With the release of Windows Server 2003, Microsoft Corp. reached an important milestone. Both its desktop and server operating systems are now, by any reasonable measure, finished. With the demise of Windows Me, the company met its goal of extending the NT architecture to all x86 systems. Windows has become the unified platform it was always meant to be.
Some analysts insist that the jury is still out on stability and security. I don't believe that's so. Most of my IT colleagues who have been working with Windows Server 2003 throughout its various release stages find this to be not only the best Windows ever, but an impressive work of engineering viewed from any perspective. Coders invested substantial effort in code review, combining leading-edge automated tools with a new culture of individual accountability to find and close exploitable holes in the platform. Then came the performance tuning. Both projects produced impressive, sometimes remarkable results; Microsoft's technical staff really knocked this one out of the park. Everything learned in the server effort will strengthen Windows XP. Customers would probably now be content if Microsoft fixed bugs and plugged security holes, but otherwise left its operating systems alone.
Microsoft now has the time and resources to focus on two things: 64-bit Windows and Microsoft applications. The latter subject I'll take up another time, but 64-bit Windows is extremely relevant in light of AMD's April 22 launch of Opteron.
The fact that Windows was absent from the launch events for both Itanium and Opteron sent Intel Corp. and Advanced Micro Devices Inc., along with the IT market, the signal that if you want the benefits of 64-bit computing, you'll have to get them from Unix (a term that, through common usage, is now understood to include Linux). I still don't consider Itanium to be a Windows platform. In my view, the best software match for Itanium is HP-UX, with Linux second and Windows a distant third. I haven't tested BSD on Itanium yet. Maybe that could push Windows to fourth place.
Opteron had the potential to play out differently. A chip that straddles the line between 32 and 64-bit computing should be perfect for Microsoft and its customers. Before Microsoft embraced the Wintel partnership again, its executives fairly gushed over how sweet Windows looked on AMD's 64-bit architecture. AMD featured Windows, often exclusively, in every public demonstration.
If Microsoft leaves AMD and server architecture partner Newisys at the altar to appease Intel, as it appears it will, the tears will fall in Redmond, not Austin. Through a remarkable team effort that preceded even the availability of early silicon (developers used a frustratingly slow software simulator), AMD, SuSE, and CodeSourcery fixed up the GNU compilers and Linux kernel to run in Opteron's 64-bit mode. Those changes are now merged into the tools and the Linux kernel. That means that on the day it launched, Opteron was an officially supported Linux platform. Downloads of Mandrake's 9.0 Linux for Opteron were available for nearly a month prior to Opteron's launch. SuSE 8.0 was installed on the Newisys demonstration and development systems shipped prior to the Opteron launch.
Microsoft trails Unix on two of the three hardware platforms that are set to dominate the server landscape. With its work on the 32-bit x86 effectively complete, it should either focus its firepower on 64-bit Windows or cede the performance server market to Unix.