Programming Windows: Transitions (Premium)

Microsoft’s revenues had long been dominated by Windows and products that required Windows, like Office. But by 2007, Office had become Microsoft’s biggest business and Windows Server had emerged as a third core business. And so it is perhaps not coincidental that the software giant moved to shore up the Windows client by splitting Windows Vista to include more premium product versions, as had been done before successfully with Office. The results were immediate: in the quarter ending March 31, 2007, over 70 percent of initial Vista consumer sales were of premium versions, and premium Vista versions contributed 1 percentage point of growth to sales from PC makers.

The numbers were interesting. Windows client revenues surged an incredible 67 percent to $5.3 billion in the quarter, it reported, and Windows accounted for about 37 percent of Microsoft’s total revenues of $14.4 billion. But it had deferred at least two quarters of Vista-related technology guarantees: Windows client had really delivered just $4 billion in revenues in the quarter, or about 28 percent of Microsoft’s revenues. And the deferrals were over, so future quarters would deliver a more accurate telling of Vista’s success.

Then, on July 19, 2007, Microsoft reported that it had earned over $50 billion in annual revenues for the first time, a gain of 15 percent year-over-year. The firm touted the “solid customer acceptance” of Windows Vista in the announcement, but it also mentioned that it would drive growth in the coming fiscal year, which stretched from July 2007 through June 2008, though a long list of coming new product offerings that included Windows Server 2008, formerly called Longhorn Server, but not Vista.

For all its inevitability on PCs, Vista was a laughingstock, and not just because of Apple’s incessant “I’m a Mac, I’m a PC” TV ads, which Microsoft silently ignored. It had taken several years to come to market and didn’t deliver on the most interesting of its original promises. It ran slowly on existing PCs, and even those PCs that could run it often didn’t have the graphics hardware to display the beautiful Aero UI, with its glass windows and transparencies. There were rampant compatibility issues with both software and hardware, a problem that was exacerbated in the 64-bit versions of the product. And Microsoft was forced to keep selling Windows XP, especially for a new type of low-end and inexpensive portable PCs called netbooks.

To fix these problems, Microsoft had brought in a fixer in Steven Sinofsky, the man who had successfully released the previous four versions of Microsoft Office, itself a complex and dominant platform, on time and on schedule over the past several. Sinofsky had been promoted and sent to the Windows team in early 2006 to focus on post-Vista versions of Windows, like “Fiji,” which became Windows Vista Service Pack 1 (SP1) and “Vienna,” which became Windows 7. So all would be well.

Eventually. In the meantime, the Windows customer base suffered through several months of silence, wondering what Microsoft would do to fix Vista’s obvious problems. And when. This was an unfortunate introduction to the cone of silence that would characterize Sinofsky’s overly secretive and data-driven Windows organization. The transparency of the Jim Allchin years was over.

As I wrote at the start of this series, the history of Windows is largely one of reactions. Reactions to industry trends, to competitors, and to the visionary, almost messianic belief of Bill Gates that he and his company could determine the course of history and push technology—and Microsoft’s revenues—to new heights. But Sinofsky’s ascendence into the lead position of the Windows organization represented yet another type of reaction, in this case to the regime that came before him. Sinofsky was the anti-Allchin, and he was determined to remove every trace of Allchin’s influence over Windows, from the product to the personnel who were responsible for it. That process began in earnest in early 2007 after Allchin retired. And it would end in a terrible “not invented here” pogrom that would see Sinofsky trying to replace key parts of Windows and erase the past while forever leaving his indelible stamp on the product as he plotted secretly to lead all of Microsoft.

Fortunately, that wouldn’t happen. But Sinofsky did have one major and permanent impact on the Windows organization that persists to this day. He instituted a policy of overly relying on telemetry data to understand which features that customers were really using the most, and then he used that data to justify cutting features and functionality that only served a minority audience. Understanding the limits of this type of thinking would require a book of its own. But long story short, it’s easy to tailor any data to match your expectations, and Sinofsky went on to ride this path to oblivion.

We’ll get to that. For now, simply know that Microsoft’s customers in 2007 waited several months to discover how and when the software giant would get around to fixing the very real and obvious problems with Windows Vista. And that happened because Sinofsky refused to communicate anything until he was sure his team could deliver it. And when that time finally came, they would consistently over-communicate and over-explain ideas that, frankly, didn’t warrant that much attention. Vista’s ills were as easily described as they were fixed.

In August, Microsoft finally came clean about its plans to fix the biggest Windows Vista complaints with a Service Pack 1 (SP1). And it would take longer than expected: Windows Vista SP1 wouldn’t ship until early 2008, the company admitted. But this was a good thing, it claimed in a tacit condemnation of Jim Allchin and his transparent communication approach.

“We know that providing timely guidance on release plans is important, but that it’s equally important for us to provide more accurate guidance that they can be confident in as they build their own plans,” Microsoft senior vice president Jon DeVaan said. “For Windows Vista SP1, that’s meant waiting until we had a higher-level of certainty in our plan, including what was going into it and when we could reasonably expect to meet the quality bar, to share information broadly. Finding the right balance between communicating earlier and more often versus later and more precisely is something we’ll continue to refine by listening to our customers.”

Windows Vista SP1 would not include new features, Microsoft said, drawing a direct comparison to the disruptive Windows XP Service Pack 2 (SP2), which it now saw as another Allchin mistake. Instead, Vista SP1 would focus on addressing some of Vista’s biggest problems: performance, reliability, and compatibility.

“We think Windows Vista is one of the best versions of Windows we’ve ever released,” DeVaan added, leading the clear-eyed to wonder why it wasn’t the very best version yet. But Allchin was out, Sinofsky was in, and his team was more interested in promoting themselves than they were in fixing Allchin’s product and legacy.

That month, Microsoft shipped a test version of Vista SP1 to a small group of testers and then followed up that release with a broader Beta 1 version in September. Thanks in part to Windows Update, which made it easier to deliver smaller bug and security fixes over the Internet when needed, Microsoft was changing how it handled service packs. Now, they would only include rollups of previously released and new fixes, plus support for some emerging hardware standards.

“We can test Wireless-N and certify it [for Vista] now even though the spec isn’t technically finalized yet,” I was told, as an example. “We [don’t have to] wait on Windows 7.”

Windows Vista SP1 was being co-developed with Windows Server 2008 (formerly Longhorn Server), and they would share the same kernel. “Windows Server 2008 and Vista SP1 are aligned,” the company confirmed. “They are a common engineering project, two separate products that share the same kernel and code base.” They would also ship simultaneously in the first quarter of 2008.

But the new Windows team couldn’t stop slagging its predecessors. In an October 2007 presentation, engineer Eric Traut discussed what he described as MinWin for Windows 7.

“A lot of people think of Windows as this really large, bloated operating system, and that may be a fair characterization,’ he said. Noting that a full Vista install image takes up over 4 GB on disk, MinWin, by comparison, was a tiny 25 MB, and Microsoft was working to make it even smaller, he said. The presentation had the desired effect, with multiple news outlets excitedly reporting that Microsoft was about to replace the bloated Windows Vista with a more svelte and lighter Windows 7.

Historical sidenote: The final versions of Windows 7 would take up about 3.2 GB on disk, depending on the product edition.

As I noted at the time, however, they should have Googled the term MinWin first. MinWin was not new to Windows 7, and I had first reported on the term back in a May 2003 article describing Microsoft’s componentization efforts for Longhorn, which, of course, became Windows Vista. Yes, Microsoft was continuing and evolving that work for the next release as well. But suggesting that it was completely different was disingenuous on Microsoft’s part.

I later discussed this topic with Microsoft technical fellow Mark Russinovich, who rose to fame in his pre-Microsoft days by reverse-engineering the NT kernel and publishing the simple registry change that would turn Windows NT 4.0 Workstation into Windows NT 4.0 Server. The Windows 7 version of MinWin, he said, was a more sophisticated approach to carving off a bottom chunk of Windows that removed any dependencies between MinWin and the rest of Windows. “Obviously, there’s a certain architectural goodness there,” he told me, noting that the new MinWin effort was also more sophisticated than the Server Core work that went into Windows Server 2008.

Microsoft released Windows Vista SP1 to manufacturing on February 4, 2008 alongside Windows Server 2008, revealing that it would begin delivering the update first to its Volume Licensing customers in March. “Windows Vista SP1 includes quality improvements that help enhance reliability, security, and performance,” the company noted blandly.

But the enthusiast community—admittedly somewhat beaten up and ragged after the years-long drama of Longhorn/Vista—was already looking ahead to Windows 7. And for some time, we had to make do with various leaks, which were identified as early pre-release milestone builds in the M1 and M2 trunks. Sources inside of the Windows organization explained to me that there would be three pre-Beta milestones, so we still had M3 to get to. And that there would be a visual refresh after M2 that Microsoft was keeping secret so it wouldn’t leak ahead of the official announcement.

“Windows 7 is the product of two teams working together: the WEX (or Windows Experience) team, which has primary responsibility for client releases as well as most user-focused features, and COSD (Core Operating System Division), which has responsibility for the kernel, networking, device support, and so on,” I was told. “COSD operates on a longer runway than WEX, meaning that work from that team will show up later in the cycle than work from the WEX team. SteveSi [Steven Sinofsky] is only in charge of the Windows team. Jon DeVaan runs COSD.” That would change, chillingly, in the future.

My source also told me about another major shift in Windows. Sinofsky’s team was pulling many of the built-in apps—like Mail, Calendar, and Photo Gallery—out of the product and placing them in the hands of the new Windows Live team. This was done for several reasons, most notably because they could now be updated more frequently and not just when Windows was updated. But this also contributed to Windows 7 having a smaller on-disk size. Microsoft would later promote Windows Live by noting that Windows 7 wasn’t complete until users had installed those apps as part of something that would be called Windows Essentials.

I was also prepped for the fact that Windows 7 would be much like an Office release: a minor, iterative update that was marketed as a major new release.

“It’s important to understand one aspect of the SteveSi philosophy, which might be phrased as ‘there is always another version’ or, more generally, the philosophy of developing release-rhythm,” the source continued. “Under Allchin, Windows tended to get into the ‘big-bang or dot release mindset [where] each release either qualified as a dot-release (i.e. 6.0 to 6.1 — a minor update or refresh, usually ignored by the senior folks) or a ‘big-bang’ release, which would try to change the world in one go. Very few dot-releases ever lasted as such … Big-bangs tend to collapse under their own weight (see: Longhorn).”

“SteveSi is a much more measured guy and [he] believes in regular releases, though his idea of ‘regular’ might still be long for some people,” he continued. “So teams will have visions that stretch to Windows 8, and you’ll see a first, thoughtful implementation in Windows 7 that sets the stage for Windows 8 (and also gathers important data about how the feature is used, to ensure that the complete implementation is better). Different parts of the product will get focus with each release, ensuring that there’s always something that is a killer feature, but there’s no attempt to make sure that everything gets an overhaul in every release. Over time, the whole OS will show improvements across the board, but the focus and long-term planning ensure that it’s a smooth upward path, not a series of spikes and (unfortunately) valleys.”

My source added that this was the model that Sinofsky had used while leading Office. If you were to compare the latest version of Office to its predecessor, you would see little change. But if you compared it to a version from several years earlier, you would see huge improvements. “Office 2007 looks drastically different,” he said, “but very few new features were added to the individual apps. The focus was on adding the ribbon to the core apps, and many old features were exposed by the ribbon, which makes it look like much more was added than actually was.”

I also learned that Microsoft would rev Calculator, Paint, and Notepad in Windows 7 but not rewrite them using Windows Presentation Foundation (WPF). WordPad, meanwhile, was also getting a modern refresh that would include the ribbon UI that first debuted in Office and support for new XML-based document formats.

“Sitting on the desk of Steven Sinofsky, Microsoft senior vice president of the Windows and Windows Live Engineering Group, and the man most directly responsible for Windows development, is a document titled ‘Shipping Seven’,” I wrote in January 2008. “I haven’t seen it, and I don’t have a copy of it. Sinofsky, of course, would like to keep it a secret: he’s among the cagiest of Microsoft executives, and no particular fan of mine. To date, Sinofsky and Microsoft have erected a wall of silence around 7, fearful of the leaks and transparent communication that ultimately doomed Vista as something that just didn’t live up to the company’s lofty promises. Microsoft’s goal this time around is to under-promise and over-deliver.”

Despite the silence, a schedule emerged: Microsoft would publicly reveal its plans for Internet Explorer (IE) 8, a key component of Windows 7, at the MIX ’08 Conference in early March 2008. It would then reveal its vision for Windows and the future of PC computing at the Professional Developers Conference (PDC) 2008, in October. And then it would unveil some low-level Windows 7 details at its Windows Hardware Engineering Conference (WinHEC) 2008, which had been rescheduled from its usual spring timing to that fall to better align with Sinofsky’s schedule. The company had publicly stated that it planned to ship Windows 7 in 2010, but my sources said that the real timeline was 2009.

At the MIX’08 Conference, Microsoft unveiled Internet Explorer 8 as expected and issued a Beta 1 version that Microsoft labeled a “developer preview release” at the last minute to lower expectations.

“IE 8 will be a major upgrade to Microsoft’s venerable browser,” I wrote at the time. “What we know today is, of course, mostly developer-oriented. But I have some information from sources at Microsoft that suggest a lot more is on the way. For example, there will be a new user interface, one that’s not here today in Beta 1. This UI will be based on customer feedback, which told Microsoft that the discovery of Favorites and adding to Favorites was too difficult in IE 7, that users want to resize the search bar and move the Stop/Refresh and Home buttons to new locations. Users requested more consistency to the tabbing functionality, and they wanted better exposure for help and the Send Mail feature.”

Microsoft had also experimented with adding the ribbon UI to IE 8, but it had found that UI was too ponderous for the browser and backed off. Here are the prototypes I was shown:

In March, I finally got my hands on Windows 7 build 6519—on DVD, no less—and sources told me about some of the “not invented here” changes that were coming to this version, including the removal of the classic Start menu, a feature Windows Vista had retained for upgraders. Windows 7 would at least retain the Classic desktop theme, I was told, but it would use a classic version of the new Start menu only.

“Windows 7 in day-to-day use is exactly what distantly viewed screenshots and virtual machine installs had suggested: It’s an update to Windows Vista with an extremely similar user experience and performance,” I wrote of the build. “It reminds me, very much, of the very first ‘community test’ preview build that Microsoft shipped for Memphis back in, oh, late 1996: it was stable, fast, and had just enough going on to be interesting. Windows 7 build 6519 is just like that. If you’re not paying attention, you’re clearly running Windows Vista. Look a bit deeper and, oh my, what’s that? And that? There are hundreds of minor changes.”

In May, Microsoft finally started discussing Windows 7 with the public, and it did so with a blog post by Chris Flores that was short as it was short on the details. But it was clearly aimed at getting in front of those—like me—who were leaking internal information, and others who were just spreading rumors.

“Yes, we are working on a new version of Windows,” he wrote. “As you likely know, it’s called Windows 7. We are always looking for new ways to deliver great experiences for our customers. This is especially true of Windows, where we’re constantly examining trends in hardware, software, and services to ensure that we continue to drive the innovation that has both made Windows the world’s most popular operating system and has provided a foundation on which our partners built great products and businesses. When we shipped Windows 2000, we were already working on Windows XP, and we started working on Windows Vista even before we released Windows XP. So naturally, we’ve been thinking about the investments we made in Windows Vista and how we can build on these for the next version of Windows.”

“What is a little different today is when and how we are talking about the next version of Windows.  So, why the change in approach? [He never explains why.] We know that this is a change in our approach, but we are confident that it will help us not only to build even better products, but also to be more predictable in the delivery of our products. We also know that this change has led to some confusion, so we would like to share information today that will hopefully clear up some of this.”

Windows 7 was a major release, Flores claimed, answering one major question about this version. But he deflected the rationale for that designation, too: “It’s hard to describe any product that is used by millions of people and worked on by thousands of engineers as anything else. [But] one of our design goals for Windows 7 is that it will run on the recommended hardware we specified for Windows Vista and that the applications and devices that work with Windows Vista will be compatible with Windows 7.” This, it turns out, was a deliberate mischaracterization: Sinofsky was actually planning to lower the system requirements with Windows 7, a move that would be celebrated as the team over-delivering.

“We are well into the development process of Windows 7, and we’re happy to report that we’re still on track to ship approximately three years after the general availability of Windows Vista,” Flores concluded in his content-free post. “As always, we will be releasing early builds of Windows 7 prior to its general availability as a means to gain feedback, but we’re not yet ready to discuss timing and specific plans for any Beta releases. In the meantime, customers can confidently continue with their Windows Vista deployment plans.”

Concurrent to this bizarre non-announcement, CNET published an interview with Sinofsky that quoted an earlier internal post in which he had defended his secretive policies.

“I know many folks think that this type of corporate ‘clamp down’ on disclosure is ‘old school’ and that in the age of corporate transparency we should be open all the time,” Sinofsky had written. “Corporations are not really transparent. Corporations are translucent. All organizations have things that are visible and things that are not.”

“We want to make sure that when we do share information, that the information we share is accurate and reliable, and that we have in place the mechanisms for feedback such that the feedback is really taken seriously with respect to our plans,” Sinofsky told CNET. “The reactions that we’ve had to some of the lessons learned in Windows Vista are really playing into our strategy of getting together a great plan for Windows 7, and working with all the partners in the ecosystem in a very deliberate way, such that the end result is a very positive experience for all of us … For the enthusiasts, who are really excited about Windows, well, first, I share their enthusiasm. And second, we’re really going to focus on making sure that when we talk about the product, that they’re getting information that is really what we’re doing for the product.”

Sinofsky noted that “three years” was the right amount of time for any Windows release and confirmed his plans to ship Windows 7 three years after the general availability of Windows Vista. He cast shade on his predecessor repeatedly while pretending to not want to rehash the past. And he deflected several attempts to learn specifics, turning the interview into much ado about nothing. In short, he took a long time to say nothing at all, a communication style we would all soon become very familiar with.

In May 2008, Microsoft revealed the unexpected: Windows Vista had sold 150 million units since its release in late 2006, and it was now selling faster than its well-respected predecessor, Windows XP, had in a similar time frame. This would not do, so Sinofsky OK’d a first public demo of a new Windows 7 technology, with lieutenant Julie Larson Greene showing up during a Bill Gates and Steve Ballmer appearance at the All Things D conference to show off its multi-touch support; Gates was leaving Microsoft that summer to focus on philanthropy.

“Windows 7 multi-touch is a chance to leapfrog the competition, and it builds on work the company pioneered in earlier versions of Windows and in its Surface smart table,” I wrote at the time; remember that the iPhone had been released the year before. “And as Microsoft CEO Steve Ballmer noted, while Apple gets a lot of press for its work, that company plays in a much smaller market. ‘We’ll sell 290 million PCs this year and Apple will sell 10 million PCs,’ he said in a ‘do the math’ moment. ‘They’re fantastically successful, but so are we’.”

In any event, Larson-Green showed an electronic version of finger painting called Touchable as well as touch-enabled photo organization and mapping applications. She also showed off a virtual piano that played music as she tapped the onscreen keys.

“We’re at an interesting junction,” Gates said when the demo concluded. “In the next few years, the roles of speech, gesture, vision, ink, all of those will become huge. For the person at home and the person at work, that interaction will change dramatically.”

I was less impressed than the non-technical hosts of All Things D.

“The problem is that multi-touch is an evolution of the interfaces that came before it, and one that augments, rather than replaces, its predecessors,” I opined. “It’s an evolution, not a revolution, and it’s not a one-to-one replacement, because many tasks are simply easier to perform with more traditional interfaces. And don’t even get me started on the smudge-tastic nature of touch screens.” The future of computing, I declared, was voice, not touch.

On August 14, I received an email message from Steven Sinofsky.

“Howdy,” it began. “We haven’t been in touch directly in a while, but I wanted to give you a heads up about a new blog that we are starting this week.”

A while? The last time we were in touch, Sinofsky had threatened to send me a cease-and-desist order because I had leaked information about the product that had become Office 2007.

“As we lead up to a series of events this fall that will feature Windows 7, we thought it would be fun for us to start a blog about how we make Windows 7. The blog will be hosted (and written) by myself and Jon DeVaan and will focus on the overall engineering aspects of building Windows 7.  Think of it as a companion to the overall external communications about Windows 7. I ask that you not discuss or write about this till the blog posts since we aren’t really letting folks know in advance and we would not want to offend anyone. The blog will be hosted on http://blogs.msdn.com. –Steven Sinofsky”

That blog would go live just a few hours later, and to my knowledge—I did ask around—he hadn’t provided anyone else with a heads-up like that.

But no matter. Sinofsky was about to start blogging about Windows 7.