MOQ had a good week.
Not a “someone shaved 14 milliseconds off a demo” kind of week. A much more useful one.
It got a job.
MOQ goes to Tokyo
A VTuber fan event in Tokyo used Bootstrap’s AIAI service commercially: 4K video and audio went from a performer at home to fans at the venue, with audience voices and venue video coming back in real time.
The performer stayed home. The pixels went out.
Honestly, that may be the cleanest explanation of MOQ’s appeal I’ve seen.
The performer could keep using OBS Studio and VTube Studio. The rendered performance travelled over the network; the avatar’s original model files stayed at home. Less setup at the venue. Less explaining which folder contains the person.
Video goes one way, audience interaction comes back the other way, and the whole thing belongs to the same real-time experience. No need to build a small zoo of unrelated protocols and hope they remain synchronized.
Interop got real
The MOQ community also held another interoperability session on September 2, targeting transport draft 18.
The published interop notes cover fetching media, publishing tracks and checking that relays preserve the details they receive. The sort of things that look obvious until someone else implements them.
And, as all good interoperability sessions eventually do, it invited the familiar conversation:
“Your implementation is broken.”
“No, your test is broken.”
“Oh.”
Meta’s moxygen actually failed parts of its own test suite when run across a real network. Some of the failures turned out to be timing problems in the test harness itself.
This is good news.
Protocols become real when independent implementations begin finding embarrassing bugs in each other. Connecting two copies of your own software proves considerably less than developers would like to admit.
Advertising has entered the chat
Paramount’s Steven Riedl published a new ad creative signaling draft covering ad creative identity, tracking and verification metadata over MOQ.
This may be the strongest evidence yet that MOQ is growing up.
Eventually every promising media technology reaches the same milestone:
Someone asks, “How do we bill for this?”
SCTE-35 can tell you an ad break is happening. This work starts addressing which creative fills that break and how the surrounding measurement information travels with it.
The proposal puts those details on a companion event timeline alongside the media. Sending them ahead of the ad gives the player time to prepare its tracking, instead of discovering its accounting obligations with the first frame.
Low latency is exciting.
Getting paid remains popular.
MOQ discovers DNS
There was also an updated draft on DNS and local discovery for MOQ endpoints.
This sounds spectacularly boring.
Which is excellent.
At some point every protocol has to stop discussing elegant transport semantics and answer a question from operations:
“Cool. Where is the server?”
The draft covers both looking up a service through DNS and discovering one on the local network using mDNS. Useful when the answer is a nearby relay, rather than an address someone pasted into a configuration file.
Work also continued around TLS behavior, cache availability signaling and 0-RTT.
These are the topics nobody puts on the conference banner, but everybody needs before putting a system into production.
Open source gets plumbing
MediaMTX added MOQ stream forwarding.
Again, gloriously boring.
A stream can arrive over MOQ and be forwarded onward to another MOQ endpoint. That starts making real relay topologies and gateways considerably easier to assemble.
Forwarding is built in: configure a destination and MediaMTX handles the connection, without a separate FFmpeg process just to pass the stream along.
The MOQ concept car is slowly acquiring brakes, mirrors and cup holders.
Nobody puts cup holders on the prototype.
Customers always ask for them.
And now there is a MOQ boat party
IBC is next, and MOQ activity is ramping up across vendors including Broadpeak, Bitmovin, Oracle and Nanocosmos.
There is also another MOQTail boat tour planned in Amsterdam for Sunday, September 13 at 18:00. Places are limited, with invitations confirmed separately after registration.
So the protocol has officially reached the point where people are discussing cache semantics during the day and getting on a boat afterward.
TCP had to wait decades for this.
The bigger picture
The most encouraging thing about this week was how little of it depended on a flashy new feature.
The conversation is moving toward interoperability, discovery, monetization, security, caching, forwarding and operational behavior.
Those are wonderfully unsexy problems.
They are also the problems technologies encounter shortly before people begin depending on them.
MOQ is starting to look less like an interesting protocol experiment.
It now has paperwork.