Munchnets: bioscale networking

Rohit Khare Rohit@KnowNow.com
Wed, 17 Oct 2001 14:54:09 -0700


[excerpted from a discussion with a colleague about possible DARPA 
research directions. I haven't expounded on my dream of munchkins 
recently, so this was a brief snapshot of what I think the problem 
statement might be -- never mind solutions! :-) -- Rohit]

==========================================
Create a prototype biological-scale network.

To reach beyond Internet-scale, even Interplanetary IP (IPN) scale, 
we need to reconsider how to build computing systems with absolutely 
no measure of central control. How can coordination algorithms 
practically scale when in the absence of any organizational arbiter 
(e.g. without assumptions like unique Ethernet MAC addresses, IP 
addresses, domain name control, or even user name control)? Even 
where arbitration is necessary, physical means should be considered, 
such as assuming that GPS-aligned clocks are globally available, for 
example.

The goal is to build a one-pound device containing a battery, 
low-power, load-adaptive central processing (aka async integer 
cores), and parasitic communications interfaces that will try to take 
in any and all connections available in the immediate environment, 
from Gbps to bps (ethernet, bluetooth, irda, UWB, etc) and any local 
storage (flash, microdrive, optical ROM caches of major public Web 
sites) to provide reliable peer-to-peer multimedia messaging.

A battlefield dataphone that can work anywhere, anytime, securely, 
and efficiently.

In some sense, this is within reach of conventional technology, it 
would seem: iPAQs with Wi-Fi and microdrives running ESMTP... but 
there are deep, abiding assumptions of central control littered 
throughout such a system, all the way down to the radio arbitration 
in 802.11b, I'm sure.

There will also have to be foundational economic research to 
establish rules for the spontaneous emergence of order -- these 
devices will reinvent systems of banking, currency exchange, and 
speculation in the markets for bandwidth and storage.

The guiding patterns for organizing this next level of computing will 
have to be biological: adopting the point of view of a zygote 
attempting to establish meaningful relationships with the existing 6B 
humans.

This kind of breakthroughs posited here are essential for 
"command-and-control" of MEMS systems and other future, agile, 
military and civilian sensor platforms, from automobiles to 
free-flying airplanes. Without shattering the paradigm of today's 
consciously-controlled and fragile Internet, we will not be able to 
leverage the coming universe of trillions of devices across, on this 
planet and on other bodies.

Strictly speaking, these technologies are below the level of the 
Semantic Web, yet will rely extensively on cutting-edge work in 
knowledge management and knowledge representation emerging out of Web 
Services, XML Metadata, and related fields.

In the immediate term, this is surely of military interest in the 
continuing struggle to maintain communication integration among 
diverse platforms, services, and nations in today's archetypal 
multinational force deployments. The stories of friction arising from 
incompatible data transports -- much less unified models for battle 
awareness (mapping, materiel identification, IFF, etc) are legion.

There is also an immediate dual-use need from homeland defence 
scenarios, since there is no prototype yet for a crisis-management 
network that can be deployed within minutes, and expanded 
incrementally and continually, while enforcing proper separation 
between the multiple agencies sure to be overlapping in the field of 
any such crisis.

So there you have it: the most audacious goal in computing technology 
I can articulate: networking without administrators.