Video Conferencing
Not logged in

Feature Requests

Wish list

These are “nice to have” features with some caveats.


While the network itself is peer to peer, the roles in a video conference have a central point, the moderator. The moderator collects the streams that are created at the endpoints, mixes them, and distributes the result to everybody.

This is easy when the central instance is a well connected server. With a p2p network, you have to distribute the load of both collecting the data and distributing it, but the central point of control (the moderator) remains.

Life streaming

net2o's basic paradigm is that of storage objects, so every object (file) has a hash, and sending participants can authenticate objects with a signature. Life streaming however means new information is added on the fly.

Participants send metadata information on every update: (size, hash) for every object, and a signature for all objects.

A complete video stream consists of several objects:

Video data encoding depends on what's available as hardware encoder with the participants.

Tradeoffs: Software video encoders have good quality, but are slow. Hardware encoders have lower quality, but are fast and don’t consume much power.

Life streaming requires real–time distribution of data from one to many participants. net2o uses an avalanche tree distribution for streaming: Data stream blocks are partitioned. Each partition is send to a tree of nodes. The branch nodes send it forward to further nodes. The leaf nodes send it forward to nodes from other partitions. This makes sure that every node has the same load.

Avalanche tree

The dangling edges that point nowhere show that it is possible to add one further node to this tree at any point in time. The tree allows to have low–latency streaming to many participants. The fanout here is four; in reality, it can be significantly bigger.

For timeshift streaming alone, a much simpler relation is possible: an ordinary queue. Every watcher starts by being added to the tail of the queue, watching the video stream right from start. It takes that stream from the one next in the queue, which got it recently. If the delay between two participants becomes too small for that simple queue, the real–time streaming tree concept can be used for them.

Timeshift streaming watchers who skip over parts have to change the position in the queue. Since they then don’t have the entire video, they can‘t distribute all of it. Everybody still can contribute as much as he consumes, it is still a fair share.

Collection tree and mixing

Each active node in a video conference creates a stream of data. Mixing the data is distributed; sending it directly to one central instance will easily overwealm that one.

The own recording is available afterwards for remixing the entire conference with higher quality, or in case connection losses took out participants and people want to check what they missed.