[p2p-hackers] Where do bright minds discuss p2p technology?

[Emin Gun Sirer]

Hi Jacob,

I am not sure what the Britney-problem is exactly (rather, I think the
Britney problem is the sheer crappiness of her music, but you seem to be
referring to something far more technical and possible fixable), but 
three things struck me about your note:

  - Since you are interested in caching, take a look at recent work on
    how caching can be used to improve the performance of DHTs. Coen
    and Shenker have a nice paper on how sqrt(N) caching in 
    unstructured p2p systems can improve performance. Our work on
    Beehive (which is the engine behind CoDoNS, CobWeb, and Corona)
    showed how O(1) lookup times could be achieved on top of O(log N)
    DHTs with very modest replication costs.

  - Since Britney is involved in the problem definition, it's a safe
    bet that lawyers/RIAA/MPAA will have to be involved as well. You
    might want to look at recent work on anonymous communication
    systems, especially source rewriting systems (Tor, onions, crowds,
    mixes, ap3b, tarzan, etc), dining cryptographer networks
    (Herbivore, Chaum's DC-nets), and constant broadcast systems
    (P5). Anonymity is hard to combine with high performance and
    scalability, but there has been lots of interesting work in the
    last few years. 

  - I encourage you, as well as everyone else on this list who is
    interested in designing, building and deploying large scale
    distributed systems (of any kind, not just peer-to-peer), and is
    interested in pursuing graduate studies, to apply to Cornell. Do
    drop me a note if you are thinking of applying or have applied.

There is quite a bit of systems building activity here and this is an
exciting time to be working on P2P. Existing infrastructure services
are so broken in so many ways that there is a huge opportunity for
impact.

- egs

Jacob Madsen writes:
>
>As it is now, i'm not working on p2p stuff, but next semester (after new year) 
>it will change.
>I'm about to start on a "final project" and I have chosen to work with 
>structured DHTs.
>
>I'm not done defining the project yet, but here is what i have in mind:
>1. Implement several known and structured DHTs (like Chord, Kademlia, ...) 
>with a common interface and compare them by running some tests.
>2. Solve the "britney problem" either by using caching, a distributed
>sloppy hash table or some other method I'm not aware of yet.
>3. Make "nodeId theft"/spoofing impossible. Right now I'm not sure how I will  
>do it.
>4. Implement a simple filesystem where all nodes have a synchronized index of 
>the directory and file structure. Again, I'm not sure how I will do it.
>
>If someone got any suggestions or papers about this, then I'll really 
>appreciate to hear about it/them.
>
>/Jacob