MOQ Wiki

CMSF - CMAF Compliant MOQT Streaming Format

8 min read

2026-06-07: Tobbe’s CUE catalog validator finds CMSF draft-01 bugs → PR #23 OPEN June 4 “Fix version examples and minor typos in draft-01” (+5/−5). Companion to moq-wg/msf PR #177: Torbjörn Einarsson’s MSF/CMSF catalog validator (announced on #moq June 5, validates strictly against draft-01 definitions) surfaced wrong version example values + typos in CMSF -01. Pure editorial fixes for a future -02; the validator additionally covers LOCMAF packaging, exercising the CMSF/LOCMAF init-reference split that came out of the June-3 initData thread. See moq-msf for the validator detail and interim-meetings — CMSF stays on the London Day-2 Will Law slot.

2026-06-04: Will Law submits draft-ietf-moq-cmsf-01 to Datatracker June 3 — first CMSF revision since -00 (Dec 2025), ~187 days since adoption. Final pre-submission clean-up was moq-wg/cmsf PR #22 MERGED 11:32 UTC by wilaw “Fix I-D nits” (+7/−6, 1f, “Add LOC reference and update abstract for clarity”) following yesterday’s PR #21 MSF-01 reference bump. 2nd MoQ WG draft revision to land in 2 days alongside MSF -01 (June 2); structurally pairs the MSF/CMSF anchors for the London Day-2 35-min MSF/CMSF Will Law slot. wilaw also posts companion “MSF updates” + “CMSF updates” announcements to the IETF MoQ mailing list June 3 — first per-spec author-led mailing-list update the wiki has tracked from wilaw. Submitted same week as LOCMAF -00 (June 2 by Tobbe + Hugo Björs) — the LOC/CMSF/LOCMAF/MSFTS individual-draft landscape is now stratified by initData semantics: CMSF carries full CMAF chunks, LOCMAF delta-compresses CMAF chunk metadata, LOC strips fMP4 entirely, MSFTS carries MPEG-TS. Carry-forward: with CMSF -01 + MSF -01 both on Datatracker, the moq-msf Issue #153 initTrack design problem (re-surfaced June 3 in Tobbe’s LOCMAF Slack thread) becomes the lead open editorial item — afrind suggests “init data as a track property” / “only publish the init subgroup in groups where it changed”, wilaw replies “currently we do not allow track init properties to change once publish has begun” (Slack June 3 thread under Tobbe’s LOCMAF announcement, see discussions-2026-06).

2026-06-03: PR #21 MERGED June 2 13:54 UTC by wilaw “Update CMSF document with new references and JSON format” (+69/−20, 1f, closes Issue #20 “Update CMSF to reference the new MSF ID draft-01”). Updates CMSF’s MSF normative reference from draft-ietf-moq-msf-00 → draft-ietf-moq-msf-01 (published same day on Datatracker) and adjusts the JSON structure for CMSF versioning + initialization data. Companion editorial work to MSF -01 submission: with MSF -01 finally published June 2 after 134-day cadence, CMSF main is brought into reference alignment within ~2 hours so any reader of either spec finds a coherent cross-reference. No new CMSF revision submitted — Datatracker still shows draft-ietf-moq-cmsf-00 (Dec 2025, expired status). The reference-bump-on-main pattern matches MSF’s pre-submit work and the SHA on the main branch now reflects the MSF-01-aware reference set, awaiting the next CMSF -01 cycle to land on Datatracker. Carry-forward: with MSF -01 published + CMSF main reference-bumped, London Day-2 35-min MSF/CMSF Will Law slot now has both anchor docs in current state; the next CMSF cycle will need to absorb the May 30 Gwendal-vs-Suhas zapping debate (Issue #122) + the May 25 emsg signaling thread (Issue #16) + the new JSON structure adjustments from PR #21.

2026-05-31: Gwendal Simon vs Suhas Nandakumar editorial debate on Issue #122 “initial text on zapping” May 30. Gwendal May 30 13:23 UTC: “I find this PR over-explaining something that is quite simple in theory. Do we need so many examples for such a basic thing? The issue #110 was raised during the Boulder meeting (i was only the notetaker). The idea was to mimic a mechanism such as HESP, where one track has much more keyframes than other tracks, to enable fast switching/zapping. In this regard, the parameter in the manifest helps…” — challenges the example-heavy approach as over-engineering. Suhas May 30 21:43 UTC: “we can simplify text if needed, but on how do we use aspect, i wonder how much of that should be in this spec? may be an appendix section?” — proposes a structural split (normative spec text vs informative appendix). Significance: first cmsf editor-vs-contributor design pushback the wiki has tracked — normally cmsf editorial work has been suhasHere monologue + late wilaw review. Gwendal’s “this is simple in theory” framing is structurally analogous to his May 27 chair-rebuttal on SWITCH — challenges a published artifact’s framing rather than waiting for show-of-hands at interim. Carry-forward: zapping/fast-switching is one of the two open CMSF schema threads alongside emsg signaling (Issue #16 reactivated May 25 wilaw asks Gwendal 3 follow-up questions on catalog EMSG signaling); both are slot candidates for the London Day-2 35-min MSF/CMSF Will Law slot.

2026-05-26: Will Law CMSF spec edits May 25: PR #19 MERGED May 25 09:19 UTC “Clarify media content and group packaging requirements” (+2/−3, fixes Issue #12 by yekuiwang) — drops the §3.3 “MUST contain media content encoded in decode order” line (redundant — ISOBMFF already guarantees decode-order) and changes §3.4 from “one or more contiguous Groups of Pictures (GOPs)” to “one or more contiguous independently coded sequences of media samples”. Rationale: GOP is a loose term and moq-transport already replaced GOP with “independently coded sequence” per moq-wg/moq-transport#951; the type of media could be other than video (audio sequences also need the same treatment). Issue #16 reactivated May 25 09:41 UTC by wilaw — original Feb 5 Denver MSF meeting agenda issue from Gwendal Simon covering emsg boxes in CMAF in CMSF + init segment per track for key rotation + Extension parameters for DRM key rotation; wilaw now asks Gwendal 3 follow-up questions about emsg signaling: (1) catalog signaling of EMSG presence so players know to parse ISO boxes, (2) per-track scheme_id_uri indicating payload (example: "emsg": "urn:scte:scte35:2013:xml"), (3) multiple emsg tags per track. Carry-forward: emsg signaling is the third active CMSF schema thread alongside #12 (now closed) and the May 22-23 moq CMSF pipeline implementation (PR #1444); Gwendal’s emsg response likely lands as a follow-on PR before London Day-2 Will Law slot.

2026-05-23: CMSF unified pipeline lands in moq via PR #1444 MERGED May 22 21:12 UTC (+1278/−14, kixelated) — kixelated forks AWS’s #1429 (which had superseded #1408), strips out-of-scope C API + caller-driven group boundaries, ships the MSF-catalog core. Third AWS-vs-kixelated design-cycle resolution in 9 days (#1413 close → #1408→#1429 50% shrink → #1429→#1444 33% shrink); total AWS net code in merged result is ~25-30% of original #1408 scope. Architectural principle: “CMSF is CMAF with a different catalog format” — hang::Catalog serves as intermediate representation, single import pipeline serializes to MSF or Hang catalog formats. First fully-working CMSF muxer/demuxer in any tracked open-source MoQ implementation.

draft-ietf-moq-cmsf-01 | 19 pages | Submitted 2026-06-03 | Datatracker

Authors

Abstract

CMSF extends moq-msf by introducing CMAF-packaged media support. It describes the syntax and semantics for adding CMAF-packaged media to MSF, enabling delivery of CMAF and LOC-compliant content over moq-transport while maintaining all existing MSF capabilities.

Relationship to MSF and LOC

  • moq-msf is the base streaming format
  • moq-loc is the low-overhead container for interactive use
  • CMSF adds CMAF packaging for traditional OTT/broadcast workflows
  • All three work together: CMSF inherits MSF’s catalog, timeline, ABR, and LOC support

Design Tension

The MOQ ecosystem has two container philosophies:

  1. LOC - Minimal overhead, WebCodecs-native, optimized for real-time
  2. CMAF - Industry standard for OTT, broad tooling support, higher overhead

CMSF bridges this by allowing CMAF content to flow through the same MSF infrastructure alongside LOC content. luke-curley’s compressed-mp4 proposal is another approach to reduce CMAF overhead.

See media-packaging for a full comparison.

History

CMSF originated as draft-law-moq-carp-00 (“CARP - a CMAF compliant implementation of WARP”) submitted by will-law in November 2025. The draft was adopted by the MOQ working group and published as draft-ietf-moq-cmsf-00 in December 2025.

ContentProtection Signaling

PR #18 (merged Apr 14 by will-law) added ContentProtection signaling to the CMSF spec. The design, proposed by Torbjörn Einarsson (Eyevinn), is roughly based on DASH/DASH-IF content protection signaling:

  • Defines contentProtections as a top-level array of protection descriptors with refIDs in the catalog
  • Each encrypted track references applicable protections via a contentProtectionRefIDs array
  • Supports the same attributes as DASH/DASH-IF: Widevine, PlayReady, FairPlay, and ECCP (clear key)
  • Addresses issue #8

Implementations:

Key rotation is not yet defined — future work can build on this for live catalogs with new refIDs.

Open Issues & PRs

  • Issue #17 (Apr 14, DenizUgur) - Explicit signalling of DRM/C2PA key-rotation or init segment updates. Proposals discussed at MSF Feb 5 meeting: (a) separate init-segment track with sync concerns, or (b) inline update signaling. Links to the broader moq-msf initTrack debate.
  • PR #19 (Apr 14, wilaw) - Clarify media content and group packaging requirements (fixes #12)

Related

External Links

Graph View

DraftsConceptsPeopleImplDiscInteropOtherTags
DraftsConceptsPeopleImplDiscInteropOtherTags

Backlinks