Wired quotes
The Pirate Bay (TPB) isn't happy with the fact that BitTorrent Inc. is no longer making additions to the BitTorrent source code entirely open and so TPB is hard at work on a new file sharing protocol that will address many of the shortcomings of BitTorrent.
The new protocol, which will create files with the extension .p2p, will be backward-compatible with .torrent, and yet, somehow, reportedly will be designed to limit the effectiveness of both spammers and anti-piracy organizations.
So far no real details are available and frankly those goals sound mutually exclusive — how do you simultaneously make something more transparent and harder to track?
Current limitations ?
1) Current P2P nets don't use upload bandwidth optimally
2) Old nodes have no incentive to communicate with new nodes (without data)
3) Rare data is as valuable as old data
Quotes:
We are building a protocol and format from scratch, but a lot of the suggested torrent improvements can be used and implemented in the new protocol.
File-based instead of bundle based (like .torrent files) tracking would really increase the availability of content and really make it harder to have large spreads when uploading fakes.
In comparison with a torrent file, which has a complex encoding, the ".p2p" format is simply a file in XML format containing the list of files in the bundle. This enables each file to have file-hashes making life more difficult for fakers and spammers. Or you could use/extend .metalink XML files, which already contain this information and are supported by many download programs.
Running TPB, I've been thinking a lot of how to make the content cleaner for the end user. Spam-blocking, fake-blocking, and being able to clean file-names in bundles is a big thing that I would like to see more of in the new protocol.
Also having the same content from several uploaders is something we see a lot of, as well as on most of the other major trackers, so being able to share peers between trackers and between users with the same files in different bundles on the same trackers (through hash search services) would really increase download speed and availability.
Counter measures to defeat traffic analysis would be interesting. This would also lead to a semi-anonymous system that would allow for plausible deniability. i2p already meets the need for provable anonymous p2p, although at a higher overhead than might be considered acceptable by this protocol, but without much wasted bandwidth this protocol could provide for sufficient obfuscation.
