So KT starts to download the torrent from the beginning of the file, right? At least that's what I've observed so far. And another thing I observed/believe is, through the end of the torrent it finds less peers that it can download from. If every torrent client starts downloading from the beginning of the torrent, the end of the torrent will be scarce in the pool. So, do you guys think it would be a better idea to download totally random parts of the torrent? Or is it already there?

Chunk selection is done like this :

- If there are music or movie files, we prioritise the first chunk of these files so the preview feature can work.
- Normal chunk selection selects the rarest chunk first (this helps the swarm), to be more precise the rarest chunk a peer has.
- There is a limit on the chunks we can select, so that we do not have to write to much data to disk at once. This limit will gradually grow until all chunks are available. We violate this limit if we cannot find a chunk for a peer to download.

In the beginning there is a bias for the first chunks, but this will disappear.

So selection is pretty random.