Re: DHT vs Gnutella

Quoting Arne Babenhauserheide <[email protected]> from ml.gnutella.dev-forum:
:Am Dienstag 05 August 2008 16:13:06 schrieb Nikolay Raspopov:
:> Please don't forget that DHT is NOT a Gnutella. And instead of developing
:> Gnutella future you wasting time by creating yet another DHT-like
:> protocol... IMHO its too "dot-com-boom-ish'...
:
:The DHT can supply a missing piece to Gnutella: searches by Hash. 

Nikolay: we're not developing another DHT-like protocol. The architecture
of the DHT is Kademlia-based, and this is a given nobody here is contesting.

Now of course the messages echanged to create the DHT structure are totally
different from the one used by Kad (e-mule) or by bittorrent's DHT.  But
this is purely incidental, as the DHT data architecture and purpose are
completely different, and this is what makes them separate instances.

Arne: the DHT is going to bring more than searches by Hash.  It's bringing
another structural dimension to Gnutella.

Sure, at first the DHT will just be a repository of distributed storage
and lookup of things by hash.  But a DHT gives you a topological structure
that can be further exploited.  I'm sure Peng's goal is to get there.

At present, I view the DHT as an evolution step for Gnutella.  It's the
same kind of evolution that occurred when LimeWire proposed the Ultrapeer
and Leaf organization of the former unstructured Gnutella world.  It brought
us the ability to do QRP, push-proxies, OOB query proxying, optimize the
bandwidth usage of leaves.  These evolutions were also helped by other
extensions such as GGEP (another great evolutionary step), and of course
vendor messages.

The first immediate usage of the DHT structure will be getting magnets to
be useful.  That won't be the last usage.  But first let's get an
inter-operable DHT out there whithin Gnutella and use magnets as the showcase
to the world that Gnutella is still capable of evolving.

Raphael