Guild Companies

A long, long time ago, which is about two years in the computer business, Windows 2000 was just starting to roll out and Microsoft was beginning to talk about the future "Whistler" and "Blackcomb" versions of its server operating systems. Late last year, as Microsoft struggled to explain how .NET would fit into these environments, along came "Longhorn," a stop-gap version of the Windows platform that had some better meshing with .NET. This week, Microsoft has killed it.

Now, we appear to be back where we started. Let me explain, if I can extract something that looks like reality from all of the convoluted gobbledygook that gets said about future operating systems.

The whole point behind the Whistler and Blackcomb Windows operating system roadmaps was for Microsoft to get its desktop and server operating systems all on the same 32-bit Windows NT kernel, extend it to 64-bit Itanium processors, and then keep them all more or less in lockstep. This has apparently been about as easy as herding cats, forget steers for the moment.

Whistler was originally expected in mid-to-late 2001, and the desktop version of the operating system is the only production version that has seen the light of day. We know it as Windows XP and Windows XP Professional. In early 2001, Whistler server, which is now called Windows .NET Server 2003 and which is the first version of the Microsoft platform to support 64-bit Itanium chips and NUMA clustering, was pushed out because it was not ready for primetime and Microsoft was struggling to get companies to move to Windows 2000 Server and Advanced Server. Then in the fall of 2001, in the aftermath of the terrorist attacks and a sharp downturn in IT sales worldwide that still continues, Whistler got pushed out to the first half of 2002. Then along came chairman Bill Gates' Trustworthy Computing initiative--a valid goal at Microsoft to make its products more secure and robust--and Whistler was pushed out to the end of 2002. Last we heard from several server vendors, no one is counting on seeing Whistler much before February 2003. That's three years after Windows 2000 launched.

Blackcomb was supposed to be in the market in late 2003 or early 2004, and it was supposed to have support for a what was then called the Next Generation Web Services at Microsoft and what is now called .NET. From a practical standpoint, what made Blackcomb interesting is support for advanced I/O and memory architectures, expanded clustering, and dynamic server partitioning--the kinds of things that are available in high-end Unix machines today. It is also expected to have tightly XML integrated into itself, not bolted onto the side as is the case with Windows 2000 and Unix servers today.

Earlier this year, during one of the many times when Whistler deliveries was being pushed out and Blackcomb was pushed out even further, Microsoft divulged an interim Windows desktop and server release called Longhorn, which was supposed to come out shortly after Whistler. Well, the desktop version of Longhorn is still expected sometime next year as a rev on Windows XP, but the server version has been sacrificed because Whistler server won't be out until then.

Even if Microsoft had enough features and gizmos to add into Longhorn server to make it interesting, corporate server buyers and system administrators will avoid doing two server upgrades in a year--or doing a jump from Windows NT or Windows 2000 directly to Longhorn--like the Mad Cow Disease. Getting the vast installed base of Windows NT 4.0 customers to move to Windows 2000 will be hard enough, and Microsoft seems content to focus on that. Whistler, whenever it does finally make it out the door, will undoubtedly be focused primarily at high-end customers who need 64-bit processor or NUMA support. Companies selling big Wintel boxes are plenty miffed that it is so late, and the continuing delays in Whistler are one of the reasons why the uptake of the Itanium processors from Intel has been so muted.

Microsoft is at a disadvantage compared to most Unix and Linux operating system vendors. Linux is still building up steam and is largely a pull--customers want it, they don't need to be sold on the idea. And with Unix vendors, they simply force customers buying the hottest new hardware to move to the latest operating system release, and thereby can get at least a predictable upgrade pattern. Microsoft didn't have to sell the upgrade from Windows 3.51 to Windows NT 4.0, especially after a few service packs were out. Windows NT 4.0 was a pretty good server, and only after a lot of marketing, demonstrations that it was a more stable operating system, and the impending end of life for Windows NT at Microsoft and its software partners did Windows 2000 start taking off. Getting server customers to move to Longhorn--unless it was a dramatically more stable and secure operating system than Whistler--would have been impossible. Most companies want to get on a two-year or three-year upgrade cycle for their servers, and Longhorn just doesn't fit in here. Such long upgrade cycles are the main reason why Microsoft is keen on moving to a services-based licensing scheme for its software. It gives a relatively steady revenue stream when customers aren't changing anything--which is in many ways as ideal for customers as it is for Microsoft.