Researchers To Release an Anonymous BitTorrent Client
torrentfreak.com
torrentfreak.com
Right now, it's hard to tell. I like that their current protocol is published as an IETF draft [1], but it doesn't include anything about their 'proxy layer' which is touted in the TF article.
From the details in this article, it seems to work a bit like Tor, which raises a few questions: who is going to run the ProxyNodes: is that their business model: making people pay a subscription to access their vast network of ProxyNodes? (Will they claim innocence because "they don't inspect the content passing through", even though it'll be mostly pirate content).
PPSP (the IETF version) looks like a more modern and lightweight BitTorrent that runs on UDP. I like that, but as every network, we'll have to see the adoption rate. People stay on BitTorrent the same reason they stay on Facebook: everybody's there.
The Tribler guys are not the only one to want to do P2P on top of UDP though: uTorrent came out much sooner (and with reasonably good adoption) with μTP [2].
[1] https://datatracker.ietf.org/doc/draft-ietf-ppsp-peer-protoc...
[2] http://blog.bittorrent.com/2009/10/05/changing-the-game-with...
This is true in general, but there are some major players who could just as easily move people to adopt. i.e. if one day TPB decides that magnet links doesn't solve as many problems as this, people would certainly follow. If anything, at the beginning people probably wouldn't mind having a different protocol or client just to run TPB seeds.
I could be wrong, but that seems like a big showstopper.
If Wikipedia is to be believed Tor mainly uses Onion Routing with a set of servers as the core of the system.[1] The Tarzan paper mentions why this might be problematic and suggests instead a decentralized P2P system, using a mix network with (what you might call) onion routing and cover traffic.
The authors claim this eliminates the "single point of failure" and the effectiveness of even network-wide analysis. And since Tarzan should be ideal as a drop-in middle-layer, I'd think this would fit an application like BitTorrent.
Unfortunately we never got to use Tarzan in our project, so I of course only have the paper to go by.
[0] ftp://163.25.117.117/gyliao/iSmart/1-P2P/2-Anonymity%2BIdentity%20Management/Tarzan-A%20Peer-to-Peer%20Anonymizing%20Network%20Layer.pdf [1] http://en.wikipedia.org/wiki/Tor_(anonymity_network)
There's also OneSwarm which has been out for a while: http://www.oneswarm.org/
It's beginning to annoy me how HN rejects double postings in general, but includes the fragment identifier again allowing unlimited reposting.
You can also use E2DK or Gnutella in I2P. Or plain HTTP. Or mails. Or IRC. Much nicer than being limited to just Bittorrent.
As for speed, if you get a nice swarm and your I2P peer is well integrated, 100Kilobytes/s are possible.
There are some "outproxies" but only 2-3 and they are slow. I2P is not meant to intermingle with the normal internet.
It is a very, very hard challenge, and I wish them luck.