When a peer is super-seeder/intialially seeding, and a KTorrent client joins, the unneccesary data can me Gigabytes when there are fast peers, which is quite a little lot to me.

Can there be done anything about this? Because this is just too much...

Have you got a torrent showing this problem ? I don't see any obvious reason why we would get a lot of unnecessary data when connected to a superseeder.

I have thought about this some more, and I think it is quite possible :

- superseeder notifies a couple of clients that it has chunk X
- several clients download it
- ktorrent joins swarm
- sees that there is only one chunks availabe and starts downloading this
- because there are multiple clients who have it, ktorrent will assign them all to that chunk

The result: massive overdownloading when the last 2 steps are repeated over and over (superseeder will only give a peer a new chunk when it has seen a have message from another client, i.e. the peer must share the chunk before it gets a new one to download)

I never liked superseeding, but clearly we need to solve this.

Yeah, it seems KTorrent does not cancel pieces that much when there are a lot rare pieces available at the same time and when it are the only one available.

Cancelling happens, but when a piece is already in transit, it has to be fully downloaded.