Smart Machines

Sun Microsystems has recently put software on the Web that any of us can tap into just as if we were running it on our own computers. That has a lot of companies wondering, Which makes the most sense -- concentrating computing power in desktops, in corporate servers, or, as Sun might urge, out on the Web?

I found some unexpected perspective on the question at my mechanic's. One day, on my way out of the shop, I passed an obviously ancient and festively painted engine flanked by two flywheels the size of car tires and festooned with fittings of mysterious purpose. I asked the shop owner, Dan Socci, about it. He explained that the cast-iron machine, which took up about as much space as a dishwasher, was a single-cylinder gas engine of typical design, circa 1910. He offered to start it up, which involved turning several brass knobs and giving the flywheels an energetic yank. Once the engine caught, it seemed to suffer from a form of mechanical narcolepsy in that the explosions that drove the exposed piston were separated by periods of relative silence. Socci assured me that this was how the machine was designed to run.

He then gave me a little background. In the early 1900s such engines, which in larger versions put out as much as 40 horsepower, were staples on farms and in small factories. As he explained more, a thought occurred to me: the evolution of the way in which American factories distributed mechanical power to individual workers in the 40 years bracketing the turn of the century seemed to neatly parallel the evolution of the distribution of computing power over the past 40 years. Maybe there was a lesson there. After further checking with Rod Groenewold of the Antique Gas & Steam Engine Museum, in Vista, Calif., I pieced this together:

Before reliable steam engines became available, in the latter half of the 19th century, rivers were the main power source for factories, mills, dairies, and the like. Think of the setup as the industrial age's version of "time sharing" -- the term coined to describe the approach to distributing computing power in the late 1950s and early 1960s. Forty years ago, most companies couldn't afford their own mainframe computers, so, using telephone lines, they tapped into the mainframes of specialized providers such as IBM, much as the mills lined up along a river took their power from a common source.

The advent of relatively affordable steam engines -- and then of even cheaper, better gas engines like Socci's -- did for factories at the turn of the century what minicomputers did for companies in the late '60s and early '70s. For the first time, businesses could afford to produce their own mechanical (or processing) power by acquiring a single engine (or minicomputer) and letting workers tap into the power via a shaft that ran the length of the factory and was affixed by belts to individual machines (or via an officewide network that was affixed to "dumb" terminals). Eventually, engines became small enough to fit inside the machines and cheap enough for businesses to buy in numbers -- ushering in the PC era of mechanical power.

But at just about the same time, in 1920, a government-subsidized network was becoming available -- the electric-power grid. Why should factories pay for engines at all when they could have power delivered at low cost to their buildings and then funneled to machines equipped with ultracheap electrical motors? If computing continues to follow the same model, then the days of placing powerful computers on the desks of individual workers may indeed be coming to a close. Why should businesses pay for local processing power when they can have it delivered at low cost to their buildings via the Net and then funneled to ultracheap devices -- call them "Web motors" -- that do little but process on-line information?

Then again, factory owners never had to worry about hackers.

David H. Freedman is a contributor to Inc.