The assertion we made in CI No 3,011 that without vigilant regulation, the computer industry will always tend to a monopoly much more rapidly than any other industry is an uncomfortable one – especially as for most people, the kind of regulation that would be required would be unacceptable. Microsoft Corp and Intel Corp share their monopoly of the desktop and growing incursion into the mid-range and towards the high end with Windows NT on Pentium Pro because of the unique nature of the computer industry – a characteristic so arcane that it is extraordinarily difficult for outsiders like Michael Blumenthal and Louis Gerstner to recognize it, let alone understand it and its implications. Simply put, it is that no other capital good can cost you many millions of dollars upfront yet leave you not knowing for between two and five years thereafter whether it was a good investment or a complete waste of money.

By Tim Palmer

We all know of examples of the latter, computer projects that started out with the highest of hopes, which ultimately had to be abandoned and written off when costs ran out of control and it became clear the project would never deliver its early promise. The first generation of data processing managers quickly learned to be pathologically risk-averse, constantly tending to give their business to the computer builder likely to be around the longest, which, it soon became clear to all, was IBM Corp. Today the average user is much less worried about their supplier dying because they buy their applications off the shelf. The risk – and the risk-aversion has now largely transferred to so-called independent software vendors – which sounds as if it refers to software retailers or distributors, where it is actually used to refer to software developers. It is these people that have largely decided that we shall use Microsoft software on our desktops and are increasingly despairing of the still far from harmonized Unix world and writing for an environment where they know there will be only one standard, with one dictator ruthlessly laying down the law, Windows NT. And since Intel Corp makes so many more of each iteration of its chips than any other manufacturer, very few are interested in writing for Windows NT on anything other than Pentium Pro. If, like Digital Equipment Corp, you desperately need people to want to run Windows NT on your Alpha RISC, you have to write something like the FX!32 emulator and do all in your power to persuade your users that it really will run NT applications written for iAPX-86 just as quickly on your own Alpha RISC as they run on a Pentium Pro.

Quick and dirty

And some will certainly buy your pitch for now, but every year, your marginal costs become ever higher, and every year, a few more of your already ravaged customer base gets cold feet and switches to NT on Pentium Pro servers from Compaq Computer Corp or Dell Computer Corp, and are amazed at how much cheaper they turn out to be in practice. None of this is any compliment to the excellence of Microsoft’s software: the company has made no progress since 1981 when it bought something called Quick & Dirty Operating System and turned it into MS-DOS. It was quick and dirty because although very similar to CP/M, it eschewed the elegant and difficult ways in which CP/M did things for the careless, slapdash and thoughtless. And very little has improved since then: Microsoft software and bloatware are synonymous, and bad software doesn’t become good software just because memory is cheap. But IBM’s championing of MS-DOS, coupled with Microsoft’s astuteness in retaining the right to license it to any other manufacturer that wanted it, ensured that the wisely risk-averse would gravitate towards it in ever-increasing numbers – to the point where there is virtually no mistake that Microsoft can make that will make much difference to its dominance of the desktop – unless a totally new computing paradigm catches on. It could turn out to be the so-called Network Computer, but the amount of wishful thinking behind the hyping of that concept makes its ultimate success very uncertain. So Microsoft and Intel have a monopoly. What harm does that do the customer? Their computing is far cheaper than it would be were the desktop to be shared by half a dozen different chips and incompatible environments. The biggest software developers need to keep only one version of their major applications current instead of half a dozen and so can justify lower prices. And because the vast majority of people are ill-equipped even to conceive of what computers might be capable of doing today if the diversity that is killed stone dead by Microsoft’s monopoly, there is no great sense of loss or of outrage as a result of that monopoly. So what, if anything, should the authorities – in practice, the US Justice Department – do about Microsoft’s monopoly?

Unbundle everything

Should it, for instance, foster a diversity of processors by insisting that successive versions of Windows should be written for a different chip each time? Such interference would be so transparently silly that it would never fly. Should it define the functions that belong to the operating system and insist that Microsoft bundle nothing else? Perhaps, but who is to decide on the definitions? Should it – echoes of IBM settlements two decades ago – insist that Microsoft unbundle everything so that users could buy – and pay for – only the bits of the operating system they wanted? It would be good for developers of things like compression software, but it would make the basic operating software much more expensive. Is that a price worth paying for limiting the ill of Microsoft’s monopoly? Should Microsoft be forced to split into two companies, one for operating software, the other for applications such as Access and Office? It’s seductive, there is no obvious reason why it should raise computing costs much – but how much would it really do to bust open the monopoly? The awkward truth is that Microsoft or its successor companies will continue to enjoy their monopoly until the next paradigm shift in computing, be it the Network Computer, or, more likely, something like optical computing that needs a quite new kind of programming – and rely on the not-invented-in-Redmond syndrome!