Microsoft is a Y2K problem. Here's why.

Some people have speculated that Microsoft may finally 'get it right' with Windows 2000, which would have dire consequences for the Linux phenomenon. In other words, if Windows 2000 turns out to be a reliable high quality operating system (unlike its predecessors Windows 3.1, Windows 95, and Windows 98), then Linux will fade in importance, since it will no longer have weak prey.

These people are operating under a fallacious assumption. In this writers opinion, Microsoft does not know how to get it right, so never will. It is a question of culture -- the myriad intertwined set of human activities behind any technological or social movement.

I'm 45 years and have been programming for thirty of those years. I have been programming professionally using Microsoft products for the last nine years (since Windows 3.0) and have programmed on UNIX more or less continuously since 1976, when UNIX was entering its seventh year of evolution. I think Linux is the archtype quality UNIX of our times, but my comments apply equally to FreeBSD and to the culture of UNIX in general.

The main difference I see between Linux and Windows is that with Windows, one is always having to deal with poorly specified APIs (Application Programming Interfaces), poorly written documentation that mixes marketing hype with fact and which is always always incomplete.

Open a page to technical documentation in Windows, and you will see statements like: "When you open a form bound to a record set, the fields on the form will display the values in the current record." OK, fine, what if the record set is empty and there is no current record? (I.e., no antecedent for the anaphoric reference 'the current record'.) Well, good luck. The documentation was mute on the subject. This example is a distilled stylized example from a case I experienced using Microsoft Access. It is typical, not rare. Always, without exception with Microsoft specs, one is forced to play a guessing game and do what could be dubbed 'programming by experimentation'. True, with any system, some degree of learning is accomplished by guessing and experimenting. And any specification ever written always contains some fine elements which are subject to interpretation, i.e., guessing. But with Microsoft products, one has to guess all the time.

When I first broke open the shrink wrap to Visual Basic 5, I was not prepared for the fact that the language had been modified, was not backward compatible with VB4, and that once I recompiled my current VB4 projects under VB5, the system would convert my code to the new language and I would be unable to run my existing code in VB4! But that actually happened. There is no published specification of the Visual Basic Language, as far as I can tell.

Recently I needed to apply the Microsoft Index Server for a client. My workstation is NT server, so I clicked on the Index Server Help menu to remind myself about what the Index Server does. Nothing happened. Click. Nothing. (Or maybe it complained that some help file was missing, I don't remember.) I was not suprised. My NT system was, in Micorsoft time, rather ripe, having been installed about one year earlier. And I am used to things not working in Windows, aren't you? So I was facing the standard 'do I reinstall from scratch or do I upgrade' dilemma that all Windows users experience every year or so. I chose to upgrade, from NT Service Pack 3 to Service Pack 4. I went out to the Microsoft Web site and tried to decipher the marketing hype in order to determine just what the Index Server consists of and how to upgrade it. It was a like a chinese puzzle trying to figure out which option pack and/or service pack to install first. I will spare you the details of my ensuing pain, but a couple iterations of 80MB downloads and four or five iterations through the installation script later, the process actually converged. I had my Index Server back. The user interface for web server administration was completely different than before and Index Server itself had (and has) obnoxious bugs (c.f. dejanews, search on ["Index Server" problem]). But the help menu did now work. A day gone by.

It is a question of feel. What the system 'feels like' when you try to perform system administration, when you want to program it, even when you want to use it.

Of course, it is not just the documentation. Basic design mistakes abound. Let me give you a simple example which has been with us since DOS. And note that even the newest Microsoft operating system, Windows NT, ('New' Technology -- gag), has a DOS command prompt window. On a DOS command line, when you type in a command and then back up with the left arrow and then start typing to make your correction, you overwrite your previous characters, clobbering characters which were previously there. With UNIX, you insert new characters at the insertion point, and the characters which were there are shifted to the right. The DOS decision is wrong because typically when you make a typo you wish to go back and replace N characters by M different characters where M is not equal to N. Hence, you want to go back, delete the mistake and enter the new characters. With UNIX you do that. With DOS, you have to always hit the INSERT key to get into INSERT mode before you start typing, otherwise you end up typing over good stuff before you remember that your command line editor was designed by Bill Gates and Company and you have just clobbered some good part of the line you are correcting. All all all Microsoft design is permeated with such slightly off-kilter conceptions.

It gets old.

After years of this, I realized something one day. DOS was invented by BASIC programmers. The company Microsoft was founded by BASIC programmers. They were programming with BASIC and foisting 8 character file name length limits on the world starting in 1980. In 1980, UNIX was already 10 years old, already had introduced effectively unlimited file name lengths and directory hierarchies, had multitasking, worked on 16 bit computers, did not have an idiotic 1MB memory limit, etc. etc. etc. What we have here is that some guys in Albuquerque got lucky, were good hackers (with BASIC interpretors), were great marketers, and took over the planet.

Now, I don't have anything in particular against BASIC, per se. It was my first programming language in High School. It is good for doing lots of stuff. It dominated as it has because it is an interpreter and (like LISP and like Perl), an interpreter is a very convenient thing to have around to write small programs fast. OK, fine, BASIC has its place and there is a reason that non-computer-scientists took to it like ducks to water.

But still, why are we stuck with Visual Basic in 1999? Well, same reason. It is easy to program the first 90% of a task and Microsoft, over the years, has managed to hack in enough semblance of modern programming constructs that one can manage, via programming by experimentation techniques, to cobble together a solution.

Entire industries were born to remedy mistakes embedded within DOS: memory managers, file system administration tools, and myriad utilities were provided by a host of third party companies who owed there very existence to design mistakes in DOS. People who installed Windows are constantly asking their mechanically-inclined or engineering-savvy friends to help them reformat disks, reinstall programs, and help find their way through the maze that is Windows, all as a favor, and at no cost to Bill Gates and Company. It could be argued that the cost to the economy in lost time in circumventing and working around the frustratingly bad design of DOS and Windows far exceeds the personal fortune of Bill Gates.

We have left unaswered the question of why Microsoft has succeeded so well in spite of these design flaws. Although that is a separate subject, it deserves comment. Did Microsoft succeed because they got product to market faster by cutting corners? Maybe sometimes, but, in general, putting out a bad piece of software actually increases the cost to the producer in the long run. I think the real reason behind the success of Microsoft is what one friend of mine calls 'survival of the adequate'. If you can put something out which has glitches but works, if it runs on the Industry Standard Architecture (ISA bus), if it costs less than products by and for yuppies (the Macintosh), if some genuinely good applications are somehow made to run on it, then the cost-conscious public buys it, naively (and understandably) expecting it to work. Finally, by fiat you draft the community of programmers to be in a constant state of beta-testing your products, get them all to do programming by experimentation to figure out how to use your products and develop further genuinely good applications for it, much like a rat figures out how to get through a maze, and voila, world dominiation.

It used to work that way. Linux is here now. Linux is a better operating system and, if my thesis is correct, will always be better. Linux costs even less so the public will like that. Many programmers have awoken to the con job which was pulled on them. Coordinated efforts to produce a good looking GUI desktop with standards-based application interoperablity are rapidly being deployed on Linux. If the applications start to come around, good bye system crashes, welcome Linux.

Will Windows 2000 solve the problem? I can only speak from personal experience. I use both NT Server 4.0 (build 1381, Service Pack 4) and Linux (Redhat 5.2, kernel 2.0.36). I mostly use NT Server as my workstation, but do run it as a server to test some development. I run one Linux box both as a workstation and as a file, web, database, and setiathome server. I am typing this very document into an emacs running on Linux but which is being displayed on an X server running in NT.

I have seen numerous situations where Internet Explorer or the File Explorer crashes and takes NT down with it. Happened under Service Pack 3. Happens under Service Pack 4.

In contrast, I have seen Netscape crash many times in Linux, yes, but never did a program which crashed in Linux take down the operating system with it. My Linux machine has been up for over six months across mulitple reconfigurations of virtual IPs, installation of new versions of basic server software, and switching from an analog dedicated modem connection to a DSL connection, without once needing to be rebooted.

If Windows 2000 really is Windows NT, then I'm not optimistic that Microsoft is about to 'get it right'.

I urge you to view Microsoft kind of like you view the Y2K problem. A bunch of hacks were made many many years ago which, through time, have been patched and grown over. Under it all, there is the corporate identity which, after all, comes from the originators of the corporation. Time buries the clunky designs, the inability to write good English, the seeming utter ignorance of the concept of a complete specification, and the arrogance of either ignoring, not caring about, or trying to out do standards. But the corporate identity, the corporate culture fostered by Microsoft remains. The players change but the game stays the same.

UNIX came out of the Universities and corporate research labs. Originated by the brilliance of the likes of Ken Thompson at Bell Labs, extended by the U.C. Berkeley Computer Science department and Bill Joy later, and with a cast of hundreds, if not thousands, augmented and improved up through today. Stallman, Torvolds, Gosling and so many, so many good artist scientist engineers have made Linux, Free BSD, Apache and all the open source efforts the new paradigm -- a different culture has coexisted along side the commercial hacks of DOS.

The 'feel' one has when one is programming, using, reading about UNIX is different than with Windows because it is a question of that culture. All of the myriad underlying skills, computer science background, design decisions. Hard to pin it down to any one thing or even a finite list.

Windows 2000 will be a Y2K problem which cannot be fixed because it has with it all the trappings of the culture which has produced it over the past twenty years.

Linux, open source, Internet standards, and just plain good science, may yet change the rules. Then, finally, we will be playing a new game, and the referee won't be Bill Gates.