Weekly Update - 4/7/2017 (Protocol / Video Segment)

Hello every!

This week, we’ve made progress in the protocol design and the media server.

The Livepeer protocol has gone through some nice improvements as we continue to bang it against different threat models and scalability challenges. We’ve mapped out a staking mechanism which makes fake useless work economically irrational, and fairly distributes network fees and block rewards while still maintaining high quality transcoding. We are also very excited about the Truebit project from the Ethereum foundation and its potential to be an important piece of our validation mechanism. Here is a video and a podcast about the Truebit project for anyone who isn’t familiar.

For the media server, we have standardized video length / format in the network so transcoding can be validated on-chain. We did this by making video validation / segmenting a necessary step when a RTMP stream is published to a node. The segmenter uses FFmpeg to create 2-second HLS video chunks with consistent leading keyframes, and we will use that as the unit for transcoding work. As a side effect, now a RTMP stream will be automatically turned into a HLS video stream, consumable from the original node. (You just have to add .m3u8 at the end of the URL when requesting a video) Your can check out our LPMS and Go-Livepeer PRs on Github.

Also related to Truebit…it is an important project for verification off-blockchain work, but it is also very expensive, with estimates being that it costs 5-50x the cost of the original work in order to ensure verification.

One of the innovations of the Livepeer protocol is working out “scalable Truebit”, in which you can verify only certain segments of work randomly (say 1 out of every 1000), and still provide the correct incentives for the transcoder to have behaved correctly on all segments in order to earn the transcoding fees/reward.