Ants in your P2Pants

Tim Finin writes: "Anthill is a framework being developed at University of Bologna to support the design, implementation and evaluation of P2P applications, viewing them as instances of Complex Adaptive Systems, typically found in biological and social sciences. In Anthill, desired properties such as resilience, adaptation and self-organization correspond to the "emergent behavior" of the underlying CA system. An Anthill system consists of a dynamic network of peer nodes; societies of adaptive agents (ants) travel through this network, interacting with nodes and cooperating with other agents in order to solve complex problems. The source code for Anthill v1.0 is available for downloading. MORE on this is at ebiquity.org."

Comments

You know

[autopr0n]

Those words do have meaning. And while they can be used for empty hype, there existance alone is not an indicator of vapiedness.

And besides how can you consider "implementation" and "complex" buzzwords but not 'dynamic', 'peer nodes' or 'downloading'?

A few other agent frameworks

[Biff]

Distributed Agent systems aren't anything new, they have been a hot topic in the academic world for quite a while now.

The BT Zeus project is an open source agent framework that includes lots of nice stuff (visual development / visualisation / intellegent agents etc.) See www.btexact.com/projects/agents/zeus/ for more info. I feel obliged to mention SoFAR an academic focusing on DIM (Distributed Information Managment) Agents (or at least they where last time I was there).

What he's talking about

[Alsee]

Akk, that post did NOT explain what you're talking about. For the benefit of the confused I'm pretty sure he's trying to come up with a P2P network without using a central address to find the network. Looks like he wants to use a CODE-RED / NIMDA method and send "hello?" packets to random IP addresses at a rate of 1 per second looking for anyone else that happens to be running the P2P.

Randomly spamming the internet with packets is a very questionable plan. You WILL get complaints. That aside, if you *do* go forward with the random address search I'd suggest you start and several addresses per second with an exponentially decaying search speed (Multiply the delay by 1.01 for each address you test). It's a much more efficient systems. Early random tests are more valuable than later ones. If there's a significant network you'll find it faster, and if there isn't then the amount of spamming is self limiting.

In addition, his math is all wrong. First - "0.000118... probability of colliding per attempt" would requre a half million friendly P2P servers. Second he's assuming 255^4 IP addresses. He should be assuming 254^4 - and then checking approriate RFC for rurther reductions. Third - the formula for % per minute is wrong - maybe he calculated it right and wriote it wrong, because the number he got looks about right (but "right" is still "wrong" because is assumes a wrong % per test).

And a final observation - the time to "first circut" would also happen to be the time to spam every existing IP address at least once (and more likely several times).

-