Supposing instead we deployed mosquitoes rather than space shuttles: there are so many of them, and each uses so few resources, that it's not important that any particular individual survives. Thus we can do away with most of the special handling, and the (wireless) communication only needs to be good enough to contact a few other mosquitoes in the immediate neighborhood. In brief, rather than insulating our processing from the environment, let's make it a living part.
Like mosquitoes, our devices will be responsible for feeding themselves, whether by harvesting RF fields, light, chemical reactions with the environment, or (like the Atmos clock) small changes in air pressure or temperature. Because the ``food'' supply might not be constant, and because of the inevitable accidents that will permanently put processors out of service, we face the challenge of creating a self-organizing system that acts reliably at a global level even though made out of elements that are inherently much more unreliable and haphazardly connected than those with which we're currently accustomed to work. The manner in which data, instructions, and messages will diffuse throughout such a system, and the automatic provision of redundancy, will probably correspond more to metaphors drawn from biology, physics, and chemistry than to current computer engineering practice.
Doctoral dissertation research in the Object-Based Media Group at the MIT Media Laboratory by William Butera seeks to make processors part of the built environment, dispersing them throughout building materials like wallboard or laminated tabletops, or even suspending them in a viscous coating that can be painted onto surfaces. Why would we do this? Not only does this allow very simple personal devices to take advantage of almost unlimited amounts of environmental computation and storage, but the connectivity and memory in such a system is itself computational: the path from point A to point B can pass through as much processing as required to compress/decompress/reformat data, index or analyze it, relate it to other information, and perform other needed services. Agents of various sorts can reside in the environment, watching information as it goes by, or wandering about looking for it. Membership in this ecology is not limited to the permanently installed devices; indeed, any nearby object can (if it incorporates one of the mosquitoes) communicate its presence and knowledge, and become a full member of the ecology.
Some more interesting opportunities emerge if we create an ecosystem of more than one ``species'' of processor. By mixing in a sprinkling of infrared sensors we can add fire/surveillance functionality; by adding some small proportion of processors with wider-radius communication capabilities (at the cost of more ``food'') we can improve routing of long-range messages. Pressure sensors and bistable electronic ink capsules can be added to turn surfaces into touch-sensitive displays. By virtue of the way in which they are deployed, environmental computational elements can gain a huge degree of context awareness that can be employed usefully in communications, entertainment, or personal information management.
Rather clever for a bunch of mosquitoes, no?
V. Michael Bove, Jr. is head of the Object-Based Media Group of the MIT Media Laboratory in Cambridge, Massachusetts, and a founder of WatchPoint Media, Inc.