Metcalfe's column: From the Ether

Computational fabrics are a promising possibility, or so say scientists working with DARPA, our Defense Advanced Research Projects Agency (http://www.darpa.mil). Of course, they said the same thing 30 years ago about packet switching.

My introduction to computational fabrics began at the American Textile History Museum (http://www.athm.org), in Lowell, Massachusetts. I went there because I live on Kelmscott Farm (http://www.kelmscott.org), which conserves rare livestock, including historic breeds of sheep.

Sheep produce wool. And wool is among the fibres shorn, carded, spun, knitted, woven, dyed, and printed into fabrics on the ancient textile machines in Lowell.

DARPA's Information Science and Technology (ISAT) Study Group (atlas.ida.org:8500) met that week in Woods Hole, Massachusetts. There I got to join brainstorming on self-configuring sensor networks, immersion techniques for enhancing human memory, mobile information assurance, and, coincidentally, computational fabrics.

The net's web is said to be woven, and the earliest computers controlled Jacquard looms, but neither of these is a computational fabric. DARPA is brainstorming about using textile technologies to weave computers, or what I call textile-area networks (TANs) of nanocomputers, or smart materials.

Leading the brainstorms were Elana Ethridge from Systems Planning Corporation, Anant Agarwal and Tom Knight from MIT, Abdelfattah Seyam from North Carolina State University, Philip Kuekes from Hewlett-Packard Laboratories, Dick Urban from DARPA, and Robert Parker from the University of Southern California.

Today, computing devices are built using rigid laminates -- chips and circuit boards -- which are patterned, etched, and drilled. But what if computing devices were instead woven -- warp and weft -- from spooled high-tech fibres, filaments, threads, and yarns? Why not sprinkle nanocomputers -- smart sand -- on substrates of woven power filaments and communication fibres? Should they be glued or soldered like sequins or stuffed in knitted pockets or hooked like velcro? Or should computational fabrics be more than substrates? Should they be transistors and their connections woven from doped yarns?

Imagine a yarn conducting on one side and insulating on the other. Woven one way, the conducting side makes contact, but the other way, not. Design software would program looms to lay out huge arrays of woven fabric devices, much as laminated chip devices are photolithographed today.

Should fibres connect computing elements point-to-point, or should large conducting surfaces be woven, through which cellular Ethernets might broadcast TAN packets?

Maybe two is the ideal number of interconnects per nanocomputer, with woven voltage and ground threads carrying power, clock, and data together. Maybe fabric nodes would include fibre batteries or even solar power cells. Maybe some fabrics will be wet, incorporating biological components to which nourishing fluids will be carried through hollow fibres in a woven tissue.

TANs are likely to shine in applications where large areas need covering, with conformability and perhaps even unfoldability. For example, imagine large woven displays, electronic paper, speakers, phased-array antennae, solar collectors, sensor arrays, mirrors, health-monitoring suits, and of course DARPA stealth shields.

When actuators can be sprinkled, how about steerable airfoils? How about satellite ground stations that drop from the air and unfold in computed camouflage?

Hey, how about an inflatable PC powered by nanoturbines spun by escaping air, as from a bagpipe?

Now, instead of today's laminated programming paradigm -- design, build, compile, and run -- there would be a defect-tolerant textile programming paradigm -- build, measure, design, compile, and run. Redundant filaments and routing would bypass fabric defects.

Where DARPA would not want a bullet taking out a whole fabric, TAN packets would be routed around bullet holes, which brings us back to packet switching.

Now, if this sounds far-fetched, imagine suggesting in 1969 that packet switching would lead to Napster.

Bob Metcalfe seeks your help in getting his new paperback into its third printing. Go to http://www.bookarea.com and buy Internet Collapses and Other InfoWorld Punditry.

Join the newsletter!

Error: Please check your email address.

More about Defense Advanced Research Projects AgencyHewlett-Packard AustraliaIDAImmersionMetcalfeMITNapster

Show Comments

Market Place