From: Eugene Leitl (Eugene.Leitl@lrz.uni-muenchen.de)
Date: Mon Jul 09 2001 - 03:42:42 MDT
On Fri, 6 Jul 2001, Brian D Williams wrote:
> I'm still reading and evaluating, but it's pretty clear that to build
> anything of any size you need hardline connections. How big it can get
> before routing updates eat the bandwidth is the question, but still
For years I keep harping about how currently used routing techniques are
so very b0rken. Cellular wireless by its very nature is ideal for geodetic
routing, which uses local-visibility routing tables strictly. Meaning, all
the admin traffic is purely local, among immediate, and
neighbours'-neighbour nodes. Meaning, the thing is infinitely scalable, as
bits are sufficiently cheap to be able to address the entire visible
universe at Planck scale, if you think you need to do that. Mapping the
entire solar system down to cubic micron is a piece of cake.
As a boon, if you use geographic coordinates for node addresses (obtained
from connectivity constraints or plain old GPS), you can't get address
collisions (no two nodes can occupy the same place at the same time), and
you can use switching instead of routing, if you design the packet header
layout right.
> pretty darn interesting....
>
> The big problem with ethernet is the distance limitation.
High-quality links are local by nature. This doesn't translate into
latency, if you use (eventually purely photonical) cut-through switching
as described above, using store-and-forward as fallback.
-- Eugen* Leitl leitl
______________________________________________________________
ICBMTO : N48 10'07'' E011 33'53'' http://www.lrz.de/~ui22204
57F9CFD3: ED90 0433 EB74 E4A9 537F CFF5 86E7 629B 57F9 CFD3
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 08:08:35 MST