Re: RANT: wiring up the neighbourhood

From: Eugene Leitl (Eugene.Leitl@lrz.uni-muenchen.de)
Date: Tue Jul 15 1997 - 23:17:48 MDT


On Sun, 13 Jul 1997, Mark Grant wrote:

> [...]
> > routing and switching. Boxes, knowing their spatial address
> > (from GPS, and/ultrasonic, or vis LPS), can select the nearest
> > one.
>
> I don't see the advantage of this; why should the boxes know my spatial
> location rather than where I'm connected to? Otherwise when I move I'll

This is one step above the physical transport layer. Upon it, you can
mount a DNS-like abstraction. However, I simply don't like global
distributed databases. Machines don't need syntactic sugar, anyway.

> have to update my address all the time. It would presumably work OK for
> fixed sites, but be really bad for mobile users.

A solution: you leave a message stream reflector in your home (sweet
home) node, you leave a trail of forwarding agents as you travel. When
two agents discover they communicate via a relais, they snap the pointer,
and start communicating directly. When you move, you report your new cell
address to the last agent. _Very_ limited flooding (just few bits of the
binary ID) should catch you even at Mach 5.

If you move permanently, you give notification to all relevant parties,
so that the bulk gets redirected to your new location. A bounce agent
left in place will forward you the rest, while also notifying the senders
about your address change.

There's a million of ways to make this practicable. The chiefest reason
is that you don't need routing anymore, switches (which are vastly simpler
and dumber) are sufficient. So let's chuck out TCP/IP.

> Mark



This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 14:44:35 MST