I think this project has a chance in ten of working. But I hope it does.
I set up Riot with a group of friends in an old used laptop last year, and it has been working surprisingly well for us. I hope we can try this out soon!
Note re live streaming: a minute of delay between streamer and audience makes interactivity (even just Q&As at the end of a presentation) essentially impossible, and Peertube's particular design makes it unlikely that'll improve by much. It does make it somewhat cheaper to host though.
I'm working on a self-hostable live video streaming platform, building on solid, well-known technology like MPEG-DASH's low latency mode and a relatively dumb (and easily replaceable) content delivery network.
Preliminary tests suggest that my system can get to about 3 seconds, which makes interactivity between streamer and audience a lot more feasible. The downside is that this costs money - about 12 cents per GB through e.g. Fastly, on top of hosting and origin bandwidth costs.
True, but esport isn't most streamers either. The Mario Maker streamers and speedrunners I follow have heavy interaction with their chat on-stream, for example, and couldn't switch to a system which gave them significantly higher latency.
That's an interesting case study! In that particular case, how much is the maximum latency you think it can afford? And do you think the latency must be equally low for every viewers, or is there a vocal minority of viewers who interact with the streamer while others watch passively? If it's the later, maybe there would be a way to keep a reasonable (same as your twitch example) latency for those viewers.
As someone who occasionally streams, max latency would be around a consistent 7-10 seconds - that makes some interaction patterns (the quick "which choice do I pick?") harder but still allows some reasonable level of feedback.
Part of the reason I personally watch streams is to feel connected to the rest of the audience - when something good happens we all cheer immediately, when something bad happens there's immediate commentary. Being split up from part of the audience would cause me to feel weird - as one of the vocal minority I'd be cheering in a much smaller crowd, and as one of the passive viewers I'd always be wondering what the vocal minority were on about.
There's also another type of stream that's popped up on reddit's RPAN - just random people chatting and interacting with their audience. That's not really feasible in Peertube's model.
Note that there's some types of streams that Peertube's model could be great for - pre-planned event shows without much interactivity, live captures of current events, and so on. GDQ, various music and performance events, broadcasts from protest zones, and so on could use this.
Thanks. I asked, bc I had a vague recollection of their having open-sourced a key part of their astonishingly low-latency streaming infr, and that's the 1st thing I'd look into if I were entering this space.
(I'm not affiliated to PeerTube, I'm just a user)