From: Robert J. Bradbury (bradbury@aeiveos.com)
Date: Sat Mar 17 2001 - 00:17:45 MST
On Fri, 16 Mar 2001, Spike Jones wrote:
> Nay, for if we reduce the communications bandwidth, we need not
> go to optical frequencies. These wispy interconnect cables, a few
> dozen microns or so in diameter, can be used to carry simple
> electrical signals at a few thousand baud. With sufficiently high
> impedence, communications between nodes could be carried out
> using very little power, a nearly negligible fraction of that collected
> by the nodes.
So you have "isolated" nodes which handicaps the power of the net
unless you have an extremely rich communications "shorthand"
where a few bits translates into trillions on decompression.
> The question becomes, what is the nature of the communication
> between nodes? Will a few thousand bits per second be sufficient?
> spike
What is the size of the computation you want to do? Are you sim-ing
just Spike Jones' mind (that will fit in a node) or humanity (that
won't [~10^5 "minds"/node]). You may be able to get away with
separating humanity across multiple nodes (must humans communicate
with most other humans at extremely low bandwidths) but you *will not*
get away with that strategy of distributing posthumans across the
nodes. Post-humans *know* that when you say "explain this to me"
it is going to involve a kilo-peta-byte transmission and they aren't
going to want to wait a billion years for it to come across your
slow data links.
You have to assume the information at your disposal is within
a few orders of magnitude of the number of atoms in a node
(perhaps even *more* than the number of atoms in the node if
you can map the information onto electron levels reliably).
If your node weights a kg -- how do you hand > 10^20 bits
of information (a fraction of what you have available)
over to another node for review?
Hmmmmm?
Robert
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 08:06:26 MST