On Thursday, October 26, 2006 Sam Berlin wrote:
> FWIW, LimeWire tries to determine if the file being transferred
> is previewable, and if so will prefer swarming from the beginning
> of the file (but also swarm from random sections).
Right. LW and BT might be operating in the different domains,
though. In BT universe, most of the sources do not have the full
copy of the file - thus the need for the "rarest first" approach,
and the negative effects of the sequential download on the pieces'
availability. Basically, in BT even in the stable network state
(approximately constant number of downloaders over time),the
sequential download results in the linear reduction of the chunk
availability as you move from the beginning of the file to the end.
Almost everyone you see has the first chunk, half of the people -
the middle chunk, and only a few seeders - the last chunk. And
flash crowds make this probability drop even more pronounced.
On the other hand, I wouldn't be surprised to find that the
majority of sources in LW have the full copy of the file; that was
definitely the case before the network grew into millions and the
download mesh was introduced, but still may be true even today. In
this case the network as a whole does not care about the download
order, because there's no need to replicate the rarest pieces fast:
all pieces have roughly the same number of copies anyway. So you
can download in any order you want - including the sequential order
- and this previewing feature won't hurt anyone. With BT, the
situation is more complicated.
Basically, the more you rely on the download mesh and less -
on the global search to find the sources, and the more is the ratio
of downloaders to seeders, the closer you become to the BT model,
and the more difficult it is to stream stuff.
Best wishes -
S.Osokine.
26 Oct 2006.
-----Original Message-----
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] Behalf Of Sam Berlin
Sent: Thursday, October 26, 2006 12:15 PM
To: theory and practice of decentralized computer networks
Subject: Re: [p2p-hackers] BT2 & Question
FWIW, LimeWire tries to determine if the file being transferred is
previewable, and if so will prefer swarming from the beginning of the
file (but also swarm from random sections). For unpreviewable files,
or if the computer's been idle for a certain amount of time, the
entire process is random (but attempts to keep the number of chunks to
a small number, building off 8 or so random sections of the file).
Sam
On 10/26/06, David Barrett <[EMAIL PROTECTED]> wrote:
>
>
>
> Red Swoosh (www.redswoosh.net) does sequential P2P swarming for "on demand"
> streaming video playback. For example:
>
>
>
>
> http://edn.redswoosh.net/www.redswoosh.net/AskANinja.wmv
>
>
>
> What's the difference between "chunks prioritization within time window"
and
> "sequential downloading"? How can you stream without sequential
> downloading?
>
>
>
> -david
>
>
>
> ________________________________
>
>
> From: [EMAIL PROTECTED]
> [mailto:[EMAIL PROTECTED] On Behalf Of
> Florent THIERY
> Sent: Thursday, October 26, 2006 6:54 AM
> To: theory and practice of decentralized computer networks
> Subject: [p2p-hackers] BT2 & Question
>
>
>
> http://tech.groups.yahoo.com/group/BitTorrent/message/5481
>
> I don't get why they are not implementing a Bittorrent variant for
> play-while-downloading (chunks prioritisation within time window). Read a
> paper once about bittorrent mod, saying that's more efficient than
> sequential downloading. Ever heard of swarming-enabled p2p protocol for on
> demand ?
> _______________________________________________
> p2p-hackers mailing list
> [email protected]
> http://lists.zooko.com/mailman/listinfo/p2p-hackers
>
>
>
_______________________________________________
p2p-hackers mailing list
[email protected]
http://lists.zooko.com/mailman/listinfo/p2p-hackers
_______________________________________________
p2p-hackers mailing list
[email protected]
http://lists.zooko.com/mailman/listinfo/p2p-hackers
