[Fwd: RE: [decentralization] What is p2p? (long)]

Date view Thread view Subject view Author view

From: Gregory Alan Bolcer (gbolcer@endtech.com)
Date: Wed Nov 29 2000 - 08:15:12 PST

Todd Boyle posted this to decentralization, so I would apologize
for reposting it on FoRK except for the fact that the Nokia Rooftop
project for building Mesh networks is really cool. It seems like a commercial
deployment of the Consume.net project starting out in a variety of cities
(and discussed previously also). [2] The Nokia whitepaper talks about their
Nokia AIR operating systems and the requirements they chose to highlight.
This includes ad hoc node deployment, network adaptation including changes
in link quality and automatic network topology configuration. They use a
spread spectrum with node authorization, although the paper doesn't mention
who and how it takes place. Likewise, with data encryption.

The best part, a call for neighbor management protocols. The also
mention that they allow network syncrhonization through a fully distributed
algorithm that eliminates the need for a master clock.


[1] http://www.nwr.nokia.com/rooftop/
[2] http://www.consume.net/

-------- Original Message --------
Subject: RE: [decentralization] What is p2p? (long)
Date: Wed, 15 Nov 2000 17:23:29 -0800
From: "Todd Boyle" <tboyle@rosehill.net>
Reply-To: decentralization@egroups.com
To: <decentralization@egroups.com>

Thanks, Clay. What a pleasure to read.

> My argument is that p2p is a class of applications that takes
> advantage of resources at the edges of the internet, resources like
> cycles, disk, and human presence.
> The full argument is attached below. (There's an HTML copy at
> http://www.shirky.com/writings/what_is_p2p.html if you prefer reading
> in your browser.) Comments appreciated.

You're definitely on the right track. There is a lot of imprecise
terminology. I'm not always sure when a new "thing" crosses the
line and actually needs a whole new word but this is getting close
to that point.

English continually overloads meanings upon existing nouns and
verbs. At the opposite extreme I guess are the Asian character
sets; obviously the academic or cultural mandarins were in the
habit of creating new pictograms everytime they wanted to insert
a new meme into the cloud.

What I think you've done is identified a variable that exists
along at least 3 dimensions and called it decentralization...

  X ------------------------------------------------------>
           decentralization of physical routes taken by data.
| \
| \
| \ decentralization of locations of data and metadata, etc.
| \
| \
| \
of effective control

Much of the engineering going into the P2P is motivated by desire to
change some of the political and economic outcomes of the existing

You could get everything you want by changing the laws, but, alas
that cannot be done so, we try first, designing various programs that
escape from social control, censorship, regulation of various kinds.
Napster, obviously. Freenet, Gnutella, Publius.

If the objective is truly democratization and decentralization of the
final outcomes, the answer will not emerge until somebody comes up
with both a new client and a new network architecture that actually
decentralizes the physical routes taken by the data as well as the
locations of files.

Completely decentrallized routing can happen in some embryonic ways
with wireless, as seen on SFLan and SeattleWireless. There are
groups in Portland, Canberra and London as well, according to Matt
Westervelt, http://www.seattlewireless.net/ who spoke to the
first meeting of the eastside wireless group Monday night.
http://seattlewireless.net/archive/dev/threads.html They're talking
about running 802.11 cards in ad-hoc or promiscuous mode, and
figuring out a way to route many-to-many. Nokia bought Rooftop
for this purpose http://www.nwr.nokia.com/rooftop/ which creates
true mesh networks on the fly. Is that cool or what!

What you want is something that only connects with the party you want
to communicate with, and nobody else, and which drills laterally
across the edges of the network to accomplish that connection.
Something which does not depend on the DNS system or any other
centrally, politically controlled server or database. You want
something that can use DNS but also works with other namespaces, even
user-defined namespaces and local tinpot dictators, authorities and
registries. Believe me---whatever happens, if it is nice and orderly,
it will be neither secure, nor useful. If you don't see illegal content,
then it cannot possibly be secure for commerce, or private with regard
to censorship, regulation etc.

You also want a network that always encrypts EVERYTHING it carries and
upon which it is impossible to send anything unencrypted. You want a
system in which every storage is always encrypted. Let the intruder
expend huge resources sorting thru the vast, worthless, desert of low
value content.

Some lessons from the last 5 years: nibbling around the edges, building
a little "decentrallization" isn't going to work. Encryption has been here
through the internet's growth, giving the author nearly complete control
over many of their own privacy issues but hasn't been adopted. You may
think, that's a chicken and egg problem? I don't. I think the problem is
that all the fsking wires go out of my house, down to the telco and into
government controlled routes from end to end. And the software is
Windows, designed from the ground up to be controlled from the server.

I'm furious about stunts like Carnivore which illustrate clearly, how all
of government as well as any party in any carrier or ISP or server along my
route can have actual control of data. Let's skip the rest of this smarmy
novel we're reading and go to the last page, in which

 - the physical routes thru the network will be decentrallized,
 - all of the data will be totally encrypted, and
 - the client will be radically hardened and private.

* Todd F. Boyle CPA http://www.GLDialtone.com/
* tboyle@rosehill.net Kirkland WA (425) 827-3107
* XML accounting, webledgers, BSPs, ASPs, whatever it takes

To unsubscribe from this group, send an email to:

Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Wed Nov 29 2000 - 08:22:36 PST