Agreed. All the tech popularizers and business people focus on the non-exclusive license for DOS they gave IBM as the smartest decision Bill Gates ever made, and I guess it was, but for my money a very close second was hiring Dave Cutler and his people from DEC to build a real OS kernel. Without the NT kernel to fall back on, Windows would have collapsed under its own weight long ago.

On the other hand, one of the worse ones was the fiasco involving OS/2 2.0, which was much closer to NT than Win9x was and for example could have been used to legally attack DR-DOS. Why did it take 10 years after Intel released the 80386 before 32-bit programming became popular? I have a thread about it here: http://www.os2museum.com/wp/?p=516&cpage=1#comment-3413

> Windows XP was a much bigger deal than Windows 95, being the first consumer release based on the NT kernel

I hung on to 98 until 2001 for gaming, but at that point most stuff worked perfectly on 2k. While it wasn't intended for "home" users, 2k was the preferred OS among my group of friends until XPsp2 or so.

I really don't get why the NT kernel wasn't introduced to the "home" market until XP.

NT5/Windows 2000 was delayed by a couple years (supposedly due to the corporate networking stuff). The hardware was certainly ready for it by about 1998.

There was also a companion "Windows 2000 Home" edition announced, however it was rolled into XP while Windows ME was released as largely unwanted stopgap.

I was a big proponent of Windows 2000 as well. Though not marketed/positioned as such, it was really the start of the merging of the consumer and corporate OSes... based on NT kernel, but had support for DirectX and other APIs Microsoft never bothered to support in older NT releases.

I think Windows 2000 Pro was the peak of the windows line.

Windows 2003 was also very good for similar reasons, though you did have to do some configuring to make it more of a workstation OS than a server OS.

How terrible third-party Windows applications were. Writing cooperative multitasking applications is hard and much blame was assigned to Windows when it should be on applications.

I agree with you up to a point but given the hardware of the time, it was certainly possible to make the OS more robust than it was prior to the switch to NT everywhere, and to Microsoft's credit they did make the correct decision.

Now, despite the lingering snarkiness from the *nix-hipsters, Windows is incredibly robust... even more so today since in Vista and on even the graphics driver can crash without an OS restart in most cases.

True. The BSOD bloggers love to write about, for instance, was mainly caused by faulty hardware or defective third-party drivers. The OS, conservatively, entered panic mode to prevent damage to data. Somehow it seems acceptable for a Unix kernel to panic on such conditions but when it happens to a machine running Windows, that must be because Micro$oft programmers are so stoopid...

Windows 95 could preemptively multitask, though. Its "instability" was largely due to it not having a HAL.

Windows 95 could preemptively multitask 32-bit code. But its core was 16-bit, and there were many legacy 16-bit apps, all of which were cooperatively multitasked.

"The whole point of a doomsday machine is lost... if you KEEP IT A SECRET. Why didn't you tell the world, EH?"

(It doesn't matter who's first. It matters who gets it into the hands of double-digit millions of end-user CPUs.)

[Edit: I was replying to a "But, but, BeOS had protected memory first!" comment.]