From: Jeffrey Kay (jkay@ENGENIA.COM)
Date: Fri Dec 22 2000 - 08:07:30 PST
Great perspective and I think this really highlights the main problem with
P2P as an "industry". Basically, you could argue that the Internet is the
same as Napster in this regard because DNS effectively acts as the
centralized mapping service between systems much in the same way that
Napster acts as the directory service for MP3 files. So I agree with you in
What we're really befuddled about in P2P is the definition. What really
qualifies? Groove uses a central rendezvous point also -- do they qualify
as P2P on the basis of their virtual server implementation? Engenia allows
every node (desktop or otherwise) to serve objects to other nodes, with an
installation defined object discovery system (could be centralized, could be
distributed) -- do we qualify? AOL Instant Messenger doesn't connect nodes
directly at the transport layer, but uses a messaging switch (essentially)
to pass messages directly from user to user -- would that qualify? Gnutella
uses a friends of friends architecture to do object discovery -- would that
qualify? Magi uses rendezvous points also -- does it qualify? E-mail?
Greg Bolcer defines Peer-to-Peer in his 6 December paper as "any
relationship in which multiple, autonomous hosts interact as equals.
Peer-to-peer communication refers to these real or virtual connections
between corresponding systems." It's a pretty broad definition that doesn't
reference intermediaries or protocol stack position. As a result, this
definition would likely cover every example that I listed above. I'm not
sure that this definition is strong enough to really define an industry and
that's what P2P really needs at the moment -- a good definition to separate
"thems that are and thems that ain't".
BTW, just in case everyone didn't see the e-mail, it looks like the next
Intel Peer-to-Peer Working Group Meeting is as follows:
Peer-to-Peer Working Group Demo Showcase and Meeting
February 7 - 8, 2001
Santa Clara Marriott
The Peer-to-Peer Working Group will hold its second meeting February 7 - 8,
2001. The meeting will include a demo showcase, a general session and a
members meeting. The proposed agenda follows:
Wednesday, February 7, 2001
0700-1300 Demo Showcase Setup
1300-200 Demo Showcase Open
Thursday, February 8, 2001
0800-0900 Meeting Registration
0900-1100 General Session
1200-1300 Members Luncheon
1300-1700 Members Meeting
-- Happy Holidays, peer-to-peer, of course --
Jeffrey Kay <firstname.lastname@example.org>
Chief Technology Officer, Engenia Software Inc.
"First get your facts, then you can distort them at your leisure" -- Mark
"Golf is an endless series of tragedies obscured by the occasional miracle"
-- Sports Illustrated
"If A equals success, then the formula is A equals X plus Y plus Z. X is
work. Y is play. Z is keep your mouth shut." -- Albert Einstein
From: Strata Rose Chalup [mailto:email@example.com]
Sent: Thursday, December 21, 2000 9:05 PM
To: Lucas Gonze
Subject: Re: ARTICLE: Despite the buzz, peer-to-peer startups lack
Oh, come ON!!!
Napster isn't a real P2P solution, Napster is a bunch of independent,
not able to share data between them servers that are "centralized"
through a round robin MX or layer 3/4 load balancing or equivalent.
Napster *is* a centralized server with a fat pipe, it's just a pile of
them instead of one.
Client lookup to server, followed by client to client contact is a
centralized solution relying on a distributed data model. Napster she
go poof, Nappieboys canna talk to each other unless somebody put up new
A truly P2P service would rely on dynamic discovery of root-server
equivalents, kinda like Anubis
(http://www.virtual.net/Projects/Anubis/), something I wrote up in 1994
and presented at Defcon 2 (my first and only) but never bothered to
publish 'cause it was so damn obvious. You can put all kinds of layers
of indirection in there to minimize data obligations (or maximize them),
use out of band pre-negotiated channels to exchange root server
information, decide you have transfinite resources and do a subscriber
multicast ring for root lookups, etc etc etc.
All P2P is just a mapping service between arbitrary-string-A and
network-reachable-address-B, with however many layers of indirection you
want to put in there. That's all. No magic. Everybody can be a client
and everybody can be a server. Just like the OLD net, the one they
called ARPA, before people got used to thinking of a bunch of glorified
terminals with 8086 chips in them as "computers" instead of "dumb nodes"
and thus confused everybody else into forgetting what "host" and
C'mon you guys, you know this stuff. As St. Dogbert says, "Don't make
me come over there..."
Lucas Gonze wrote:
> The scalability of Napster doesn't just apply to disk and pipe, it also
> to scale as a tastemaker. At Walmart, the need for economies of scale
forces them to
> provide one-size-fits-all choices. As Napster grows it supports smaller
> How could something as timely, stupid and irresponsible as those fake
> cuts ever show up on a centralized server? The degree to which Napster is
> to their users would be nearly impossible to reach with a centralized
> is like the Santa Claus problem. Santa really doesn't have time to find
out that I
> like orange mittens but not red, but if each kid has a parent to track
their taste it
> is completely practical.
> - Lucas
> > You're saying Napster could have been done by anyone with a big server
> > farm and a fat pipe? Isn't that true of most P2P projects? Couldn't
> > be run on a big Beowulf cluster at NASA? I think the point of Napster
> > that Shawn didn't have a server farm, and didn't have a fat pipe, and
> > P2P allowed the users to get around that resource limitation.
> > I don't know whether, historically, Shawn thought of the resources first
> > or the (alleged) copyright circumvention first, but even if there were
> > copyright issues, P2P file sharing let them solve a problem through
> > when they couldn't afford a centralized solution.
> > -Matt Jensen
> > NewsBlip.com
> > Seattle
-- ======================================================================== Strata Rose Chalup [KF6NBZ] firstname.lastname@example.org VirtualNet Consulting http://www.virtual.net/ ** ISP/ASP Systems Integration, Architecture, & Project Management ** =========================================================================
This archive was generated by hypermail 2b29 : Fri Dec 22 2000 - 08:13:12 PST