[bittorrent] DHT - Clarifying the spec

Tien Tuan Anh Dinh t.t.dinh at cs.bham.ac.uk
Sun Mar 4 16:33:57 EST 2007


  Isn't only Bittorrent client considered as mainline ? I dont think 
many of us here using that one.
  Anyways, Azureus had already followed the footstep and implemented 
Kademlia.

I believe in a generic DHT infrastructure that supports both Kademlia, 
Pastry as well as others, and that would make easier to adopt any 
algorithm we wish.

Alan McGovern wrote:
> The reason why is that that's what was chosen by mainline, so in order 
> to be compatible with mainline we have to use kademlia. No other 
> reason really.
>
> Alan.
>
> On 3/3/07, * Tien Tuan Anh Dinh* <t.t.dinh at cs.bham.ac.uk 
> <mailto:t.t.dinh at cs.bham.ac.uk>> wrote:
>
>     Adrian Ulrich wrote:
>     >
>     > Checkout the 2nd Kademlia paper:
>     >  http://www.scs.stanford.edu/~dm/home/papers/maymounkov:kademlia.ps.gz
>     <http://www.scs.stanford.edu/%7Edm/home/papers/maymounkov:kademlia.ps.gz>
>     >
>     >
>     Sorry if these have nothing to clarify the spec, but i was wondering:
>     WHY using Kademlia for Bittorrent dht ?
>
>     1. This is one of the structured P2P systems with the least number of
>     papers/analysis. Moreover, i haven't stumbled to any formal/detailed
>     evaluation of this protocol yet.
>
>     2. Pastry offered the same functionality as it exposes DHT interface.
>     Plus, Pastry has been studied in detail and its routing has very low
>     latency: median of 1.7 compared to the optimal IP latency. I believed
>     Pastry is one of the best DHT systems out there, and it is definitely
>     not close-sourced. Tapestry is also good.
>
>
>     _______________________________________________
>     BitTorrent mailing list
>     BitTorrent at lists.ibiblio.org <mailto:BitTorrent at lists.ibiblio.org>
>     http://lists.ibiblio.org/mailman/listinfo/bittorrent
>
>




More information about the BitTorrent mailing list