Something to learn from Minis,


I'm pretty negative on the more highly romanticized aspects of the mini revolution. The thesis that a network of minis is more powerful than a clump of ditto, and that the latter is more powerful in turn than a standard 370 or 1106 costing the same, is thermodynamically unsound. There is, however, one profoundly encouraging thing about minicomputers and microprocessors: because of the limited support the manufacturers could originally offer, and because of the amazing diversity of user applications - especially in dedicated mode most programming was done in machine language or with the help of very primitive assemblers and debugging aids.

The programmers had to get along without Fortran or Cobol, without an operating system, without foolishness like virtual memory. The result was, and is, remarkably efficient and fast-running applications software, contrasted with the murky, OS/VS/PL1-encumbered software written by 370 victims.

But now the suppliers are getting rich; DEC is already very big and very prosperous, and others are in train. So what happens? Word size, goes up, to 32 bits in some cases. With big production runs, memory cost comes down and memory sizes increase, The manufacturers now offer Fortran and Cobol, and threaten their innocent customers with operating systems. Nonstandard, of course!

Doesn't 'have to happen, people! We can go on using the new machines the way we do now. And we can use the 370s and the Univacs and Honeywells the same way: we can learn how to operate big machines efficiently by watching the mini users. The greatest benefit of the whole game could be not the intrinsic value of built-in minicomputers and one-chip microwonders, but the way that they demonstrate to bemused data processing managers how much work they should be getting out of their big machines. It isn't the sophisticated peripherals that are slowing the Major installations down; it's the layers and layers of unnecessary systems software!

The mini girls and boys have shown the old-timers what can be done. Now, let's do it!

Dr. Herbert R. J. Grosch ist Editorial Director der Computerworld