Historical Perspective

Again I find myself indebted to David Solomon and Mark Russinovich for providing the opportunity to write a few words about their newest edition to a series of books on Windows Internals. It has been over three years since the last publication in this series, and this passage of time has seen two major releases: a very significant update to the client system, and another very significant update to the server system, currently being readied for shipment.

Two of the growing problems faced by the authors of a book such as this are tracing the implementation evolution of the Microsoft Windows NT system and documenting the way in which feature implementation has changed in each version. To this end, the authors have done a remarkable job of providing examples and explanations throughout the book.

(Left to right) David Solomon, David Cutler, and Mark Russinovich

I first met David Solomon when I was working at Digital Equipment Corporation on the VMS operating system for VAX, and he was only 16. Since that time, he has been involved with operating system development and teaching operating system internals. I met Mark Russinovich more recently but have been aware of his expertise in the area of operating systems for some time. He has done some amazing work, such as his NTFS file system running on Microsoft Windows 98 and his "live" Windows kernel debugger, which can be used to peer into the Windows system while it is running.

The beginnings of Windows NT started in October 1988 with a set of goals to produce a portable system that addressed OS/2 compatibility, security, POSIX, multiprocessing, integrated networking, and reliability. With the advent and huge success of Windows 3.0, the system goals were soon changed to natively address Windows compatibility directly and move OS/2 compatibility to a subsystem.

We originally thought we could produce the first Windows NT system in a little over two years. It actually ended up taking us four and a half years for the first release in the summer of 1993, and that release supported the Intel i386, Intel i486, and the MIPS R4000 processors. Six weeks later, we also introduced support for the Digital Alpha processors.

The first release of Windows NT was larger and slower than expected, so the next major push was a project called Daytona, named after the speedway in Florida. The main goals for this release were to reduce the size of the system, increase the speed of the system, and, of course, to make it more reliable. Six months after the release of Windows NT 3.5 in the fall of 1994, we released Windows NT 3.51, an updated version containing support for the IBM PowerPC processor.

The goal for the next version of Windows NT was to update the user interface to be compatible with Windows 95 and to incorporate the Cairo technologies that had been under development at Microsoft for a couple of years. This system took two more years to develop and was introduced in the summer of 1996 as Windows NT 4.0.

The following version of NT saw a name change to Windows 2000 and was the last system for which the client and server systems were released at the same time. This version was built on the same Windows NT technology as the previous versions and introduced significant new features such as Active Directory. Windows 2000 took three and a half years to produce and was the most tested and tuned version of Windows NT technology produced at the time. Windows 2000 was the culmination of over eleven years of development spanning implementations on four architectures.

At the end of Windows 2000 development, we embarked on an ambitious plan to implement new versions of the client and server systems, which would include new enhanced consumer features and improved server capabilities. As plans developed, it became clear that implementation of the server features would cause a lag in the implementation of the client features, and therefore, the releases were split. In August of 2001, Windows XP Professional and Windows XP Home Edition were released, and a little over a year later, in March of 2003, Microsoft Windows Server 2003 was released. In addition to the Intel x86 architecture, these systems contained support for the Intel IA-64, marking Windows NT's first move to 64-bit processing.

This book is the definitive work on the internal structures and workings of Windows XP and Windows Server 2003. In addition, it offers a glimpse into the future of Windows' move to 64-bit computing by covering AMD's introduction of the x64 architecture (AMD64) in 2003 and Intel's announced support (EM64T) in February 2004. A fully supported x64 client and server release is planned in the first half of 2005, and this book contains many insights into the implementation details of the x64 system.

The x64 architecture is the beginning of a new era for Windows NT at a time when the x86 architecture is beginning to show signs of old age. This architecture offers 32-bit x86 compatibility at speed to protect legacy software investments, and provides 64-bit addressing capability to address the most ambitious of new applications. This will protect 32-bit software investments while providing Windows NT with a breath of new life well into the next decade and beyond.

Although the Windows NT system has undergone several name changes over the past several years, it remains entirely based on the original Windows NT code base. As time has marched on and invention has thrived, the implementation of many internal features has changed significantly. The authors have done a laudable job of assimilating the details of the Windows NT code base and its differing implementations from release to release and platform to platform, and of producing examples and tools that help the reader understand how things work. Every serious operating system developer should have a copy of this book on his or her desk.

David N. Cutler

Senior Distinguished Engineer

Microsoft Corporation