NavigationAnnouncementsUpcoming eventsRecent blog posts
|
linkswww.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
<<<<<<<<<
|
SearchAnalysisNewsReviews |