On 07/07/2012 04:13 PM, Alexander Lehner wrote:
On Sat, 7 Jul 2012, Doug Jones wrote:
[...]
I used the trick mentioned earlier in this thread. I trimmed the
?torrent off of the end of the URL and started downloading the entire
file through Firefox. After a few megabytes I paused the download,
copied the partial file into the download folder used by my bittorrent
client, and told it to torrent that file. It contacted the tracker at
amazon, which is indeed working, and the tracker found one peer (you),
and sent you those megabytes.
So the amazon tracker is working. And the file is indeed present on
the amazon server. But the amazon tracker isn't using the amazon
server as a seeder!
So something is still wrong.
I now managed to download the zip files and seeding them now (for depedia).
Il will seed others also, in the hope, that the torrent network comes
back to live.
Off-topic soon, how-to:
I used azureus/vuze.
First downloaded the depedia*.7z files (as described earlier, without
.torrent).
Copied them into my torrent download folder.
Asked vuze to download the torrent 'depedia*.7z.torrent' from the
http://dev.thewikireader.com/language-packs/
page.
Got warning to continue/restart download, acknowledged that.
Started to seed then...
---
Just wondering:
I was downloading the files from web with approx. 600kB.
This is much faster than any torrent network will provide and faster
than most other download sites.
Who - if ever - will have to pay for this download speed?
Does it make sense at all here, to use torrent network to distribute
such kind of data?
A.
I assume that Openmoko is paying for the bits downloaded from that
Amazon server. Of course, when we ask our bittorrent clients to
download and seed the files, Amazon doesn't see the traffic going
between other peers so Openmoko doesn't pay for that part. This is part
of the reason we use bittorrent -- we are volunteering to take some of
the bandwidth load ourselves instead of asking some centralized server
to cover all those costs.
Another reason for using bittorrent is this: Suppose a group of peers
already have, between them, a complete copy of a given file. (No one
peer needs to have all of the chunks; each chunk just needs to be on at
least one peer in the group.) Then we just need to get a copy of the
corresponding .torrent file to each peer, by whatever means, and then
the peers can collectively act to build a complete copy of the big file
on each peer. No centralized tracker is needed, as long as the peers
can find each other. This brings the fully decentralized robustness of
the Internet to torrent distribution. (Okay, DNS isn't fully
decentralized yet, but that's another story...) This is what Tribler
does. It allows peers to find each other without a tracker. I expect
all bittorrent clients will eventually gain this capability.
As to the question of whether bittorrent makes much sense for these
particular files: That's a good question.
If you have a bunch of peers, and one of them can pump out data at
600kBps and the others are much slower, then the fast one will
presumably hand out most of the chunks and the slower ones won't do much
at all. If the fast server is worried about cost, then it could
throttle itself down when there are other peers seeding. But throttling
itself down to zero (effectively that's what's happening now) is also
undesirable, especially if the other peers don't have a complete copy
between them yet ;-)
_______________________________________________
Openmoko community mailing list
community@lists.openmoko.org
http://lists.openmoko.org/mailman/listinfo/community
