feat(egress): generic TS output sink (UDP/RTP/FEC/ST 2022-7) + PCR-aware pacing

[t0ms]

Context

#1828 added SRT egress (MoQ broadcast -> moq_mux::ts::Export -> SRT messages), but the
wiring is SRT-specific and its review explicitly deferred pacing: send timestamps use
Instant::now() rather than media PTS, so a tune-in keyframe burst is released as a burst,
and "a proper fix needs a PTS-exposing API change in moq_mux::Export". #1799 also lists
RTP/UDP, FEC, ST 2022-7, and multicast as broadcast egress needs. This issue proposes a
small shared egress abstraction so transports are pluggable, plus the PCR-aware pacing the
SRT work punted on. Part of #1799; builds on #1828.

Proposal

1. PCR-aware pacing + a PTS-exposing Export API (do first)

• Add an additive API on moq_mux::container::ts::Export to expose per-output timing
  (PTS/PCR) to the caller, the same surface feat(moq-srt): serve broadcasts back out over SRT (m=request egress) #1828 needs and that fix(moq-mux): author DTS for B-frame MPEG-TS export (decode timeline not emitted) #1836's faithful DTS
  option also wants. Solve once.
• A pacing component releases TS toward the transport on the media clock (PCR) instead of
  wall-clock, fixing the feat(moq-srt): serve broadcasts back out over SRT (m=request egress) #1828 tune-in burst for every transport, not just SRT.

2. A generic egress sink

• A small trait, output sink, that consumes paced TS packets/bytes (+ timing) and writes
  them to a transport. Refactor feat(moq-srt): serve broadcasts back out over SRT (m=request egress) #1828's SRT path to be one implementation.
• Backends, in order:
  1. raw UDP (unicast + multicast)
  2. RTP/UDP (RFC 2250 MP2T payload), unicast + multicast
  3. RTP + FEC (SMPTE 2022-1 / Pro-MPEG COP3 column+row)
  4. ST 2022-7 seamless protection switching (dual-path hitless merge)
  • (SRT stays as the existing sink.)

Where it lives

A dedicated egress crate, or extend the egress shape moq-srt introduced; the relay core
stays media-agnostic. Coordinate with moq-srt (which is dev).

Branch

dev to sit alongside moq-srt egress. The moq_mux::Export timing-API addition is
additive (non-breaking).

Notes

A private reference implementation of RTP, FEC (2022-1), PCR pacing, and a decoder-safe
start gate exists; offered as design background and a starting point, not a wholesale
merge, real upstreaming carries standards-conformance (RFC 2250, SMPTE 2022-1/-7) and
interop tests (TSDuck, a hardware IRD).
(Written by Claude)

[coderabbitai]

🔗 Related PRs

#1808 - moq-gst: fix plugin license + broadcast-aligned timestamps [merged]
#1809 - fix(moq-gst): deterministic moqsrc pad names so CMAF playback works [merged]
#1812 - fix(moq-mux): carry all distinct SPS/PPS/VPS through transmux, not just the last seen [merged]
#1815 - feat(moq-mux): generic verbatim MPEG-TS carriage (mpegts catalog section) [merged]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!

[kixelated]

I'm not sure about the second part. It's feasible but right now we're just targeting WebRTC, not RTP in general.

[t0ms]

Update: PCR-paced egress + a PTS-exposing API have landed (on dev) — re-centering this issue

A chunk of what this issue asked for is now implemented on dev:

• feat(moq-mux,moq-srt): PTS-exposing TS Export API + PCR-paced SRT egress #1845 — PTS-exposing TS Export API + PCR-paced SRT egress. This is the phase-1 pacing primitive this issue called for, now real for the SRT sink.
• feat(moq-srt): serve broadcasts back out over SRT (m=request egress) #1828 (SRT egress / m=request), feat(moq-rtmp): RTMP play mode (egress) #1829 (RTMP egress), feat(moq-rtc): H.265 + AV1 egress, unify NALU reshaping via moq-mux #1827 (RTC H.265+AV1 egress) — three egress gateways now exist, so the "serve a broadcast back out" pattern is established.

That changes the shape of the remaining ask. Rather than a standalone TS sink, the gap is now:

1. A generic egress sink abstraction that the existing SRT/RTMP/RTC outputs and a new UDP/RTP output all implement, so PCR-aware pacing (from feat(moq-mux,moq-srt): PTS-exposing TS Export API + PCR-paced SRT egress #1845) is shared rather than per-sink.
2. UDP / RTP output (raw TS over UDP and RTP/MP2T) — the classic contribution/distribution transport, still missing.
3. FEC — SMPTE 2022-1 / RFC 2733-style row/column FEC for RTP, the SRT-recovery equivalent for the UDP path.
4. ST 2022-7 — seamless dual-path (hitless) protection, the satellite-replacement-grade reliability tier.

Suggested sequencing: (1) sink trait → (2) UDP/RTP → (3) FEC → (4) ST 2022-7. (1)–(2) are the near-term unlock; (3)–(4) are the "comparable to Zixi/RIST" tier.

This pairs directly with the opaque TS lane (#1861): an opaque whole-mux source feeding a PCR-paced UDP/RTP+FEC sink is the end-to-end native contribution path. Happy to take the UDP/RTP + FEC backends once the sink trait shape is agreed.

(Written by Claude)

[kixelated]

I don't want to add an import/export module for every possible transport. If they're common enough then sure, but IDK it's not hard to make a tiny layer on top of moq-mux. I think I need a concrete customer ask first.