www.humancasting.org

Radio is inherently a broadcast medium. The internet is inherently store-and-forward. Proxy Caching Mechanism for Multimedia Playback Streams in the Internet http://www.ircache.net/Cache/Workshop99/Papers/rejaie-html/

Proposed Humancasting network architecture http://humancasting.manilasites.com/pictures/viewer$8

(Cron or any job scheduler) + (Napster or any filesharing tool) + (OPML or any playlist format) + (WinAmp or any mp3 player) + (signed playlists to allow for editorial voice) =

music and text/flash?instead of dj voice

Push is the key

what is SYN/ACK?

http://tipster.weblogs.com/discuss http://www.sourceforge.net/projects/swarmcast/

The Case Against Micropayments http://www.oreillynet.com/pub/a/p2p/2000/12/19/micropayments.html

"Paris Metro Pricing model. http://www.research.att.com/~amo/doc/paris.metro.minimal.txt add text-to-speach processor (with a skin or filter?)

>http://openapplications.org/challenge/index.htm http://www.thetwowayweb.com/soapMeetsRss

grid computing virtualisation http://groups.yahoo.com/group/decentralization/message/1060

http://www.superopendirectory.com/about

Gartmer onP2p http://groups.yahoo.com/group/decentralization/message/1065

http://www.thetwowayweb.com/payloadsForRss Pro Mojo http://www.oreillynet.com/pub/a/p2p/2001/01/11/mojo.html

I'd like to see a system similar to this, but using some system of identifying other peers available that are "close by" (in network terms). The most obvious method of determining "peerness" would be DNS, but we are all aware of the problems of DNS & P2P systems. I'm thinking perhaps of some kind of client based trace-route program, and an algorithm which compares its traces with other peers in an attempt to find reasonably close matches.

>>> As I understand Freenet, it's slightly better than that. Freenet uses something like consistent hashing so if a node doesn't have a document you're looking for, it at least knows which of its neighbors is more likely to have the document. So at each hop a request gets a little closer to the document. >>>

There is still the problem of the Freenet topology itself. A node's neighbors are (AFAIK) purely random and do not reflect the underlying topology. So there might be another Freenet node very close to you on the Internet, but it might be far away on Freenet. Thus Freenet does not necessarily deliver data from the closest node. On Network-Aware Clustering of Web Clients Balachander Krishnamurthy and Jia Wang Proceedings of ACM SIGCOMM 2000, Stockholm, Sweden, o hop count o latency measure o bandwidth probe o physical distance o etc o weight combinations of the above. simple ping (either IP or application level). It contains information regarding hop count, router congestion/load, bandwidth, and end-node load. We (www.vtrails.com) developed a p2p application for streaming (one to many to many) the one being a coordinating server that includes an algorithmic module beining in charge of mapping the ip request and sort them (network wise & connection wise). mojo method http://groups.yahoo.com/group/decentralization/message/1111 swarmcast description http://groups.yahoo.com/group/decentralization/message/1121 serialcast or multicast?? http://www.techrepublic.com/printerfriendly.jhtml?id=r00720010103ggp01.htm http://www.peertal.com/directory/ http://www.frankston.com/public/essays/ContentvsConnectivity.asp clay speaks!! You are not mistaken. We all live in an interated prisoners dilemma, so there are rewards for co-operation that come from the growth in the system as a whole.

This is where I think Mojo Nation is blowing smoke. They've make a big deal about the Xerox Gnutella study, and use it question the intelligence of the user:

"In a similar vein, Napster and other distributed client-servers are built on the shifting sands of volunteerism. Freeloaders and parasites cannot be controlled. The freeloader gains all the benefit of the whole system and pushes the cost to those foolish enough to give away their resources."

As someone who has been foolish enough to give away my resources almost since Napster launched, I can say (along 10's of millions of others) that far from being foolish, this is one of the best software choices I've ever made.

I half-recommend* (or recommend with trepidation) Non-Zero by Robert Wright. The first third of the book notes that life is a daisy chain of non-zero-sum games, and that there are non-zero-sum economic games as well. What Napster understood was that resource allocation could be non-zero, i.e. non-Pareto optimal, if it leveraged unused resources correctly.

mccoy The "Paris Metro Pricing" model is a market-based distributed resource allocation tool to provide what Dr. Odlyzko argued was the least complex mechanism for providing best-effort quality of service when dealing with network congestion. It is not necessary to expose this sort of a system to the user and it can exist simply as an optimization mechanism within the infrastructure, but a tool that provides both distributed load balancing by shifting users towards under-utilized resources and allows for basic QoS (only if it becomes necessary, otherwise everything can run flat-out as fast as possible) can sometimes be a useful thing to have available.

Both BearShare and LimeWire have features new releases that highlighted fairly extensive freeloader protection built-in to the clients. EDonkey 2000 has also implemented a rather clever mechanism for accomplishing something similar to our swarm downloading architecture by letting users who host the same file to answer queries for different byte ranges within that file. This mechanism is not as aggressive about marshalling lots of agents to a specific download tasks but as a passive replication and parallel downloader it seems to be a good idea. more mccoy signpost http://groups.yahoo.com/group/decentralization/message/1207 On NAT http://msdn.microsoft.com/library/default.asp? URL=/library/techart/Nats2-msdn.htm lucas on swarmcast