[p2p-hackers] In search of the Darknet....

[Duncan B. Cragg]

>> And there is definitely some confusion in the use of 'small world
>> network': P2P researchers use it one way (for optimising the topology),
>> F2F-ers another (apparently, then, as a synonym for F2F) - this may be
>> the source of the apparent mix-up in the 'Freenet 2' slides.
> 
> As far as I know, both camps are referring to the same thing. A network 
> is called a small world if it has a small diameter relative to its size 
> (like a random network) and a high degree of clustering (unlike a random 
> network). Social networks seem to have these characteristics, so we 
> might expect F2F networks to have them too.
> 
> The routing algorithm in Freenet 0.7 is based on Kleinberg's much more 
> specific definition of a small world [1] - a regular lattice with a few 
> additional random links that follow a certain probability distribution - 
> but there are other kinds of network with small world characteristics, 
> eg scale-free networks [2].
> 
We could say an F2F network arises from meatspace connections being
replicated in the P2P network.  So an F2F network exists /a priori/ and
can't be jiggled around much for the sake of routing, whereas our P2P
algorithms can do what they like to reap the small-diameter benefits of
small-world networks, perhaps using coderman-style implicit hints, from
the social activity visible, to help.

The point, though, was what Freenet is up to, and here we seem to have
both: we definitely have an overlay network topology optimisation (nodes
around the ring swapping position), but we also seem to have an F2F
aspect, with a reference to out-of-band friendly connections. It may
/have/ both, but I'm struggling to see the connection, as the slides
seem to jump from F2F to Kleinberg... It doesn't explain what 'long'
or 'short' connections mean in the F2F context.

Please help by explaining, as I think I'm handicapped by only seeing
slides and not having been at Defcon! Or perhaps I'm just dim... =0(


Duncan