) — ✅ content-dedup built (AisGate, generic (entity,content,window) keying, reusable); the cross-mode ingest-fan-in application is deferred (=ECO-9)
  • XM-6 Cross-mode distress/emergency overlay (ADS-B 7500/7600/7700 + TC28 + AIS-SART/MOB + STD-C EGC distress + future DSC / COSPAS-SARSAT 406) → one alerting surface — ✅ data surface DONE 2026-06-18: /api/state alerts[] aggregates all five sources (ADS-B emergency/squawk, AIS distress-class, STD-C distress-priority, DSC distress-alert, every SARSAT 406 beacon) keyed mode:entity, 30-min linger, {mode,id,kind,seen,lat?,lon?}; distress_alert() helper reuses each mode's already-verified fields; test feeds one event/mode + routine traffic. ❌ remaining: dashboard.html panel rendering the array (UI; needs visual QA) + console banner (note: console.rs already renders distress inline per-message)

ACARS — VHF ACARS + application layer (A1)

  • ACARS-1 Label catalogue + per-label field extractors (build label→meaning table) — ✅ label families + per-label extractors complete (Q-series 1.1, raw MIN 1.2; the ACARS-2.x text decoders are the catalogue)
    • ACARS-1.1 Decode the Q-series link-test/squitter family (Q0–Q7, QA–QX)
    • ACARS-1.2 Surface raw MIN; handle 4th-char downlink-rule edge cases (see VERIFY-2)
  • ACARS-2 Embedded text-content decoders — the big user-visible gap ★ quick win — ✅ 2.1–2.5 all done
    • ACARS-2.1 OOOI: gtout/gtin/wloff/wlin/depa/dsta/eta
    • ACARS-2.2 Free-text position reports (labels 20/POS, 4J, H1 POS)
    • ACARS-2.3 AMDAR / winds-aloft / PIREP (WMO-BUFR-class schema; NOAA dcacar ref)
    • ACARS-2.4 FLIGHTPLAN / route (FPN) + Boeing/Airbus telex / structured free-text
    • ACARS-2.5 H1 #CFB/CF maintenance family (APM_REPORT, ATA, AL, FDE, ECT, FLR, LIGHTS, MIL, MPF, PAGE, WRN)
  • ACARS-3 Application-layer completion (vs libacars 2.2.1) — ✅ 3.1–3.3 done; 3.4 deferred (see subitem)
    • ACARS-3.1 CPDLC argument readers for the bracketed-template shapes + FANSPosition placeBearingDistance + RouteClearance trackDetail/routeInformationAdditional
    • ACARS-3.2 Generic sublabel/MFI extraction beyond H1 — ✅ H2 family (libacars grammar + ARINC 620-4 App C)
    • ACARS-3.3 Reassembly-status enum (assstat: complete/in-progress/skipped/duplicate)
    • ACARS-3.4 Verify MIAM CRC + vendor real off-air media-advisory captures — ⏸ DEFERRED: needs off-air media-advisory IQ captures to verify MIAM CRC (skip-don't-fake)
  • ACARS-4 Demod / robustness — ✅ 4.1/4.2 done; 4.3 deferred (see subitem)
    • ACARS-4.1 Emit noise/noise-floor + SNR (today only envelope RSSI) — see XM-1 — ✅ 2026-06-18: MskDemod tracks a silence-gated envelope-power noise EMA (noise_dbfs()); to_message fills noise_db + snr_db. Verified vs an analytic AWGN oracle (floor → 2·σ² within 1 dB, +6 dB when σ doubles), decode path bit-identical (full bench unchanged).
    • ACARS-4.2 Syndrome-table FEC (O(1) error-position lookup, acarsdec syndrom.h) — ✅ reuses xng_dsp acars_crc; parity-guided search kept as multi-error fallback
    • ACARS-4.3 Off-air acarsdec head-to-head benchmark + CI count gate (no POA row exists today) — ⏸ DEFERRED: needs a vendored POA ACARS IQ fixture for an acarsdec head-to-head gate (none in-repo; skip-don't-fake)
  • ACARS-5 Outputs — ✅ 5.1 done + 5.2 done — MQTT sink is src/outputs/mqtt.rs (publishes <prefix>/<mode> incl. ACARS); per-label counters done
    • ACARS-5.1 Emit the dropped acarsdec-JSON fields (noise, sublabel, mfi, assstat, nested app/libacars, OOOI fields) — ties FEED-2.4 — audited 2026-06-18: mostly done — sublabel, mfi (H1 in xng-acars, H2+ in xng-mode-acars) and the OOOI fields (depa/dsta/eta/gtout/gtin/wloff/wlin via oooi::decode→app) already surface on AcarsCore/the message JSON (tests pin them). ✅ assstat DONE 2026-06-18: new AcarsCore.assstat field; apply_reassembly stamps the reassembler verdict on every CRC-ok message; extern-ingest reads it from incoming acarsdec JSON; native JSON serializes it + acarsdec-JSON UDP feed emits it. ✅ noise DONE 2026-06-18 (acarsdec-JSON feed emits the noise dBFS field now that ACARS-4.1 tracks the floor). ACARS-5.1 complete — all dropped acarsdec-JSON fields now surface.
    • ACARS-5.2 MQTT output sink + per-label / per-channel counters (see VERIFY-9) — ✅ DONE: MQTT sink = src/outputs/mqtt.rs (publishes <prefix>/<mode> JSON incl. ACARS, --mqtt); per-label counters = xng_acars_messages_total{mode,freq,label} (LiveState.acars_labels), which also fixed a latent station-mode bug (/metrics served a never-updated LiveState → all-zero counters) + made per-channel stats freq-keyed.

VDL2 — VDL Mode 2 + AVLC + ATN (A2)

  • VDL2-1 Table/codegen-driven unaligned-PER ASN.1 core ◆ big bet (unlocks the next 5 at once) — ✅ 1.1/1.2/1.4/1.5 done; 1.3 ACSE done (AARQ/AARE bodies deferred — need a captured PDU)
    • VDL2-1.1 ~44 unsupported CPDLC argument types (element walk currently stops at first undecodable arg) — ✅ arg-type coverage 22→63, walk no longer halts. (deeply-nested DepartureClearance/PositionReport optionals deferred pending a real PDU)
    • VDL2-1.2 CHOICE extension alternatives / integrityCheck / PER fragmentation — ✅ X.691 §10.6 extension-addition index, §10.9.3.8 fragmented length, integrityCheck surfaced; fixed a real bug (PMCPDLC User abort-reason 13 values → 4 bits, was 3)
    • VDL2-1.3 ACSE (AARQ/AARE/RLRQ/RLRE/ABRT) + Session (X.225 SPDU) layers — ✅ all 5 ACSE-apdu CHOICE alternatives recognized/dispatched + RLRQ/RLRE/ABRT reasons decoded; ❌ AARQ/AARE SEQUENCE bodies (need a captured PDU) + Session-SPDU auto-wire from COTP (unverifiable null-framing) deferred
    • VDL2-1.4 Full Context Management (TSAP/NSAP addrs; CMContactRequest/LogonResponse/ForwardRequest/Update) — ✅ Long/ShortTsap+APAddress, CMLogonRequest/Response, CMUpdate, CMContactRequest/Response, CMForwardRequest, CM abort reasons
    • VDL2-1.5 Plain/unprotected CPDLC PDUs + forward/forward-response bodies — ✅ CPDLCAPDUsVersion1 + ATCForwardMessage/ATCForwardResponse
  • VDL2-2 CLNP + COTP → native ATN-B2 ADS-C ◆ big bet — ✅ COTP DC/ED/AK/EA/RJ TPDUs + variable part (ATN checksum 0x08, credit, ext-seq) + CLNP option / ATN-security-label TLVs + multipart CLNP reassembly + multipart COTP TSDU reassembly landed; ❌ remaining: native ATN-B2 ADS-C (2.3) — ✅ CLNP/COTP + reassembly done; 2.3 native ADS-C skip-don't-fake (no ASN.1 module/sample)
    • VDL2-2.1 Multipart CLNP reassembly + ATN security-label TLVs (traffic-type/ATSC-class/subnetwork-type) — ✅ ClnpReassembler (segment-offset, out-of-order, NSAP+data-unit-id keyed) + more-segments/error flags; ATN security-label TLVs already landed
    • VDL2-2.2 COTP DC/ED/AK/EA/RJ TPDUs + full variable part (TPDU-size, checksum, ATN checksum 0x08, credit, EOT, extended seq) + multipart COTP reassembly — ✅ TPDU types + variable part already present; added EOT-driven multi-DT TSDU reassembly (CotpReassembler, ISO/IEC 8073 §6.6) wired through decode_network
    • VDL2-2.3 Native ATN-B2 ADS-C (ADSReport/RequestContract/Accept/Reject/PositiveAck/NonCompliance over CLNP/COTP) — ⏸ skipped (skip-don't-fake): ADS-C ASN.1 module absent from docs/asn1 and no captured ADS-C-over-CLNP/COTP sample to verify against
  • VDL2-3 XID parameter completion — TG5(0x46), T3min(0x47), GS-address-filter(0x48), broadcast-connection(0x49), frequency-support-list(0xC0), airport-coverage(0xC1), nearest-airport(0xC3), ATN-router-NETs(0xC4), system-mask(0xC5), TG3(0xC6), TG4(0xC7) + ISO-8885 HDLC param set; decode autotune freq→MHz + timers→int
  • VDL2-4 X.25 completion — RESTART-REQ/CONF, facility naming, clear/reset/restart cause + diagnostic-code dictionaries, SNDCF compression facility
  • VDL2-5 AVLC polish — SABME (0x6F), expand FRMR info-field, pin one canonical FCS octet order; cross-check the v2.5.1 249-octet block-length bug (VERIFY-3)
  • VDL2-6 IDRP RIB-REFRESH + OPEN body fields + ERROR code/subcode text; ES-IS option TLVs (0x81/0x88/0xCF/0xC5)
  • VDL2-7 Demod — --max-ppm PPM/CFO reject filter; Gardner/matched timing-error detector for weakest-burst acquisition — ✅ 2026-06-18: the preamble fit's per-symbol carrier-rotation (CFO slope) is now surfaced as freq_skew_hz on Burst/Vdl2Frame/SignalQuality (the prerequisite signal-line field for VDL2-8); opt-in --max-ppm reject (also max-ppm in the station TOML) skips bursts whose preamble CFO exceeds the limit. Verified vs injected CFO ground truth (skew tracks ±offsets; 1 ppm gate rejects a 400 Hz burst that 10 ppm/no-gate keep); default off, bench unchanged (vdl2_offair 44). ❌ Gardner TED skipped per the audit — the coherent LS preamble-fit timing already subsumes it (any non-improving TED is a no-op against the ≥42 gate).
  • VDL2-8 Outputs — full per-message signal line (noise+SNR+ppm, see XM-1); --extended-header [S][L][F][#]; aircraft enrichment (--addrinfo/--bs-db); msg-filter grammar; ZMQ publisher; raw-AVLC archive; --dump-asn1 — ✅ Part A signal line DONE 2026-06-18: SignalQuality now carries rssi_db + freq_skew_hz (VDL2-7) + EVM-derived snr_db (from the per-symbol decision residuals; ordering-verified). noise_db deliberately NOT exposed (channel-FIR-scaled floor + slow cold-start → not cleanly verifiable; skip-don't-fake). ❌ remaining (separate output features): --extended-header, --dump-asn1, raw-AVLC archive, msg-filter grammar, ZMQ publisher; --addrinfo/--bs-db enrichment is oracle-gated on XM-3 + a BaseStation DB.

HFDL — HFDL / ARINC 635 (A3)

  • HFDL-1 HFNPDU full-record decode ★ quick win (pure parsing)
    • HFDL-1.1 Split 0xD1 performance-data from the 0xD5 handler; decode the full 47-byte perf record (version, flight_leg, gs/freq_id, counters, freq_change_code)
    • HFDL-1.2 Decode 0xD5 per-GS {gs_id, prop_freqs, tuned_freqs} arrays
    • HFDL-1.3 0xD2 system-table-request field parse; name 0xDE delayed-echo; name 0x2F logon-denied with reason table
  • HFDL-2 System table + GS naming
    • HFDL-2.1 --system-table load/save persistence (cold-start enrichment) — ✅ serde save/load API. (--system-table CLI flag = follow-up)
    • HFDL-2.2 Config-driven GS name file (IDs up to 127; fill the 12 hardcoded holes) — ✅ roster pinned id-for-id to dumphfdl systable.conf (1–11+13–17 assigned; only id 12 was a real hole; 18–127 unassigned upstream)
  • HFDL-3 Aircraft-ID→ICAO cache (ac_cache, --aircraft-cache-ttl) — ties XM-3
  • HFDL-4 Positions — ✅ lift {lat,lon,utc,flight} into MessageBody::Hfdl details["position"] (perf-data 0xD1 / freq-data 0xD5 HFNPDUs; ICAO back-filled from the logon cache) + dashboard map plot merged by ICAO with 1090/UAT/ACARS (XM-2.2) + ✅ SBS feed + ✅ Beast feed (HFDL positions emit :30003 BaseStation lines and synthesized DF17 ES frames on :30005 via the shared AircraftFix adapter, XM-2.2); ❌ --freq-as-squawk
  • HFDL-5 Demod — ✅ fec_corrected + per-frame SNR/signal/CFO populated; LMS-tap count verified (VERIFY-5); ❌ remaining: FFT polyphase channelizer
  • HFDL-6 — ✅ SPDU rls_in_use/iso8208_supported flags + fuller system-table decode; ❌ remaining (output-side): Prometheus/StatsD expansion + noise-floor gauge, per-slot assignment map, zmq/file-rotation, raw-frame modes

AERO — Inmarsat Aero L + C-band bursts (A4, A5)

  • AERO-1 Full P-channel SU classifier (0x00–0x76; JAERO names the table, xng decodes ~6) — ✅ 1.1–1.4 done (the classified SU set; remaining 0x00–0x76 entries are reserved/unused)
    • AERO-1.1 Log-on/log-off control SUs (0x10–0x17) → structured AES↔GES session events
    • AERO-1.2 Call_announcement 0x21, T_channel_assignment 0x51
    • AERO-1.3 AES system-table broadcast (satellite_identification 0x0C, GES_beam_support 0x07, Psmc/Rsmc 0x05, broadcast_index 0x0A)
    • AERO-1.4 EIRP-table 0x28, P/R-control-ISU 0x40, T-control-ISU 0x41, RQA 0x61, RACK/TACK 0x62, short-LSDU 0x74/0x76
  • AERO-2 Satellite/beam resolution → tag every message with the resolved satellite (self-configuring, L-band analogue of HFDL systable)
  • AERO-3 R/T-channel named control set (access-request/call-progress/telephony-ack/RQA/ACK); verify SEQINDICATOR→(k,n)
  • AERO-4 Interpret 16-bit P-channel frame header (formatid/superframe/framecounters) for superframe lock + AFC/DCD state machine — ✅ 16-bit P-channel header parsed/exposed (format-id/superframe/frame-counters) + ✅ SuperframeLockStateMachine (acquire at N=3 in-sequence headers, lose at M=4 misses) with coupled CarrierState→dcd/afc_locked; lock snapshot surfaced in Aero details (enriches c-channel-assignment + emits a p-channel-status body for otherwise-undecoded P-channel events). Synthetic frame-counter-sequence oracle + real off-air 600 bps no-regression
  • AERO-5 C-channel voice → WAV (AMBE decode behind a feature flag) + older AERO-H LPC path; wire CChannelDecoder into a runtime Mode — ⏸ DEFERRED: C-channel voice→WAV needs an AMBE codec behind a feature flag (no open AMBE) — large, deferred
  • AERO-6 Demod — coherent 600/1200 path (close the ~2 dB gap to JAERO); populate fec_corrected; BER-vs-SNR curve — ✅ decision-directed coherent carrier path + fec_corrected (re-encode count) + synthetic BER-vs-SNR test
  • AERO-7 Outputs — expand MessageBody::Aero bodies (log-on/satellite-id/system-table/call); aircraft-DB enrichment + Aero position plotting — audited 2026-06-18: the message-body expansion is done — all classified P-channel SUs (log-on/off 0x10-17, sat-id 0x0C, system-table 0x05/07/0A, call-announcement 0x21, channel-assign 0x31-34/0x51, P/R/T control 0x40/41/51, EIRP 0x28, RQA/ACK 0x61/62, short-LSDU 0x74/76) surface as structured MessageBody::Aero{kind,details}, JAERO-layout + end-to-end tested. Remaining: aircraft-DB enrichment + map position plotting = output/dashboard, depends on XM-2.2 (cross-mode entity model + position→SBS adapter).
  • AERO-8 Aero-C consolidation (A5)
    • AERO-8.1 Fix Mode::AeroC mislabel (to_message hard-tags AeroL) + wire scan-plan/dispatch/feed — OR fold aero-c into aero as a PHY-selected burst sub-decoder
    • AERO-8.2 Typed SU classifier shared across P/R/T + bit_rate/channel tag; docs/notes/AERO.md written (C-channel dl2 descrambler still TODO)
    • AERO-8.3 10.5k A-QPSK burst path for aero-c — ⏸ DEFERRED: 10.5k A-QPSK aero-c burst path needs a real aero-c sample to verify — deferred
  • AERO-9 SwiftBroadband-Safety — roadmap watch only (FANS successor; no open PHY decoder today); no implementation — ⏸ DEFERRED: SwiftBroadband-Safety is roadmap-watch only — no open PHY decoder exists, nothing to implement

STDC — Inmarsat STD-C / EGC (A6)

  • STDC-1 Geographic area-address decoder — biggest gap, no OSS decoder does it ★ quick win
    • STDC-1.1 Rectangular (C2=04/34), circular (C2=24/44/14), NAVAREA/METAREA number (C2=31), coastal/NAVTEX (C2=13/73)
    • STDC-1.2 Emit structured area geometry fields + map plotting of areas/coordinates — ✅ geometry fields (signed °/NM) in details. (dashboard map plotting = output follow-up)
  • STDC-2 C-channel descriptor field depth — 0x7D bulletin-board full fields, 0x6C signalling-channel, 0x83 logical-channel-assignment, 0x92 login-ack (LES list), 0xAB les-list, 0xA3/0xA8 short-text, 0x08 ack-request routing
  • STDC-3 Channel-frequency decode (uplink/downlink MHz from channel-number word; formula already in STDC.md)
  • STDC-4 LES/NCS operator-name table + ocean-region long names + service long names
  • STDC-5 Follow 0x83 → demodulate the LES message channel (closes the biggest functional gap vs tekmanoid) ◆ big bet — ⏸ DEFERRED: demodulating the LES message channel is a ◆big-bet needing the channel + a sample — deferred
  • STDC-6 Text — ITA2/Baudot (presentation 6); typed presentation-7 binary capture
  • STDC-7 EGC polish — frame_number→UTC-of-day (×8.64); verify single 0xB0 vs double 0xB1+0xB2 (VERIFY-6); distress-specific position/alerting
  • STDC-8 Demod — ✅ RRC matched filter (shared xng_dsp::rrc_taps), per-frame UW BER, fec_corrected, mid-frame polarity-flip recovery (synthetic AWGN BER test); ❌ optional CMA equalizer, SatDump .frm goldens

IRID — Iridium (A7)

  • IRID-1 Frame typing — AQ acquisition uplink + ISY pattern done; ❌ NXT deferred (not in iridium-toolkit/sniffer — no oracle)
  • IRID-2 Upper-layer IP content — PPP-PAP credential frames + HTTP Basic-Auth headers in plaintext IIP/IIQ/IIR IP sessions (new decode target; ~88% of frames unencrypted)
  • IRID-3 GSM layer-3 — RR messages (Immediate-Assignment/Paging/System-Info) labelling; expose LCW layer; PCAP output (-m lap)
  • IRID-4 Positions — wire Iridium ADS-C / mt-position to SBS + web-map layer (sbs.rs is ModeS-only; ties XM-2.2); render mt-position — ⏸ DEFERRED: bare mt-position has no recoverable ICAO; synthesizing one would plant phantom aircraft in SBS/tar1090 — left web-map-only by id-policy (the web-map layer IS done)
  • IRID-5 Demod — soft-decision / Chase BCH decoding (the next weak-frame lever); UW error-correction pre-classify; explicit SIMD select; GPU/OpenCL burst detection — ✅ soft-decision/Chase-2 BCH + UW access-code pre-classify (gated XNG_IRIDIUM_MAX_EFFORT, default bit-identical; +18.6 pts AWGN; benchmark-gated no regression at 1577 CRC-OK IDA); ❌ explicit SIMD select / GPU-OpenCL burst detection deferred
  • IRID-6 Outputs — KML export; SigMF / burst-IQ capture; IBC-driven PPM clock self-calibration (-m ppm) — ✅ KML/GeoJSON/GPX export of Iridium mt-positions (via geo_entities); SigMF burst-IQ + IBC PPM self-cal deferred
  • IRID-7 Multi-receiver TDOA (-m tdoa) — ties XM-4 (IRA ECEF + IBC iri_time primitives already exist) — ⏸ DEFERRED: multi-RX TDOA (=XM-4) needs multi-receiver infra + synchronized captures — deferred
  • IRID-8 ⚑ Verify Iridium time re-epoch handling (ERA3 2025-02-14; ERA4 2026-01-14 18:08 UTC) in satellite-naming/SGP4 code — live bug risk (= VERIFY-1) — ✅ re-verified 2026-06-18 with ERA4 now active: correct (see VERIFY-1)

AIS — AIS / ITU-R M.1371 (A8)

  • AIS-1 ASM (DAC/FID binary on 6/8) dispatch table ◆ big bet — ✅ dispatch + DAC=200 Inland (FID 10/23/24/40, pyais-verified); ❌ remaining: DAC=1 IMO Circ.289 (no pyais oracle), regional DACs
    • AIS-1.1 DAC=1 IMO SN.1/Circ.289 (FID 31/11 meteo-hydro, 21 weather-from-ship, 16 POB, 22/23 area-notice, 17 VTS, 24/25/26 static/cargo/sensor, 27-30 route/text, 32 tidal) — ✅ FIDs 11/16/17/24/27/28/29/30/31/32 full + 21/22/23/25/26 header-only (spec-derived; no pyais oracle for DAC=1)
    • AIS-1.2 DAC=200 Inland (FID 10/21/22/23/24/40/55) — ✅ 10/23/24/40 (pyais) + 21 ETA / 22 RTA / 55 persons-on-board (spec-derived: UNECE SC.3/176, gpsd + e-Navigation.nl cross-checked)
    • AIS-1.3 Regional DACs (235/250/366 AtoN-monitoring, 316/366 Seaway-meteo, 367 US-environmental, 265 STM-route); validate vs pyais oracle — ✅ DAC 235/250 AtoN-monitoring full + regional header-only (DAC/FID/identification) for the rest; ❌ full 316/366/367/265 bodies (pyais has no oracle — skip-don't-fake)
  • AIS-2 Multi-fragment AIVDM reassembly across sentences (long type 5/6/8/26); type-24 Part A+B merge by MMSI; type-5 voyage merge; per-MMSI AISTracker state
  • AIS-3 Easy sub-field fills — types 1-3 (ROT/accuracy/timestamp/maneuver/RAIM), type 5 (version/dims/EPFD/ETA/DTE), type 4/11 (accuracy/EPFD/RAIM/UTC), type 18 (accuracy/timestamp/RAIM), type 19 (+dims/EPFD/DTE), type 21 AtoN (accuracy/dims/EPFD/timestamp/off-position/RAIM/virtual/name-ext), type 24B (vendor_id/model/serial/callsign + dims-or-mothership). All verified against pyais vectors (1-3 & 5 hand-decoded; 4/18/19/21/24B vs the pyais test-suite vectors). ✅ SOTDMA/ITDMA radio communication-state now decoded too (slot timeout/offset, sub-message by sync-state) — the last sub-field gap is closed.
  • AIS-4 AIS-SART / MOB / EPIRB-AIS distress tagging — fields::distress_class (MMSI prefix 970/972/974) tags distress in details, surfaced in console + dashboard vessel; nav_status 14 (AIS-SART) and Msg-14 ACTIVE/TEST text already decode. (VERIFY-4 resolved: 970/972/974 now mapped; ties XM-6 cross-mode distress overlay)
  • AIS-5 Outputs — ✅ NMEA-over-UDP (--nmea-udp, src/outputs/nmea_udp.rs; README claim fixed to "TCP server + UDP push") + ✅ NMEA tag-blocks (--nmea-tag-blocks, \s:<station>,c:<ts>*HH\ on both NMEA sinks; xng_mode_ais::nmea::tag_block) + ✅ channel_letter fix (±12.5 kHz A/B bands, '?' for non-AIS freqs instead of silently 'A') — all 2026-06-18; + ✅ per-type/MMSI filter + rate downsample + output dedup (AIS-5h, 2026-06-18: AisFilter/AisGate pre-bus in decode_loop, configured per-session in the station TOML; rate-thins dynamic position types per MMSI + collapses content-duplicate reports; generic enough to reuse for XM-5). + ✅ AIVDO own-ship encoder (AIS-5c, 2026-06-18: own_ship_position(mmsi,lat,lon) → Type-1 AIVDO sentence; fields::encode_position_report inverse of the 1..=3 decode arm, round-trip-verified through the pyais-oracle decoder, not-available kinematics not fabricated). ✅ periodic emit wired (2026-06-18: station own-ship-mmsi + a session receiver-pos → run_station injects an AIVDO fix every 30 s onto the bus, clean stop-flag shutdown). ❌ remaining: GPSd output/fusion; NMEA2000/N2K (canboat-GPL/no oracle — skip); HTTP aggregator push (= AIS-Catcher :5599, ties FEED-2.3); JSON_FULL details; single-session CLI flags for the AIS-5h filter (station TOML done)
  • AIS-6 Demod — low-CPU Pi mode (1.4× headroom is thin); verify CIC5 droop compensation (VERIFY-4) — ⏸ DEFERRED: low-CPU Pi mode is an optimization; CIC droop moot (VERIFY-4: no CIC) — deferred

ADSB — Mode S / ADS-B 1090 MHz (A9)

  • ADSB-1 Modern accuracy/integrity + intent layer — entirely absent today ★ quick win (bits already in-message) — partially done on feat/per-decoder-airframes-feeding — ✅ 1.1–1.5 done
    • ADSB-1.1 TC31 operational-status decode → adsb_status (version, NIC-supp-A, NACp, SIL, SIL-supp, GVA, NICbaro), surfaced in console/dashboard/asf-2.0. (NACv is a TC19 field — see ADSB-1.5)
    • ADSB-1.2 ADS-B version read directly from TC31. (heuristic version inference for v0 / non-TC31-emitting aircraft deferred)
    • ADSB-1.3 TC29 target-state (MCP/FCU selected alt, QNH, selected heading, AP/VNAV/APPROACH/LNAV flags)
    • ADSB-1.4 TC28 aircraft-status — emergency/priority state (mapped to label, flagged on the map) + ACAS-RA subtype flag. (full RA decode = ADSB-3.1)
    • ADSB-1.5 Accuracy fields — NACp/SIL/NIC-supp-A/GVA (TC31) + NUCp(v0), version-aware NIC (TC+supplement), NACv (TC19), SDA, HRD; folded into adsb_status (pyModeS uncertainty-table verified)
  • ADSB-2 Position/velocity completion — TC5-8 surface movement+track; TC9-18 Q=0 Gillham altitude (routed through the dump1090-verified ladder — fixed a latent −100 ft bug); TC20-22 geometric altitude; VR source + geom-minus-baro; NACv (dump1090/pyModeS verified)
  • ADSB-3 Comm-B BDS register expansion — ✅ 3.1–3.4 done
    • ADSB-3.1 BDS 3,0 ACAS/TCAS RA (high safety/intel value)
    • ADSB-3.2 BDS 1,0 data-link-capability, 1,7/1,8/1,9 GICB capability, 2,1 registration markings
    • ADSB-3.3 BDS 4,4 (wind/temp/pressure/turbulence/humidity), 4,5 (hazard), 5,3 (air-referenced state)
    • ADSB-3.4 rs1090-style density/penalty BDS scoring (vs binary validate); extend inference beyond 4 registers
  • ADSB-4 DF coverage — DF19 military ES; DF24-27 Comm-D ELM; surface FS/DR/UM (alert/SPI/ground) from DF4/5/20/21
  • ADSB-5 DF18 CF-subtype classification (CF=0 non-transponder, 2/3/5 TIS-B fine/coarse/mgmt, 6 ADS-R) → source tag (VERIFY-7)
  • ADSB-6 Mode A/C decode — decode kernel done (octal squawk / SPI / Gillham ladder, dump1090-oracle-verified); RF framing-pulse demod still deferred
  • ADSB-7 Demod/trust — ✅ graduated position trust (PosTrust grade GlobalUnambiguous / LocalContained / LocalReceiver; NIC/NUCp containment + dump1090 half-CPR-zone cap + speed-gate jump reject; surfaced in adsb_status.position_trust); ❌ phase-classified per-phase bit templates (the ~3-frame demod gap to readsb)
  • ADSB-8 Outputs — readsb-schema aircraft.json (version/nic/nac/sil/gva/emergency/nav_*/acas_ra/wind/oat); ✅ true RX-clock Beast timestamps → MLAT-feedable (2026-06-18: PpmDemod tracks a drained-samples base → per-frame absolute sample offset → monotonic 12 MHz tick on SignalQuality.rx_ticks_12mhz → Beast counter; wall-clock fallback only when absent; benchmark unchanged at 323); ✅ readsb-schema aircraft.json (2026-06-18, = ECO-4); ✅ tisb/adsr provenance type (2026-06-18, ADSB-8a: DF18 CF-class → tisb_icao/tisb_other/adsr_icao/adsb_icao_nt/adsb_other, DF17/19 → adsb_icao, bare Mode S → mode_s, no-downgrade precedence; "mlat" deliberately NOT emitted — passive single RX); ✅ docs/notes/ADSB.md Beast-MLAT bullet corrected to as-built RX-clock state (2026-06-18, ADSB-8b). ADSB-8 outputs now complete.

ECO — Ecosystem / outputs / tooling / dashboard (B)

  • ECO-1 Persistence, history & replay — on-disk message DB + retention + search UI; per-aircraft trace history; BaseStation flight/aircraft DB; dashboard time-scrubber/replay of decoded history; position-density heatmap — ⏸ DEFERRED: on-disk message DB + retention + replay + heatmap is a large persistence subsystem — deferred
  • ECO-2 Coverage / range analytics — measured/theoretical range outline (upintheair/HeyWhatsThat/polar), range rings, distance/RSSI columns + per-entity reception — ⏸ DEFERRED: coverage/range analytics is dashboard UI + range modeling (needs visual QA) — deferred
  • ECO-3 Alerting & notifications — watchlists (tail/flight/ICAO/MMSI/label/text); special-category highlighting (dbFlags mil/interesting/PIA/LADD + emergency squawk 7500/7600/7700); push (Discord/Telegram/webhook/sound); MQTT Home Assistant autodiscovery — ⏸ DEFERRED: watchlist matching is implementable, but the push channels (Discord/Telegram/webhook/MQTT-HA) need external services to verify end-to-end — deferred as a unit (emergency highlighting already done via XM-6)
  • ECO-4 REST / data API — ✅ /data/aircraft.json + /data/receiver.json served off the dashboard port (2026-06-18: readsb field schema — hex/flight/r/t/alt_baro/gs/track/squawk/nac_p/sil/version/lat/lon/seen/seen_pos/messages — merged across ADS-B/UAT/HFDL by ICAO; receiver.json carries version + receiver-pos; makes xng a drop-in for tar1090/graphs1090/VRS); ✅ incremental ?since=<unix> (2026-06-18: aircraft.json?since= returns only aircraft heard at/after that time, tar1090 poll pattern; query_since parser + filter tested); ❌ remaining: WebSocket/SSE/delta stream
  • ECO-5 Export formats — GeoJSON / KML / GPX; ?screenshot kiosk + shareable deep-link state — ✅ GeoJSON + GPX + KML 2026-06-18: /data/export.geojson (RFC 7946 FeatureCollection: Point per fix + LineString per trail) + /data/export.gpx (GPX 1.1 wpt/trk) + /data/export.kml (OGC KML 2.2 Placemark/Point + LineString), shared geo_entities() iterator over aircraft/vessels/beacons, coordinate-order asymmetry (GeoJSON/KML [lon,lat] vs GPX lat/lon attrs) tested. ❌ remaining: ?screenshot kiosk + deep-link hash state (dashboard.html UI)
  • ECO-6 Map/dashboard UX — basemap switcher/overlays/offline MBTiles/weather; persistent trails; map+table filters (alt/speed/type/callsign/source); measure/ruler, gridlines, on-map labels, box-select, tableInView, column sort, dim/contrast/icon-scale controls — ⏸ DEFERRED: map/dashboard UX is pure front-end needing visual QA — deferred
  • ECO-7 Statistics & graphs — built-in graphs page (msgs/min, per-freq, level histogram, CRC/error %, CPU/temp/uptime); expand Prometheus (per-mode/type counts, entity gauges, decode latency, FEC-corrected, reassembly, dropped/lagged-bus); first-party Grafana dashboard JSON — ✅ FEC-corrected counter + per-(mode,freq) frame/crc/level + per-label ACARS counters done; built-in graphs page + Grafana JSON = UI deferred
  • ECO-8 Health / watchdog / operability — health state machine + no-data alarm + auto-restart; continuous autogain during listen (VERIFY-7); web config/management UI — ⏸ DEFERRED: health/watchdog + web config UI is a large operability subsystem; continuous autogain captured under VERIFY-7 — deferred
  • ECO-9 Multi-receiver aggregation & integrations — turn xng ingest into a fan-in dashboard/dedup hub (ties XM-5); SignalK delta; NMEA2000/N2K; NMEA tag-blocks + community-map push; GPSd input — ⏸ DEFERRED: multi-RX aggregation hub = XM-4/XM-5 over the asf-2.0 fan-in + external integrations (SignalK / NMEA2000 canboat-GPL / GPSd) — deferred
  • ECO-10 Enrichment data — bundled/auto-updating aircraft DB + dbFlags; route/airline/operator/airport enrichment on dashboard; persistent vessel registry across restarts — ⏸ DEFERRED: enrichment needs a bundled/auto-updating aircraft+vessel DB (data dependency); partial dbinfo (country/db lookup) exists — deferred
  • ECO-11 Smaller dashboard/quality — per-mode entity expiry (not flat 300 s); message-stream scrollback backed by ring file; dark/light/scale controls; verify non-Iridium trail antimeridian wrapping (VERIFY-7) — ⏸ DEFERRED: per-mode expiry is design-ambiguous for multi-source entities; scrollback-ring + theme controls are UI — deferred (flat 300 s works; antimeridian wrap done in VERIFY-7)
  • ECO-12 Aggregator-network targets — feed/dedup against airplanes.live / adsb.lol (ODbL) / adsb.fi / OpenSky; study BelugaProject (air+sea fusion prior art) — ⏸ DEFERRED: aggregator-network feeds need each network's endpoint/protocol + accounts (external) — deferred

NEW — New modes / capabilities (D)

Priority tiers from the consolidated viability matrix. NEW-SKIP items are considered and declined — kept so they aren't re-proposed.

  • NEW-P0-1 UAT 978 MHz — crate xng-mode-uat: ADS-B downlink (state vector) + FIS-B uplink (APDU framing + DLAC text products), oracle-verified. (IQ demod + bin/Mode wiring = follow-up)
    • NEW-P0-1.1 UAT ADS-B short/long messages (US GA positions) — ✅ UAT ADS-B short/long state vectors decode in xng-mode-uat (oracle-verified)
    • NEW-P0-1.2 FIS-B weather products (NEXRAD/METAR/TAF/PIREP/AIRMET-SIGMET/Winds-Temps/NOTAM/TFR/SUA) — audited 2026-06-18: the DLAC text products (20-27, 411-413) + APDU frame headers are done and dump978-verified; the non-DLAC text products (0-13) and graphical products (51-102) are blocked on a missing oracle — dump978's uat2text prints 0-13 raw (no frame decode) and no public reference decodes the graphical rasters. Skip-don't-fake until a fixture exists.
    • NEW-P0-1.3 TIS-B + ADS-R (uat2esnt → DF18 CF=6 integration path) — ✅ 2026-06-18: synth::EsSource (Adsb→DF17 CA=5, TisB→DF18 CF=2, AdsR→DF18 CF=6); AircraftFix.source from the UAT address_qualifier (tisb_*/adsr_other); Beast routes accordingly so UAT 978 TIS-B/ADS-R keep their class on 1090. Round-trip-verified through the oracle-validated df18_cf_class (+ CPR still decodes); Beast unwrap asserts DF18 CF=6 for ADS-R. DF17 native path unchanged.
  • NEW-P0-2 COSPAS-SARSAT 406 — crate xng-mode-sarsat: FGB (T.001) message + BCH decode, amsa-code-verified. (SGB T.018 + IQ demod = follow-up — audited 2026-06-18: SGB (2nd-gen, 250-bit OQPSK+DSSS, distinct layout/FEC) is blocked on a missing oracle: no vendored SGB vectors, no amsa-code/sgb-decoder compliance kit, no published T.018 worked example. Skip-don't-fake until a reference message exists.)
  • NEW-P0-3 DSC — crate xng-mode-dsc: ITU-R M.493 symbol + message decode. (IQ demod = follow-up)
  • NEW-P1-1 Radiosondes — crate xng-mode-sonde: RS41 RS-FEC + frame (STATUS/GPS/PTU) decode, real-off-air verified (119/119 vs rs1729 rs41mod). (GFSK demod + RS92/DFM/M10/… = follow-up — audited 2026-06-18: additional sonde types blocked on a missing oracle — no vendored capture or published worked-example frames for RS92/DFM/M10 in-crate; skip-don't-fake until one exists)
  • NEW-P1-2 AIS-SART / MOB / EPIRB-AIS — done under AIS-4 (cross-reference)
  • NEW-P1-3 EOT / HOT / DPU rail telemetry — crate xng-mode-eot: Manchester-FSK + AAR S-9152 frame (unit address, brake-pipe pressure, motion, marker light + battery, turbine/valve, BCH); direction (eot/hot) by RX freq (457.9375 / 452.9375 MHz). Runtime-wired to --mode eot. (reverse-engineered field semantics per cited open decoders; demod synthetic AWGN-BER — no public IQ; train-tail map plotting = follow-up)
  • NEW-P1-4 NAVTEX — crate xng-mode-navtex: CCIR 476 + FEC-B + ZCZC message decode. (IQ demod = follow-up)
  • NEW-P2-1 ADS-L — crate xng-mode-adsl (the "ADS-K" item): EASA SRD860 message decode. (FANET/OGNTP + IQ demod = follow-up)
  • NEW-P2-2 APRS / AX.25 (incl. HAB balloons) — crate xng-mode-aprs: AFSK1200 (Bell 202) over FM + AX.25 v2.2 UI (callsign/SSID/digipeaters, X.25 FCS) + APRS payload — uncompressed + Base-91 compressed position (incl. course/speed/alt), Mic-E, weather, message, status, object, item, bulletin/announcement, general-query, PHG/DFS/RNG data-extensions, Maidenhead grid, telemetry (T# data values + PARM/UNIT/EQNS/BITS definition messages, APRS 1.0.1 ch.13, spec-worked-example verified). Runtime-wired to --mode aprs (144.39 NA / 144.8 EU / 432.5 UK). (igate feed = follow-up; demod synthetic AWGN-BER — real-RF unconfirmed: 144.39 quiet at the KSMF soak site, decode pipeline verified by spec)
  • NEW-P2-3 POCSAG / FLEX / FLEX-NEXT paging — ✅ POCSAG (crate xng-mode-pocsag: 2-FSK 512/1200/2400 multi-baud auto-detect + CCIR Code No.1 / ITU-R M.584-2, BCH(31,21,2), numeric/alpha/tone) + ✅ FLEX (crate xng-mode-flex: 1600 bps 2-FSK + 4-level 3200/6400, auto rate-detect from the Sync-1 A-code, FLEX sync/FIW/BIW, BCH(31,21), alpha/numeric/tone, off-air garbage rejection + alpha header strip), both runtime-wired (--mode pocsag / --mode flex, FLEX opened baud=0=auto). FLEX real off-air validated (929 MHz 6400 4-level US paging: recovers clean alpha pages); POCSAG demod synthetic AWGN-BER. ❌ FLEX-NEXT still open (per-session baud knob moot — rate auto-detected)
  • NEW-P2-4 ATCS — crate xng-mode-atcs: HDLC/LAPB framer + Spec-200 address/header decode. (Genisys/ARES payload + IQ demod = follow-up — audited 2026-06-18: the codeline payload protocols (Genisys/ARES/SCS-128) are proprietary AAR systems with no public spec or OSS reference; sigidwiki shows message-type tags but no payload format. Blocked — declined per skip-don't-fake; raw user_data preserved.)
  • NEW-P2-5 VDES / long-range AIS extensions — ✅ ASM via crate xng-mode-vdes: GMSK 9600 + HDLC/CRC-16 transport on the ASM 1/2 channels (former AIS 27/28), AIS Msg 6/8 ASM headers (source/dest MMSI + DAC/FID), and 7 spec-cited DAC/FID application payloads — DAC=1 FID 11 (met/hydro IMO236), 16 (POB), 17 (VTS synthetic target), 18 (clearance-time), 31 (deep IMO289 weather block); DAC=200 FID 10 (Inland static&voyage), 55 (persons-on-board) — + raw data_hex fallback. pyais + gpsd spec verified. Runtime-wired to --mode vdes (161.950/162.000 MHz). (demod synthetic AWGN-BER; sparse public VDES spec); ❌ variable-length/repeated-block ASM bodies w/o a hand-verifiable vector (FID 14/20/22/23/25, regional DACs), VDE-SAT/terrestrial data channels, full AIS-2.0 still open
  • NEW-P3 Parking lot (low priority / dependent) — FLARM (after open OGN), LoRa-APRS/Horus HAB (radiosonde rider), DMR-LRRP / P25-Unit-GPS / TETRA-SDS-LIP position PDUs (metadata only), TPMS (rtl_433 subset), Orbcomm STX (breadth flex), GTFS-realtime ingest (non-RF complement), WSPR/FT8 (HF-prop health niche) — ⏸ DEFERRED: parking lot — explicitly low-priority / dependent future modes
  • NEW-SKIP Considered & declined (documented in D) — VDL Mode 4 (regional/near-dead), DSRC/C-V2X (wrong band, sunsetting), toll/MDT, AEI/Eurobalise (passive/near-field), PTC/GSM-R/FRMCS (encrypted), weather imagery NOAA/Meteor/GOES (SatDump incumbent), cubesat/SatNOGS, Globalstar/Starlink/rocket (proprietary/encrypted), GNSS/SBAS (no entities), Drone Remote ID (off-axis; only via ADS-L Issue-2), WEFAX/Pactor/Winlink, wM-Bus, ERMES — ⏸ DEFERRED: considered & declined — terminal by definition (rationale in COMPARISON_RESEARCH §D)
  • NEW-V Verify before committing — (a) which airports/airlines still run POCSAG/FLEX ground-ops paging; (b) AeroMACS/Gatelink demand; (c) whether xng L-band front-ends pass 406 MHz for SARSAT reuse; (d) whether generic-ISM (TPMS) scope is desired (= VERIFY-11) — ⏸ DEFERRED: external market/hardware research questions (=VERIFY-11) — not code

VERIFY — Research / correctness checks (Appendix + inline ❓)

Mostly "read the code / confirm against a source" tasks; resolve these to confirm or retire the matching items above. ⚑ = possible live bug.

  • VERIFY-1 ⚑ Iridium time re-epoch correctness in satellite-naming/SGP4/TLE code (= IRID-8) — ✅ resolved 2026-06-18 (live): iri_time_unix_at() (ira.rs) selects the newest era-epoch whose base ≤ reference time; all four era bases independently checked vs MetOcean bulletin instants (ERA2 2014-05-11, ERA3 2025-02-14, ERA4 2026-01-14T18:08Z — active today), boundary inclusive of the newer era (no off-by-one), ERA2 leap-seconds applied only in-range. Post-2026-01-14 frames decode to 2026, not 2014. 4 unit tests pin it; demod bit/field-exact oracles guard no regression.
  • VERIFY-2 ACARS — raw MIN / 4th-char downlink-rule edge cases; any media-advisory v1+ exists? — ✅ resolved 2026-06-18: raw 4-char MIN surfaces on every carrier (AcarsCore.msg_num); split MIN (3-char + 4th-seq + reassembly index) on the byte path (min.rs::split_downlink, consumed by HFDL/Aero); 4th-char rule correct (A-Z→seq, digit/punct→None, used in reassembly skip); media-advisory SA is v0-only by libacars-faithful design and MIAM (ARINC 841) already covers v1/v2 CORE PDUs + file transfer. (low-pri follow-up: the native VHF/POA path doesn't yet forward the split MIN to the serialized AcarsCore — needs a shared field on AcarsCore; raw MIN is already there. Per-label Prometheus counters → VERIFY-9.)
  • VERIFY-3 VDL2 — ✅ resolved: SABME already folded into the U-frame arm (not U?); ES-IS option-TLVs present; plain/unprotected CPDLC added (VDL2-1.5); the v2.5.1 249-octet-multiple block-length bug does NOT affect xng (RS row count = div_ceil(tl_bits/1992), exact at the 1992-bit boundary — regression-guarded). (StatsD good_loud/pp_acars are output-side, tracked under VDL2-8)
  • VERIFY-4 AIS — CIC5 droop compensation in xng-dsp; mid-frame polarity-flip recovery — ✅ resolved 2026-06-18: (a) moot — there is no CIC anywhere in the DSP (repo-wide grep: zero CIC/comb/integrator hits); the AIS front end uses FIR flat-passband decimation, so droop compensation is unnecessary. (b) polarity/NRZI-ambiguity recovery is already present — NRZI decodes on level-change (inherently polarity-invariant) + an explicit π-flip trellis seed + decision-directed phase tracking. (MMSI 970/972/974 mapping: RESOLVED — AIS-4)
  • VERIFY-5 HFDL — resolved: as-built is a 7-tap symbol-spaced LMS equalizer; dumphfdl's 15 is T/2 (half-symbol)-spaced — both correct, different spacing convention (code comments fixed)
  • VERIFY-6 STD-C — resolved: 0xB0 vs 0xB1+0xB2 distinction surfaced in details; mid-frame polarity-flip recovery added (STDC-8)
  • VERIFY-7 Ecosystem — ✅ resolved 2026-06-18: (a) receiver-position pin — was missing, now implemented (session_descriptor emits receiver_pos; dashboard renders a 📡 station marker; verified live at 38.5125,-121.4925); (b) non-Iridium antimeridian trail wrapping — was missing (only Iridium sat trails unwrapped), now implemented (upsert applies unwrapTrail to aircraft/vessel/beacon trails); (c) DF18 CF-subtype (= ADSB-5) confirmed done (df18_cf_class, DO-260B). ❌ remaining: continuous autogain during listen (only survey --tune-gain exists today) — a larger feature, deferred.
  • VERIFY-8 Aero-C — resolved: AEROTypeP/R enumerator hex verified against the JAERO source (AERO-6)
  • VERIFY-9 ACARS — acarsdec mqttout.c in current f00b4r0 4.x tree (post-SoapySDR refactor)?; xng per-label ACARS counters? — ✅ per-label counters DONE 2026-06-18: LiveState.acars_labels (freq,label→count) map; decode_loop tallies each published ACARS message; metrics.rs emits xng_acars_messages_total{mode,freq,label} (escaped label values). En route, fixed a latent station-mode bug: run_station served /metrics off a fresh LiveState the decode loops never updated (frame/CRC counters always 0) — now one shared LiveState, per-channel stats freq-keyed so multi-session stations don't clobber rows. ✅ MQTT output sink confirmed present (src/outputs/mqtt.rs, = ACARS-5.2). (mqttout f00b4r0-4.x source question = external research, immaterial — xng's MQTT output is independent.)
  • VERIFY-10 ADS-B — does xng-mode-adsb already emit Mode A/C? (= ADSB-6) — ✅ resolved 2026-06-18: it does not emit Mode A/C — the decode kernel exists (mode_ac.rs: octal squawk / SPI / Gillham ladder, dump1090-verified) but is unwired (no framing-pulse demod), exactly as ADSB-6/PROVENANCE state. The README does not overclaim (its "altitude replies" = Mode S DF4/20, not Mode A/C). No action — remaining work is the RF framing-pulse demod (ADSB-6, deferred).
  • VERIFY-11 New-mode commitments — POCSAG/FLEX airfield usage; AeroMACS/Gatelink demand; 406 MHz front-end reuse; generic-ISM scope desire (= NEW-V) — ⏸ DEFERRED: external market/hardware research questions — not code
  • VERIFY-12 Beast MLAT counter usability — ✅ resolved 2026-06-18: confirmed the blocker was the wall-clock-derived (jittery, non-monotonic) Beast timestamps, not GPS absence — then fixed it: PpmDemod now tracks a drained-samples base, so each frame carries its absolute stream sample offset, converted to a monotonic consistent-rate 12 MHz tick and stamped on the Beast counter (rx_ticks_12mhz, wall-clock fallback if absent). The feed is now well-formed for an MLAT client to fit the RTL clock drift. Passive readout — frame counts unchanged (benchmark adsb_modes1 = 323). New Beast unit test. (= ADSB-8 Beast-timestamp item.)
  • VERIFY-13 Feed — confirm VDL2 ingest format preference (dumpvdl2 decoded:json vs vdlm2dec); whether Airframes exposes public IMSL/IRDM/STD-C ports + the settled Iridium feeding mechanism (= FEED-4) — ✅ VDL2 format confirmed = dumpvdl2 decoded:json (FEED-2.1); IMSL/IRDM/STD-C ports + Iridium mechanism remain external (FEED-4.2)