[LINK] Huston: 'The Future Internet and Peer to Peer'

[Glen Turner]

On Mon, 2007-06-25 at 17:55 +1000, Howard Lowndes wrote:
> Methinks that Geoff has just gotten himself in the sights of the RIAA 
> and MPAA.

Hi Howard,

Let me give some more background information.  The network engineering
problem here is that a lot of peer-to-peer networks have "supernodes"
or other concentrations of network activity, but these nodes are not
placed on the optimal points of the network.

This is because the p2p applications are not given any network
engineering data. That data is currently available, via the BGP
routing protocol, but as Geoff indirectly notes that is available
only to prearranged network routers, not to dynamically-appearing
p2p application "routers".

There are a few issues here. How can network engineering data be
made available more widely without it being misused (such as for
planning more effective DoS attacks)?  Matt Roughan at The
University of Adelaide was the first researcher to note the need
for this and he has completed the research to answer this question.

The next issue addresses your problem: will p2p applications use
network topology data? The "yes" answer's argument is: p2p software
will run faster if it knows the topology, there's a lot of problematic
topology-discovery code in p2p apps that can be removed. The
"no" answer argues that accessing the network topology feed gives
RIAA/MPAA/etc and local network administrators a point to detect p2p
applications, and a major user driver for p2p is avoidance of detection
(using port 80, etc).

The truth is probably somewhere in the middle -- bittorrent would
probably use the topology data, since it has never been fussed about
avoiding detection. Some other applications (which the new copyright
changes prevent me from naming) would continue to value stealth above
all else.

Cheers, Glen