Programming Windows: More Alpha Tales (Premium)

The Longhorn leaks and rumors continued into 2003. But the longer the schedule was pushed out, the more excited enthusiasts became. Longhorn, suddenly, wasn’t just a minor stopover between Whistler (Windows XP) and Blackcomb. It was everything.

The Wall Street Journal got things started in 2003, fittingly enough, the week of the New Year, speculating whether Longhorn would “take the blue ribbon or get slaughtered.” The rationale for this question was solid enough: with an ever-creeping feature set and non-stop delays, Longhorn was starting to look a lot like a previous and failed Microsoft project called Cairo.

Cairo, described by Bill Gates as both “a paradigm shift” and “a holy grail,” had “imploded under its own weight,” according to one analyst in an eerie prognostication of what would later happen to Longhorn too. Not helping matters, Microsoft’s Longhorn efforts were led by the same man, Jim Allchin, who had previously spearheaded Cairo. And while even Allchin admitted to the parallels, he told the publication that Microsoft “didn’t have the technology we needed to pull off [Cairo], but now we think we do. We maybe have a little more wisdom.”

“Longhorn is a delicate balancing act,” the Wall Street Journal report explained to its readers. “If Microsoft makes only minor changes to the operating system, few people will buy it. But if it makes big improvements, it will have to take more and more time to tinker with increasingly complex Windows code, raising the possibility of errors and delays, and some customers still might not bite.”

At the Consumer Electronics Show (CES) the following week, Bill Gates delivered another keynote address, but he only mentioned Longhorn once, and in passing. Instead, he focused this time on coming Microsoft consumer platforms like Media2Go—which would later be rebranded as Portable Media Center (PMC)—and Smart Personal Object Technology (SPOT), a Windows CE .NET-based system for “everyday devices” such as alarm clocks, wristwatches, key chains, and even refrigerator magnets. A SPOT watch would later appear briefly, backed by an online service called MSN Direct, but that would be the end of that.

In February, Bill Gates and his executive staff was suddenly ready to talk Longhorn, and he admitted to The Seattle Times that Longhorn was indeed a big release, and not a minor stopover.

“It’s a phenomenal step forward and very ambitious,” he said. It’s “my big-time focus now.”

“There is a lot more in Longhorn that should be exciting to software developers,” CEO Steve Ballmer added, comparing the coming release to Windows XP, which didn’t offer much new to developers.

“Longhorn will be the first release we’ve done since Windows 95 for the client where we’re able to do basically three things: (create) end-user excitement, developer excitement and new excitement for the hardware vendors,” Longhorn development lead Chris Jones told the publication. He added that Microsoft’s investments in Longhorn fell into one of two “buckets”: basic improvements to its performance, reliability, security, and privacy protections, and new ways to use a PC, including new applications Microsoft was developing and efforts to help other companies build new programs for Longhorn-powered computers.

And the leaks had started back up too.

In very late January, I had obtained a leaked version of Longhorn build 4001 that, among other things, verified previous leaks thanks to its inclusion of the infamous Longhorn “hay roll” wallpaper.

And on March 1, Longhorn build 4008 leaked. I wrote an extensive review, though very little had changed since my late 2002 Longhorn alpha write-up.

“Longhorn build 4008 shows numerous evolutionary, rather than revolutionary, improvements over build 3683, which was the focus of my Longhorn alpha preview,” I wrote. “This tells me two things: first, that the most exciting bits in Longhorn, including the new 3D video-based user interface, the SQL Server-based file system, and a new breed of exciting digital media-oriented applications and services, have yet to be introduced into the core Longhorn code, creating a bizarre pseudo-update to XP with an ugly, seemingly unfocused user interface. Second, that there’s a bit of subterfuge going on here. Looking over build 4008, I’m constantly struck by the fact that this leak seems somewhat convenient. I almost suspect it was released on purpose, to throw fans off of the real work that’s coming down the road. Whether this second point is real or imagined matters little: if you think the final release of Longhorn will even slightly resemble these early alpha builds, then you have a pleasant surprise coming.”

Well, that much was true at least. But Longhorn build 4008 did have a few improvements, including an overhauled and faster setup process, hints at future WinFS-related functionality in Windows Explorer, and numerous visual enhancements.

In mid-April, another Longhorn build, 4015 leaked, and while there were no major revelations, it did have some improvements over build 4008, including a new Welcome experience, new taskbar and sidebar options, new Explorer features, and more WinFS functionality (though it had to be disabled to make the shell usable).

The most interesting new feature, perhaps, was something called libraries, which I explained was “a virtual folder that intelligently gathers information about files on your system and presents them to the users in a collection.” In other words, where a folder in the file system contained files, a library provided a “view” that consisted of shortcuts to files and folders and presented it as if it were a folder. These libraries would aggregate related content—photos, perhaps, or all of the files related to a specific project—into a single view no matter where that content lived in the file system. The Document Library replaced My Documents, providing a collection of all of the documents on your system; by default, the Document Library collected documents from My Documents, the desktop, and Shared Documents.

Concurrent to this build, I received a teaser for the upcoming WinHEC 2004 conference that promised that Microsoft would show off the “3D graphics enhancements in Windows Longhorn.” These would include a dynamically composed desktop display that would enable compelling new visual effects like “windows tumbling onto the screen, rotating windows, warped windows, alpha blending between windows, threads, and events and other synchronization objects. The new graphics infrastructure of Windows Longhorn will also enable sub-pixel Microsoft ClearType text display technology with anti-aliasing, hardware-accelerated and resolution-independent anti-aliased 2D graphics, rich 3D graphics using a higher-level API, integrated with 2D API and controls infrastructure, glitch-free video playback, greater than 8 bits per component color pipeline, all graphical elements can be arbitrarily combined within applications and across the desktop using the Desktop Compositor, and support for higher pixel density on both LCD and CRT displays.”

Ahead of WinHEC, details about Microsoft’s plans started to leak, often inaccurately. Citing one source, a CRN report claimed that Microsoft would reduce the number of Win32 API calls from over 70,000 to just 10,000, “to help developers better exploit the next-generation Windows shell, user interface (codenamed Aero), and .NET Framework components in Longhorn, according to sources familiar with the Longhorn plans.” (The new Longhorn APIs would, in fact, be separate from Win32 and would only add to the API count.)

CRN also reported that Microsoft would replace the Windows GDI with Avalon Longhorn, which would “replace the need to do manual coding with prebuilt, extensible XAML scripts … XAML has new metatags and extensible schemas for user-interface structures and behaviors that are designed to simplify and increase the customization of the ‘jazzed up and 3-D oriented Longhorn GUI, code-named Aero’,” the publication reported, correctly. A beta version of Longhorn was “due to be launched at the Microsoft Professional Developers Conference this fall.”

Microsoft finally provided an incredible and transparent preview of Longhorn at WinHEC, which was held in May that year in steamy New Orleans. Microsoft senior vice president Will Poole kicked things off with a keynote address in which he went public with the then-current plans for the Windows release roadmap. And it wouldn’t happen quickly.

As everybody knows, Longhorn is the big goal for us from an operating system perspective that we are putting all of our efforts behind,” he said. “This is a huge, big, bet-the-company move, and it’s one that we are very enthusiastic about what we’re able to do here. The breakthrough work that we’re going to do in ‘Longhorn’ is going to really change the landscape of what consumers, what businesspeople see when they look at a new PC.”

“So the road between now and Longhorn is not super short. We’ve got some work to do. It’s going to take us a while to get there. And what you’ll see is there are a couple of major milestones, a couple of big road signs there. We went through the developer preview in March. We got some great feedback to understand what people want to see from a pure development perspective in the Longhorn platform. And the next major milestone from a developer perspective will be in October of this year with our Professional Developers Conference … Over the course of 2004 you’ll see a couple of releases in the betas for Longhorn and we’ll see that coming to market in 2005.”

What Poole had just admitted was that Microsoft, the firm that had shipped Windows XP about a year after Windows 2000, would need four years to complete Longhorn, a product that was originally envisioned as a minor release. But Poole also addressed what would happen in the interim: Microsoft would ship updates to Windows XP Media Center Edition and Windows XP Tablet PC Edition.

Earlier in his keynote, Poole had eased the audience into the coming 3D revolution by showing off a videogame called Unreal Tournament 2003. Running on Microsoft’s DirectX 9 technology, Unreal Tournament 2003 offered stunning, hardware-accelerated 3D graphics. But Poole noted that this technology could be used throughout Windows as well.

“In the Longhorn timeframe, you’ll see us drive this level of immersion and this level of fidelity right into the base level OS,” he said. “And from there what we really need is gaming graphics level cards shipping in non-gaming SKUs in the Longhorn timeframe. You’ll see some examples when you look at some of the combined DirectX and user interface and hardware work that we’re doing in various sessions here that all will fold into Longhorn.”

Poole said that Longhorn would deliver on a concept called Life Immersion, which would “embrace human factors like [Microsoft has] never done before, to really understand how to make that emotional connection to [its] customers to address all of the product requirements that I talked about, making it just work, making it something that you can invite and live with every day in new and profound ways from a technological perspective both at the hardware and with the software to deliver that immersive experience.” This was vague at the time, but by the October PDC, Microsoft’s goals there would become clearer.

Longhorn would also be completely componentized, a big change from the previous “thing on a thing” development approach, where new features were simply heaped on top of the previous code-base with each new product revision.

Taking a cue from its embedded versions of Windows, Longhorn would be logically broken down into several building blocks. At the lowest level was Microsoft’s Mark Myers called the “base OS” component, which was about 95 percent of the total Longhorn codebase. The base OS component would be completely language independent and was a subset of all of the Longhorn product editions that Microsoft would create.

Longhorn product editions, or SKUs, would be built on top of that base OS component and on top of each other. To create SKU #1, for example, which was called XP Home Edition at the time, Microsoft would add a set of components to the base OS. And to create Longhorn SKU #2—what we called XP Professional Edition at the time—it would add more components to SKU #1.

This componentization would also lead to big changes in how Longhorn was deployed. It would boot up and install much more quickly, something we’d already seen in the leaked builds, and it could be serviced more easily.

Of course, I was mostly interested in the promised new graphical effects and here I was not disappointed: the Longhorn Desktop Composition Engine, or DCE, would support wide aspect ratio and high-density LCD displays (120 dots-per-inch, or dpi, and up), and advanced hardware-accelerated 3D graphics processors. It would have some heady requirements, at least compared to Windows XP. At a minimum, a Longhorn system would feature a 1024 x 768 display with 32-bit color and hardware-accelerated 3D video.

Kerry Hammil, a Microsoft program manager with the Windows team, Client Platform, provided a variety of demos showing spinning transparent windows with no clipping, and how high DPI displays and a new generation of scalable application UIs, would change everything in Longhorn. It was exciting stuff.

In an interesting twist, all of the Longhorn demonstrations Microsoft provided at WinHEC were performed on the M5 milestone, build 4015, which I had previewed weeks ago. But the leaked build didn’t have any of these effects, so I asked why that was. Microsoft explained that the leaked build had come out of Microsoft’s main build lab, while only those builds from Lab 6—home of the Avalon graphics team—had the DCE bits installed. It was, in its own way, a unique example of Longhorn’s componentization efforts. (And, unknowingly, an early example of why Longhorn, too, would crumble under its own weight a few years later.)

Microsoft did not reveal the new Aero user interface at WinHEC, with one source telling me that it just wasn’t “ready for primetime yet.” And the firm clarified its WinFS efforts, noting that the underlying file system in Longhorn would remain NTFS. WinFS was not a file system, I was told. Instead, It was a service that would run on top of—and require—NTFS. When enabled, file letters would be hidden from the end-user, though they would still lurk there under the covers for compatibility with legacy applications.

In late July, Microsoft again hosted its annual Financial Analyst Meeting. And once again, Bill Gates and the executive staff teased Longhorn. And once again, the scope of Longhorn expanded.

“Longhorn is the next generation [of Windows],” Gates said. “It’s a big bet for us. We don’t know the exact time frame of it. It’s clearly many years of work that we’re engaging in. We’re very excited about the prototypes we’ve built, and some of the early technology proofs that we have. We’ve got a major advance in the user interface, a major advance in the API. It has this new storage capability, it’s got web services as sort of a built-in piece of the platform, and it’s very oriented around scenarios, making it easy to manage workflows, making it easy to manage contacts. For our customers, taking the idea of the way you deal with photos and music and just unifying those into the single storage metaphor. Longhorn is not just a release of the Windows client, it’s also a release wherein the same time frame you’ll have advances in Office, our server products, virtually everything at Microsoft is synchronized to build on this platform and take advantage of that.”

Windows lead Jim Allchin followed Gates, and while much of his talk focused on coming iterations of Windows XP, he also spoke of steady progress with Longhorn despite the years of time that had already elapsed in its creation. But he, too, slyly added alluded to another setback: Beta 1 would not be delivered at PDC. It would arrive the following year.

“The next step for Longhorn, we’ve had some design previews already,” he said. “We’ve shared with partners as well as ISVs and IHVs what we’re doing. We do have regular builds running of this. The next big event is PDC, which is our Professional Developer Conference, which is coming up at the end of October. We will be handing out CDs there to get the developers going, to let them see the great capabilities that we’re putting into Windows. And then later next year we will be doing Beta 1, which will be a broad beta of the technology of Windows code-named Longhorn. As Bill said, there are other pieces coming in the Longhorn wave but I’m just talking about the client here. So that’s it. It’s a great future. We will sustain the momentum in Windows and the future; we’re deeply invested in it and it looks incredibly bright.”

In his concluding remarks, Microsoft CEO Steve Ballmer tried to justify the delays.

“In the Longhorn case we’re sort of taking this to the next level because we’re absolutely trying to think about not just the next generation of Windows but the next generation of a whole series of products,” he said, noting that Microsoft was finally starting of its platform as being “bigger than Windows.” “Some people say, ‘That sounds like the highest-risk thing I’ve ever heard, Steve, cut it out. Why don’t you dribble-drabble some things out here?’ But that’s not how this industry is going to continue to be transformed.”

“We think there needs to be, not every year, but there need to be periodic big bangs both for our company and for our industry in order to rejuvenate the innovation cycle, and that’s how we think about Longhorn, as one of those fundamental big bangs.”

“People say, ‘Oh, it sounds scary. That means you might not ship it on date foo or bar or whatever the heck date you think it’s going to be.’ And I say, you’re right. That’s why we don’t try to talk about the date. We’re trying to talk about the conceptualization. We’re going to do our best; we’ve got teams working hard, great plans, good focus, but having a breakthrough and having scenarios that are integrated around that we think is Job One. Job Two is we need to be a platform that creates opportunities for other developers—independent software companies as well as custom developers—because no matter how good our integrated innovation, we’re not going to have all the answers.”

Concurrent with Ballmer’s comments about what the company was now calling a “Longhorn wave” internally, Microsoft admitted that it would, in fact, ship a Longhorn Server concurrently with the Longhorn client. The rationale was interesting: Microsoft typically released a new version of Windows Server every three years, and with Longhorn delayed again and again, the schedule was lining up. If temporarily.

Microsoft also started fielding questions about whether it would phase out .NET as a brand, a topic that would come up again and again as it provided more details about its Longhorn API plans.

“We’re trying to clarify some of our positioning around branding,” Microsoft senior vice president Eric Rudder noted in an interview that August. “When we first launched Word and Excel, we used to be very adamant about calling it Word for Windows and Excel for Windows. Eventually, people understood, ‘Oh, there’s this new platform Windows,’ and they started calling it Word and Excel and, of course, and eventually Office. With the initial .NET platform enthusiasm, we made sure that everybody saw the .NET name all the time to make sure that they understood that it was really about the new platform. It was important, for example, in the developer space to really call Visual Studio ‘Visual Studio .NET’ to make it clear that ‘Hey, this was the tools set for delivering .Net applications.’ But as people understand that .NET really is Microsoft’s platform message, when people say the word Visual Studio now, I think we’ve sort of implied that it’s Visual Studio for .NET. As that clarity comes, I think we probably won’t use .NET in as many suffix ways as we have in the past.”

In August, I finally obtained some official shots of the (non-glass) Aero UI that an engineer had used in a slide deck during a WinHEC talk the previous May, but had left them out of the publicly distributed version. These shots included depictions of a new Sidebar-based volume control, a Hardware & Devices control panel, a Device Properties page, and more.

And then in late August, Longhorn build 4029 leaked. I didn’t get my hands on it until almost a month later, but there wasn’t much in the way of new features.

October finally arrived, and all eyes turned to Los Angeles and the coming PDC. For Windows enthusiasts, PDC 2003 would be the highest of highs. Until, of course, it was revealed to be the lowest of lows, and the beginning of the end for Windows as the epicenter of personal computing.

More soon.