[p2p-hackers] Generalizing BitTorrent..

[Vaste]

Bryan Turner wrote:
 > Hello p2p-hackers,
 >
 >     I've been thinking about BitTorrent recently and had some ideas on
 > improving the protocol.  I'm sure others have had similar thoughts,
 > and I'm interested to hear your opinions.
 >
 > [situation where e.g. some peers are interested in A, some in B and
 > most in C]

I'd say the problem boils down to finding trading partners with common 
interests. (How to do this decentralized?) The Fedora guys share an 
interest in Fedora data and would likely benefit from being introduced 
to each other. They still share the interest in MAME the rest of the 
MAME-swarm has, but these two peers in particular have high common 
interest (how to measure? MB of common interest?).

But, how does this affect the randomness of BitTorrent's network? Let's 
say there's 50 or so Fedora&MAME-guys. Then these might be happy to 
connect only to each other (e.g. they've found 30 peers that are 
"better" than the rest of the MAME-swarm) and an isolated island might 
be formed. This is obviously contra-productive e.g. if there's no 
MAME-seed in the island, as then no MAME data will ever spread there.

But if one does "normal" random selection (ignoring common interest) the 
Fedora-guys might never find each other at all, if the MAME-swarm is 
sufficiently large. So what to do?

/Vaste