[p2p-hackers] Live P2P Video State of the Art

[David Barrett]

We've talked at length about massively scalable file transfer, but we've 
generally presumed fixed-length, pre-recorded files (ie, the file is 
available in entirety before sharing).  I'm curious if you have 
experience or ideas around "live" streaming of content simultaneous to 
its recording/creation.

Clearly, this isn't a new field, and streaming architectures abound.  
But while there has been extensive innovation in file sharing (DHTs, 
merkle trees, swarming downloads), I haven't heard much innovation with 
live streaming content.

So far as I know (but I'm asking you to prove me wrong), the state of 
the art in audio/video streaming is still a classic "hierarchy of 
repeaters", where the broadcaster sends to N receivers, each of which 
sends to N receivers, and so on.  There are obvious variations on this 
theme (adaptive fan out, topology optimizations, re-request of dropped 
data, jitter buffers, etc) but I'm looking for the next major innovation 
(such as was taken between Napster and Gntella for file sharing).

With this in mind, do you know of any proven techniques (or new 
research) in grid/swarming delivery of live video?  Which research, 
projects, or products would you say are demonstrating the state of the 
art in scalable, adaptive, high-quality video streaming?

-david