From observation, I could see that actually BC doesn’t sit idle and wait for pieces if they’re not available, it will download almost always some random pieces too, along with the sequential ones. I don’t know exactly how it handles this internally;
it may be that it selects a couple of peers from which it requests by the normal algorithm or that once in certain while it switches back to “rarest first” mode. The fact is, that if you watch the pieces chart you’ll almost always see some random pieces downloaded.
I’ve never seen the client sitting idle, waiting for some piece.
Besides, the client always launches requests for multiple pieces, so it’s bound to at least receive a part of the responses as positive.
Besides, since the normal peers in the swarm work by the “rarest first” algorithm, if the swarm has a reasonable enough number of peers, or if it has only a few peers if some of them have high upload speeds (as it often happens on private trackers) then it won’t make any difference, since the pieces will be rather homogeneously spread across the swarm.
The thing is, that peers don’t really “offer” pieces; they all respond to **requests **from other peers by sending or not sending the requested piece. Your client determines what pieces it can request based on the HAVE messages sent by all other connected peers, and thus it can construct a map of the available pieces beforehand, prior to starting to request different pieces.
So, by your reasoning, if the first 15 minutes are rare that’s what your client would download anyway by the normal “rarest first” algorithm.
The only way it may hurt is if let’s say the pieces in the period 15-30 minutes are rare, and your client will focus on the first 15 minutes, since that’s what it needs now, so the swarm will have one less peer which contributes at downloading rare pieces (so as to make them “not rare”) and occupies some swarm bandwidth by downloading more popular pieces from other peers. But this will be only temporary because when it gets to the next 15 minutes it will download exactly pieces which are rare. So, the whole thing is rather shifty.
Due to the inherent complex nature of the internal dynamics of a swarm, I do not pretend that I can tell with absolute certitude how much, requesting sequential pieces, would affect the swarm in a negative way (it may involve a bit of statistics math and even some empirical tests to draw a really solid conclusion), but based on the fact that in average there are very few users (even among BC users) who use this on a constant basis the overall average of sequential downloaders among all clients inside a swarm is pretty small, so the impact (even if there is any) will be negligible, in my view.
