When most people look at human settlement patterns over time, what they see is a history of transportation; but what I see is a parallel to the development of the computer industry.
Settlement patterns reflect transaction costs for information and market exchanges, that's why markets started out walking distances apart and why, today, thousands of formerly busy Canadian and American prairie towns are remembered only by a road side and a lonely graveyard.
When science based computing - as distinct from data processing - first got started in the late nineteen forties and early fifties, the world saw what amounted to an agrarian land rush with thousands of companies and individuals setting up shop in the new field of cybernetics.
So where are they now? Well, just as motorised farm equipment and trucks depopulated the prairies, the microprocessor turned these initial thousands into hundreds.
As today's desktop computing paradigm gets replaced by network computing we're down to four main contenders: embedded PPC, IBM with Linux on its grid-on-a-chip Cell, Sun with Solaris on CMT/SPARC, and Microsoft straddling the fence as it gets set to move from the old one man, one computer x86 idea to its still secret network OS on Xenon.
"Empty towns, busy graveyards" really is a metaphor for economic change in agriculture, but I think it's also a warning to those of us who work in IT. All that next generation stuff coming at us is going to drop those transactions costs a lot closer to zero a lot faster than most of us expect. So what's your Plan B?