One of the most common questions I hear now that Windows 98 has shipped is whether I think this version is the last version of Windows. My answer is no. Win98 probably isn't the last Windows. I've heard that This is the last Windows line before.

In 1991, Microsoft was trying to get Windows 3.1 out the door, and at the same time the company wanted to crush OS/2's chance of becoming a major player in the operating system (OS) market. But, OS/2 was a 32-bit OS. Windows 3.1's 16-bit nature seemed anemic in comparison. To compensate, Microsoft made a lot of noise about its then-developing Win32 programming platform, an enhanced, 32-bit version of the familiar Windows programming interface. Which OS would run Win32? Windows NT 3.1, which was due out by the end of the year. Don't buy OS/2, Microsoft suggested. Wait for NT 3.1. The 16-bit Windows is a dead end; NT will be the new Microsoft OS. It will be 32-bit and better than OS/2.

NT shipped later than 1991, of course. Microsoft held a developers conference in April 1992 to give programmers an early start on Win32. By mid-1992, the fact that NT was going to require expensive hardware was clear (to me, anyway). I expected that NT's hardware requirements might not sit well with every customer­33MHz systems with 8MB of RAM couldn't run NT but just might run OS/2 pretty well. I opined this expectation to some of the Microsofties at the conference and suggested that a Windows 4.0 might make sense. The Microsoft representatives said that no, Microsoft definitely didn't plan to release another Windows.

NT 3.1 shipped more than a year after the Win32 developers conference. By the time it shipped, its hardware requirements were common knowledge, and customers started looking more seriously at IBM's OS offering. Microsoft kept these customers from jumping ship by spreading the word that it was working on Windows 4.0, a 32-bit version of Windows that became Win95. After Win95's release, Microsoft claimed that Win95 was the last Windows. Now we have Win98. (For Mark Smith's opinion about whether Win98 should be the last Windows, see "S, M, L, XL, and XXL," September 1998.)

Why would Microsoft build Windows 2001? Three reasons: old hardware, new hardware, and old software. NT Workstation 5.0 will probably be great, but if it's everything I want in a 32-bit desktop OS, it'll probably require 300MHz and 96MB of RAM to run. I won't object to those requirements, but some customers might. Some folk think that an OS that forces them to buy more than 32MB of RAM is unacceptable. Microsoft says those folk will be happy with Win98, but I disagree. What will Win98 customers do when Institute of Electrical and Electronics Engineers (IEEE) 1394 (or another new hardware doodad I can't even imagine) becomes popular? If Win98 doesn't support the latest device and NT's too expensive for some people, OS/2 or a friendly version of Linux will have a window of opportunity.

And remember, NT will never run DOS programs. This shortcoming doesn't apply only to ancient code. I bought the DOS games Wing Commander: Prophesy and Master of Orion II a few months ago. They're hardly legacy code, but they'll never run on NT. Is Microsoft really willing to cede to a competitor the DOS market and any customers who don't want to purchase the latest machines? I doubt it.

P.S. Big grumbles this month go to Citrix Systems for not writing an Alpha version of MetaFrame. Microsoft Windows NT Server 4.0, Terminal Server Edition needs horsepower to work, and Alpha is the fastest engine available. Independent Computing Architecture (ICA) is the protocol of choice for Terminal Server users. Big applause goes to Mark Russinovich for BlueSave, a cool tool that saves NT blue screens to a text file. (For more information about BlueSave, see http://www.sysinternals.com.)