Re: DHT vs Gnutella

[email protected] Wed, 6 Aug 2008 07:59:00 +0000 (UTC)
Newsgroups gmane.network.gnutella.devel
Organization Home, Grenoble, France
Message-ID <[email protected]>
Quoting Nikolay Raspopov <[email protected]> from ml.gnutella.dev-forum:
:Hello, [email protected]!
:You wrote to <[email protected]> on Tue, 5 Aug 2008 20:11:15 +0000 (UTC):
:
: RM>In a flood model, search horizon is limited
:
:Then improve it! That why this list created! Suggest algorithms to reunite all
:Gnutella ultrapeers to one connected network.

That algorithm exists. It's called Kademlia.

: RM> In a structured network such as the one provided by Kademlia, the
: RM>topology guides you to the nodes where you are most likely to find  the
: RM>data.
:
:You cannot guaranteed that nodes will publish rare files in time especially for
:large library with thousands and thousands of keywords and hashes. Moreover
:storage nodes may go offline and you must republish all data again and again...
:insane. Users will be "happy" with infinite downloading of all this key-value
:pair data...

Just like users were happy in the good old days when Gnutella was flooded
with pongs and query hits.

Raphael