[bittorrent] Fwd: Piece downloading strategy
alan.mcgovern at gmail.com
Sun Aug 5 20:12:26 EDT 2007
(forwarded to mailing list)
These statements need to be qualified and/or justified (a reason
> "why" is needed). A linear strategy is reasonable if the client's
> goal is to stream the content to the user while downloading. Strategy
> 1 won't overload seeders any more than strategy 2 (as long as early
> peers aren't started around the same time) but produces poor
> distribution of the later pieces. The statements assume that the
> client's goal is to maintain the widest availability of pieces, but
> it should be stated and explained why this is good/necessary and
> how each strategy supports or hinders this goal.
Very true, and definitely something to be mentioned. When describing the
algorithm i will be sure to mention the scenarios where it'll be useful and
the scenarios in which it won't be.
The problems with the first two
> strategies happen largely when most peers are using that strategy,
> not when one uses it as an exception.
Also true. Those strategies will be mentioned aswell, but they will be
mentioned as sidenotes at the end as opposed to given first mention in the
section. Completely random and linear are fine under certain circumstances.
I think that the algorithm description in the document's current
> second paragraph is fine, and would continue to be so in conjunction
> with your proposed changes (replacing the first paragraph/sentence).
The second paragraph is pretty much what i'm going to write. It's just i'd
much prefer if the second paragraph was the highlight of the section rather
than a sidenote at the end, hence the reason i want to rewrite it. For the
standard case where you just want to download a file, a pseudo-random
rarest-first algorithm would offer the best strategy for downloading.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the BitTorrent