Wowza Streaming Engine MoQ Module
Proof-of-concept Java module that maps Wowza Streaming Engine's existing CMAF fMP4 output onto MoQT using the CMSF CMAF-over-MoQ mapping, with no core code changes. Explicitly not production-ready, but a useful demonstration that an existing CMAF packager can reach MoQ through the standardised container mapping.
Category
Publishers & Encoders
License
Commercial
Status
Experimental
Languages
Java
Media Formats
A proof of concept from Barry Owen, Wowza’s Chief Solution Architect, published April 2026. It takes the CMAF fMP4 that Wowza Streaming Engine already produces and delivers it over MoQT using the CMSF CMAF-over-MoQ mapping, implemented as a Streaming Engine extension module with no modifications to core code.
Wowza is explicit that this is not a production release — their words are that it would need significantly more testing and hardening first. We list it because of what it demonstrates rather than what you can deploy: that an existing CMAF packaging pipeline can reach MoQ through the standardised container mapping, without rebuilding the media path.
Why the CMSF route is the interesting part
Most vendor MoQ work so far has gone through moq-lite, which trades generality for a smaller surface. This one goes through CMSF, the working group’s adopted CMAF-compliant implementation of the MOQT Streaming Format — the path the IETF track intends for anyone already producing CMAF.
For the large population of broadcasters whose entire plant already emits CMAF, that route matters more than any latency number. It is the difference between “adopt MoQ” meaning a new pipeline and meaning a new output module. We cover the container layer this sits in as part of the MoQ draft ecosystem.
On the conceptual inversion
Owen’s write-up is worth reading for one passage in particular, because it names the thing that trips up every team arriving from HTTP streaming: HLS and DASH are pull models where clients request playlists then segments, while MoQT is push — the publisher pushes, relays fan out, subscribers receive, and nothing polls an origin for a segment URL. He notes both trigger paths, subscriber-initiated SUBSCRIBE and the newer publisher-initiated PUBLISH, and observes that this single inversion drives buffering strategy, threading model, and more.
That matches our own account in How MoQT Subscriptions Work, and it is the mental model shift that costs teams the most time.
Owen demonstrated the module at NAB 2026 on the Cloudflare booth, combining Streaming Engine, the Cloudflare MoQ Relay, and Shaka Player.