Jump to content
To block spammers, this forum has suspended new user registration ×
Comet Forums
To block spammers, this forum has suspended new user registration

being able to spot unique pieces and percentage when connected to a peer


Recommended Posts

Hi all,

When we are downloading through some peer, one wants to be able to know how much percentage of unique pieces the other peer has. What happens now is you just get the other peer's percentage number and the peer status graph. The graph is not a good visual indicator of how much unique pieces the other peer has. It would be great if this could somehow be done.

Link to comment
Share on other sites

If you actually had that information, down to the piece level, what would you do with it?

You can't force another peer to give you any pieces, let alone a specific piece.

You have no information about whether a piece is "unique" -- you can only know that if you could somehow compare bitmaps of all the peers in the swarm, many of whom you aren't connected to or have not been offered a piecemap by. The swarm is not homogenous, it "clumps", so a piece that's common in one clump is very rare or unavailable in another. People write dissertations on why this happens.

Bittorrent negotiation is a very complex process, and a very rapid one. You are hopelessly slow and present a traffic jam if you try to control the process. Decisions need to be made in microseconds, with more information than you have available. Human intervention can't speed up the process or help it along, but it can sure gum things up.

Link to comment
Share on other sites

Even if someone was able to design a way of determining what pieces were rare and requested only them, by the time they got them, they would no longer be rare, and the common pieces they passed up when looking for rare pieces would then be rare, making it a hopelessly inefficient vicious circle.

Link to comment
Share on other sites

I should add that it's a very common reaction when people are first looking at packet-switching algorithms, to think they can be improved by human management. That might be true, but only if you could operate at computer speeds. It seems to go against the grain, but you have to learn to trust the network to do its job.

Link to comment
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×
×
  • Create New...