Ability to snub peers.

[m_cable]

I know there was a big thread regarding manually banning peers, but everyone on that thread is saying that nobody has a good reason why it should be implemented. And I have a good reason, and I don't want it to get buried in that thread without the program writers seeing it.

Snubbing peers is enormously useful for seeders. Especially initial seeders without a huge pipe. If you're in a big swarm with one seeder that has only 40-50 kB/s up then invariably what happens is that you'll have dozens of people stacked at the uppermost percentage completed. Say if the torrent is 60% available, then you'll have half the swarm at 58-59%

With superseeding you can go at a steady pace up til 95-99%. But some pieces always get lost in the shuffle. At that point, it's very important to be able to "snub" (as Azureus refers to it) all the other downloaders, and concentrate only on those that are almost finished.

I know that the algorithm is supposed give priority in sending out the rarest pieces, so theoretically it should finish off those peers that are almost done. But I've been in enough swarms to know that that isn't the case. For some reason, the lone seeder seemingly always starts uploading non-rare pieces to peers that aren't even close to done.

I've had some torrents where it took 7 hours to go from 0-95%, and then an additional thirty minutes to an hour to finish off that final 5%. Even if that was the only application, of this feature then it would be worth it. But there are plenty of scenarios where this is incredibly useful.

For example, if you've been seeding a file for a while, and uploaded more than the typical 1:1 ratio, but all the other seeds have left and you remain the only full copy, then I think it's completely legitimate to go finish off a peer or two so that you can log off. This lets you keep the torrent alive, and also move on for whatever reason.

And what's wrong with giving preference to your friends. You're still seeding. You're still getting pieces out to the swarm. If you happen to have over 1:1 upload/download ratio, then there are no ethical problems.

Yes there is a possiblity of abuse, but there's also a possiblity of abuse in being able to limit your upload speed. Does that mean that we shouldn't have the ability to change the max upload as we see fit? Closing the torrent after you're done leeching is an abuse too, but should the program be hardcoded to seed until it reaches a 1:1 ratio? IMO, the possibility of abuse should not be a reason not to implement this feature. Almost anything can be abuse if a person is determined to do that.

I really think the program writers should reconsider their decision to not include this feature. It really is an incredibly useful option, and near-vital in some situations.

[silverfire]

If I remember correctly, super-seeding will only send out one copy of each piece, and will refuse to send any duplicate pieces, so any speed detriment you see should be pure myth.

[paritybit]

Wow, lets have another 6 pages on the same topic saying everything all over again. This topic is reductant. I believe your points were raised in the other topic and covered over and over again.

Then we have people trying to hold the developers to ransom by saying the client is no good without this feature etc etc. This is hardly a critical feature that isn't absolutely necessary. I believe the devs read most of the topics and this thread is reductant.

[slayers]

Ok, I don't wish this to be another endless thread like that other one but I will address your point on the superseeding because there is some validity to it.

First off, manually snubbing/banning IPs is not the answer. It is wrong for all the reasons already posted on that thread. Now, that doesn't mean that the superseeding mode can't be perfected. For a start, it was never part of the original bittorrent protocol so it's not a finished concept. Isn't there an endgame mode that would resolve the issue you mentioned?

Whatever solution is adopted it should NEVER have to rely on human intervention. That is how we get into trouble in the first place.

EDIT- I just looked up endgame and it kicks in during the final phase of a download, not upload. But perhaps something like a superseeding endgame could be developed.