[BitTorrent] XML Torrents?
ashwood at msn.com
Sat Mar 12 05:55:23 EST 2005
----- Original Message -----
From: "Kenneth Porter" <shiva at sewingwitch.com>
Subject: Re: [BitTorrent] XML Torrents?
> Looking at the spec <http://www.bittorrent.com/protocol.html>, I see a
> filename (or list of filename/length pairs), tracker URL, and a set of
> fragment checksums. Most of the size would be the checksums. How do you
> propose shrinking this to fit in a URL? I suppose you could designate
> trusted peers (eg. initial seeds) to host the checksums.
I believe the concept there is to (as we have been discussing in great
detail with very mixed results) use Merkle trees, this eliminates the long
list of hashes at the end. The proposal also includes the presumption that
multi-file torrents have to be redeigned substantially (again many concepts
and mixed results). By doing this the necessary elements of the torrent
become tracker, filename, root hash. There would have to be come changes in
the formatting of the URI, basically just encoding, but the end result could
look something like:
I assumed the inclusion of the recently (briefly) discussed hash_function
argument, and obviously SHA-512, although I am not sure "-" is valid in
URIs. Any authorities? Even at that we have the option of (assuming the
tracker or peer protocol is altered to support this, I would vote peer) the
URL can be reduce to not include the filename either.
Even without the further optimizations at 218 bytes this forms a usably
sized, if hard to remember, URI. The same information could be stored in any
other convenient format as well, with bencoding of course being the default,
an XML definition (even though so many people don't like it), etc. It fits
easily into a TCP/IP frame, ethernet frame, Wi-Fi frame, but does not fit
into an ATM frame, not sure about modem frame I haven't looked at that
information for quite some time. The frame fitting is not critical but it
can relieve overhead in a system, and was one of the primary reasons for the
size limits on URI in the beginning (IIRC 4096 bytes to match the maximum
ethernet frame, and yes I am old school enough to have actually done heavy
optimization based on this and achieved throughputs higher than the
theoretical maximum because of it).
I actually think that this poses a very interesting new area for BitTorrent,
or a BitTorrent-like protocol, and at the very least something that can help
in optimizing the system.
Yahoo! Groups Links
<*> To visit your group on the web, go to:
<*> To unsubscribe from this group, send an email to:
BitTorrent-unsubscribe at yahoogroups.com
<*> Your use of Yahoo! Groups is subject to:
More information about the BitTorrent