Programming Windows: NT Everywhere (Premium)

“NT EVERYWHERE! NT EVERYWHERE!” the crowd of mostly male Microsofties chanted, inadvertently channeling a Nazi rally and making the walls pound. I looked nervously at the co-worker who’d snuck into this building on the Microsoft campus in Redmond, Washington. We have to get out of here, I mouthed. “NT EVERYWHERE! NT EVERYWHERE!” There was Jim Allchin, smiling serenely, then chanting the words. “NT EVERYWHERE! NT EVERYWHERE!” There was Steve Ballmer, pumping his fist, somehow louder than the crowd around him, almost crazy-eyed. “NT EVERYWHERE! NT EVERYWHERE!” The coworker nodded, and we edged through the crowd, nervous. “NT EVERYWHERE! NT EVERYWHERE!” There was the door. Almost home. “NT EVERYWHERE! NT EVERYWHERE!” My hand was on the bar, I could see the light. “NT EVERYWHERE! NT EVERYWHERE!” And then it happened. “HEY! YOU TWO GUYS! STOP!”

—

Dave Cutler signed off on the first version of Windows NT—which was branded as version 3.1 to match the new DOS-based version of Windows—at 2:30 pm PT on Monday, July 26, 1993. He had been hired by Microsoft almost exactly five years earlier specifically to make this moment happen, and he had taken scores of engineers and hardware experts with him from DEC to create a portable OS for the ages, a Unix killer.

NT’s development had taken a heavy toll on this team and a slowly expanding group of Microsofties that had been let into the inner sanctum. There were divorces and nervous breakdowns. Missed births, birthday parties, and anniversaries. Canceled vacations.

Naturally, Microsoft co-founder Bill Gates was “gleeful,” author G. Pascal Zachary wrote in the industry classic Showstopper! The Breakneck Race to Create Windows NT and the Next Generation at Microsoft. He thanked the NT team for its “super effort,” predicting that Windows NT would “redefine the expectations people have for operating systems.” “No other project of this scale has ever been done before at Microsoft,” an unnamed executive added to Gates’ assessment. Left unsaid: At $150 million in costs, NT was the most expensive software program ever created at any company.

Over on the Cairo team, Jim Allchin was busy leading a team that would build on top of NT and create what had, to that date, been described as the next version of NT. Cairo would add object-orientation to the OS, whatever that meant, in the form of an object-based file system and other improvements.

“NT is a truly amazing accomplishment, probably the most ambitious project ever completed successfully in the world!” he gushed. He was obviously inspired by this, too, since he subsequently set off to create two projects inside Microsoft that were even more ambitious: Cairo and Longhorn. That both were epic failures is, perhaps, a reflection on how difficult it really is to pull off something as ambitious as NT. It was, perhaps, a once in a lifetime event.

Outside of the insular NT team, however, Microsoft’s world was changing.

The software giant—and it was truly a giant by that time—had just released Windows 3.1, and despite being a fairly minor upgrade for users, it was somehow even more successful than its mighty successor. It also arrived at a time when Apple was faltering and then cratering, and when Microsoft’s productivity application competitors were getting knocked down, and then knocked out, by the firm’s aggressive—some said illegal—business practices. This created an opportunity for Microsoft to seize control of both the PC industry and the graphical user interface. Microsoft, suddenly, was alone at the top of the market.

And a bigger team at Microsoft was already working on the successor to Windows 3.1, which would bring some of NT’s 32-bit power to its shaky MS-DOS underpinnings. Windows 4.0, like NT—like every major Windows version before it, really—would be delayed multiple times, and wouldn’t ship until the second half of 1995. But it would also include a new user interface that had been originally created for Cairo, albeit without any of the resource-intensive OOP nonsense.

Outside of Microsoft, the reaction to Windows NT was much calmer than the reaction to Windows 3.1. Reviewers and even users seemed to understand that NT was the future, somehow. But they needed to get work done, and Windows 3.1 wasn’t just faster than NT, much faster, it was more compatible. Both with the applications they relied on and, even more obviously, with the hardware their PCs required. Critics said that NT stood for “Not There.”

Not there yet, perhaps. But NT’s performance and compatibility issues made onlookers forget what a miracle this program really was. Even in its first iteration, NT was a modular, portable wonder that would take on multiple OS “personalities” and arbitrarily run software applications that had been written to different platforms and standards, and do so side-by-side in a safe, multitasking environment.

Users may have been confused about NT, but the industry wasn’t. Novell saw the writing on the wall, and it understood that NT was the biggest threat yet to Netware, so it purchased Unix System Laboratories from AT&T so that it could offer a Unix-based alternative to NT. Microprocessor makers approached Microsoft for NT ports to their platforms, resulting in versions of the product tailored for the PowerPC (IBM, Apple, and Motorola), Alpha (Digital), and Mips platforms. Suddenly, Bill Gates’ vision for Windows as a scalable platform that could meet any need, from the smallest handheld device to the most powerful servers, seemed within his grasp.

But NT did need to improve, and Cairo would keep getting pushed back further and further until it melted into nothingness when Windows 4.0, rebranded as Windows 95, emerged as the most successful product in Microsoft’s history to date, shattering the already-heady popularity of its predecessors. Cairo wouldn’t be the next version of NT. It would never happen, period.

Instead, Microsoft split the NT team into two main groups, with Jim Allchin as the overall leader of these efforts. One was working on Cairo, still, of course. And the other was led by NT architect Dave Cutler, who wanted to address the “too big, too slow” complaints that so many had leveled against his product. This version of NT, which would later be branded Windows NT 3.5 because it was partway between the past (3.1) and the future (4.0), was codenamed Daytona.

(Fun aside. A typical type-A bully, Cutler and some of his alpha-male cohorts would blow off steam racing cars at a Daytona International Speedway in Florida. Naturally, the speed of these cars spoke to Cutler’s goals for Windows NT 3.5.)

“Every first release of an operating system is a compromise,” Cutler said. “So it is with NT.”

Fixing NT was simple enough—make it faster and smaller, a task Microsoft would later undergo yet again when the Windows 7 team had to fix Jim Allchin’s mistakes in Windows Vista—but Cutler was also unimpressed with the plans for Cairo. He felt Cairo was a distraction, and one that was sucking far too many people and resources from the job at hand: NT needed all the polishing it could get, he said. And that should come first.

And so it did. Microsoft shipped Windows NT 3.5 in September 1994, and in keeping with Cutler’s portable needs, it ran on the Intel i386 and newer, DEC Alpha, and MIPS platforms. It was aligned, roughly, with Windows for Workgroups 3.11—why the different version numbers is unclear—and offered a handful of new features, including long file name and OLE 2 support. But Daytona, true to its name, was really about speed, and it showed.

In May 1995, Cutler’s team then shipped Windows NT 3.51, a minor update that added support for the PowerPC platform. By that time, Windows 4.0 had been renamed to Windows 95 and excitement was growing for the constantly-delayed major upgrade to Windows 3.1. Windows 95 was obviously going to be very successful, in sharp contrast to Cairo, which by that point was fading fast. And in a weird about-face, Microsoft prepared to bring the technically simpler Windows 95 user interface, or shell, to Windows NT instead of waiting for the OOP-based Cairo shell on which it was based.

Three days after Microsoft shipped Windows NT 3.51, it released a preview version of this new user interface called the Shell Technology Preview. Not surprisingly, it proved successful with testers, as it would later that year in Windows 95. So Microsoft began plotting a major new release of NT, called Windows NT 4.0, that was internally called the Shell Update Release (SUR). Windows NT 4.0 was finalized in late July 1996, and upon its release a month later, it basked in the glow of the successful and similar-looking Windows 95, pushing NT to new levels of market acceptance and usage.

(Fun aside. I showed up at a Software Etc. store in Phoenix on the night of August 24, 1996, when Windows NT 4.0 became available to the public at midnight. Before the event, I had attended a pool party at the home of one of my wife’s coworkers, and I had pushed into the pool with all my clothes on by some coworkers. So when I walked through the store to pick up my retail copy of Windows NT 4.0—I had written a book about the product for the education market already—I made a wet squishing sound with every step, triggering some confused looks by employees.)

Windows NT 4.0 was notable for many reasons, but in retrospect, it is perhaps most notable because it is the last version of the product to run on the Alpha, Mips, and PowerPC architectures. Perhaps not coincidentally, this is also the first version of NT in which Cutler finally agreed to move NT’s graphics subsystem out of user space and into the kernel; that design is less secure, but it provides significantly better performance, and this change finally quieted most complaints about NT’s performance for good.

Microsoft never really abandoned its portable dreams for NT, but with the next release, originally called Windows NT 5.0, running only on Intel hardware originally, the team was free to turn to the Assembly language optimizations that Cutler had previously denied, further improving performance. NT was still portable—it would later be ported to the 64-bit Intel Itanium and then, much later, to different ARM-based platforms—but much less so than in the past.

Microsoft released the first beta version of Windows NT 5.0 in September 1997, and the original plans were ambitious. This was the release where Microsoft would finally combine the underpinnings with NT with the driver compatibility of Windows 95, creating a single platform that would serve all users. Those plans were eventually scrapped until the subsequent release, codenamed Whistler, which arrived years later as Windows XP. Windows NT 5.0, oddly, had no codename, because Jim Allchin, burned by his inability to ship Cairo, suddenly didn’t like codenames.

And this is where I enter the story.

In August 1998, I flew to Seattle, Washington for the first time to attend the Windows NT 5.0 Beta 2 Technical Workshop. At the two-day event, which was held in a small conference room in a hotel next to the Washington State Convention Center, we heard from top executives on the Windows NT team like Allchin, Yusuf Mehdi, Moshi Dunie, and Mike Nash. Indeed, I asked Mehdi before the start of day one where Dave Cutler was—obviously, I was eager to the meet the man—but was told he was off “working on 64-bit stuff.”

I still have my notes from that workshop.

Windows NT 5.0 would feature the personalized menus from Office 2000, Internet Explorer 5.0, and “one-stop” mobile sync. It would be faster than Windows 98, at least on systems with 32 MB or more of RAM. It would equal Windows 98 with regards to USB, hardware, PnP, 1394, ACPI, and DirectX 6, and it would help solve the “DLL hell” problem that Microsoft had created for itself in 1985 with the shared library system it implemented in the first Windows.

Microsoft would ship Windows NT 5.0 “when customers tell us it’s ready,” I was told, when there are no “Sev 1” bugs—the worst kind—just as was the case with the first version of NT. Microsoft said that it was running its own business on NT 5.0 already, on over 100 servers and over 20,000 desktops, and that internal tests showed it was better than NT 4.0 in application compatibility, stress tests, and hardware tests.

Microsoft even discussed the post-NT 5.0 future. It was a world where NT would support 64-bit clustering, more integrated storage, consumer simplicity (with “auto everything”), an adaptive user interface, developer richness with web and Win32 unification, localization independence, and market ubiquity, where the codebase would be ported to embedded platforms of all kinds, real-time systems, and more.

It sounded fantastic, and to be fair, most of that did happen. But NT, as a brand, as an identity for something that was once truly great, would not make the leap into this future.

Two months after that technical workshop, Microsoft announced that Windows NT 5.0 would be branded as Windows 2000.

This signaled the end of an era, and that the marketers had won a crucial internal battle over the engineers who believed, as I did, that NT had a soul and meant something. It also signaled to the world that Microsoft’s plans to merge Windows NT with Windows were proceeding even though this release, Windows 2000, would not achieve that.

Before all that happened, however, I was sitting in a room in a hotel conference room, mesmerized. I didn’t get to meet Dave Cutler, but I did get to hear directly from key Microsoft executives, some of whom I’d later have personal relationships with. And I met Mark Minasi, who I would soon join over at Windows NT Magazine. He sat in front of me at the workshop, nodding his head eagerly as the Microsofties spoke.

A friend at Microsoft told me that the NT team was going to celebrate the release of Windows NT 5.0 Beta 2—build 1877 had been signed-off on the night before—that afternoon at a building whose number I now forget at the Microsoft campus in Redmond, across the lake from Seattle. We had rented a car, so I decided to head over to the campus with a coworker. We met the friend, briefly, and he pointed us at the building where the celebration would take place. That he wanted no part of it should have been telling. But we simply followed a stream of people heading inside the building.

Back then, security on the Microsoft campus wasn’t as tight as it is today. Then, you could simply follow anyone through an open door, but today, everyone is required to “badge in.” Being 31 years old at the time, I looked exactly like everyone else at Microsoft. So no one questioned my presence, especially given how packed the room was.

Jim Allchin and Steve Ballmer both spoke, and the crowd got more and more excited as it continued. Dave Cutler never showed up, and come to think of it, he was the real reason I had snuck in. And then, as Ballmer finished up, the chants began.

“NT EVERYWHERE! NT EVERYWHERE!”

We headed for the door, nervous that we had clearly interrupted something that the NT team didn’t intend the world to see. You could tell it was about to get really stupid in there.

Just as I put my hand on the bar on the door that would open it to the outside world, someone saw us exiting and yelled out, “HEY! YOU TWO GUYS! STOP!”

We froze. Shit. Here we go.

I turned around, and a man was standing there holding something in his hands. “You forgot your shirts!” he said, thrusting two SHIP IT shirts into my hands. He smiled, clapped me on the shoulder, and then turned around and took up the chanting with everyone else. Next to me was a table piled with these shirts so that attendees could grab one for themselves when they left. In my eagerness to leave, I hadn’t noticed it.

Standing there in the exit, looking in on a world in which I was really only on the periphery, I paused, wondering about this company, and this product, and the people who had made it. And I wondered, not for the last time … could I ever work at such a place?

No.

But Windows NT 5.0 was so exciting that I started the Windows NT 5.0 SuperSite, which was later renamed to the SuperSite for Windows NT 5.0—Microsoft didn’t like third-party sites using “Windows” as the first word in the names back then—and then finally the SuperSite for Windows as I started covering Windows Millennium and then non-Windows products as well.

This was where I belonged. Covering Microsoft from the outside.