Programming Windows: It’s Not Going to Work (Premium)

We should have seen it coming.

In the days leading up to the Professional Developers Conference (PDC) 2003, we received yet another warning sign that all was not well with Longhorn. Build 4051 had leaked online two days before the show, giving us a first and early peek at the Developer Preview. And … it was disappointing. Really disappointing.

It was immediately clear that Longhorn build 4051 was still a very early and incomplete build of the coming operating system. Indeed, Jim Allchin would later claim that this was the earliest in a release cycle that Microsoft had ever given such a thing to the public. But that statement, too, should have triggered red flags, as Longhorn had been in active development for over two years by that point. And Microsoft still couldn’t even deliver a proper beta to outside developers?

Build 4051 featured yet another Windows XP-like user interface, with an ugly new Slate theme replacing the Plex theme we’d seen earlier and no hints at all at the coming Aero UI.

“At the risk of sounding a bit negative, Longhorn build 4051 is, sorry, boring,” I wrote. “Once you get over the mildly amusing Slate theme, and the slow and painful hardware detection, it’s basically the same as previous alpha builds, albeit in more usable form. Explorer windows feature the Aero-like look that I first revealed [previously] and… well, that’s about it. There really isn’t that much more to say, at least not yet. We’ve installed 4051 on three machines so far and we’re not that impressed.”

Still, a coworker and I took over 300 screenshots in anticipation of future galleries and articles. And we spent much of our time at PDC alternating between keynotes, briefings, and sessions, and playing with the new build. Unfortunately, compared to the exciting Hillel Cooperman keynote demos, in particular, it never hinted at the promises of what Microsoft intended to deliver.

“We’ve spent several hours playing with Longhorn build 4051, and while we’re still not overly impressed, we have at least gotten the gist of what’s going on in this build,” I wrote during the show. “First, it’s a dog on any system with less than 512 MB of RAM, so consider that a base amount (up from 256 in Windows XP). The new content aggregator Libraries [virtual folders] are more usable in this build than in previous alphas, and it’s clear that a lot of what we’re seeing now are just placeholders for future refinements.”

By November, Microsoft was referring to build 4051 as “pre-alpha,” a nonsensical term, to further distance it from the first beta build, which it now expected to deliver in the second half of 2004, a full year from then.

In January 2004, I interviewed Hillel Cooperman and his user experience partner in crime, Tjeerd Hoek. (I will be publishing the full interview soon.) I was surprised to discover that Cooperman had worked for Microsoft for several years—I had assumed he was a recent hire—and that he had joined the Windows team to work on a version that never shipped called Neptune alongside Joe Belfiore.

“I don’t want to over-promise, but we have big dreams,” he told me. “We’re really trying to have a level of attention to detail that surpasses anything out there. And we think we’re resourced appropriately for it, and it’s really just a matter of time. We’re not going to get to everything, but we’re going to do a bunch of things we can be really proud of, and that customers can respond to. We’re working really hard on it. The thing that’s heavy lifting for us is that we don’t just have these user experience goals in a vacuum. If our goals around Longhorn were just about improving the user experience, we’d be done a lot sooner. There’s an ecosystem to involve first, and to aspire with different innovations, inventions, and platforms. But I’m glad [Longhorn] isn’t just [a new user experience].”

Because Longhorn’s special effects would require more powerful hardware than was found in most PCs of the day, Microsoft began promoting a three-tier user experience model. In the simplest mode, Longhorn would emulate the Windows 2000 user experience, complete with the classic Windows shell. This mode would be provided solely for compatibility reasons only, and for enterprises that didn’t want to retrain employees on the new Longhorn Aero user interface. The other two tiers were called the Tier 1 Experience and the Tier 2 Experience.

Longhorn’s Tier 1 User Experience was “a baseline experience” that all machines with Longhorn-capable hardware could use. It was to feature entirely new visuals, though the rendering complexity of those visuals would technically resemble that of Windows XP but would visually be a scaled-back version of the full Longhorn user experience.

The so-called Tier 2 User Experience was the full meal deal, a superset of Tier 1 with a “graphically stunning user interface” and advanced composition services that would be available to Longhorn-specific applications through a new set of APIs. This version would require the Desktop Compositing Engine (DCE) and advanced 3D hardware support.

Historical sidenote: It’s interesting in retrospect how the Longhorn requirements, spelled out so far in advance, and repeatedly, would come back to haunt Microsoft. When Windows Vista finally shipped over two years later, the biggest customer complaint was that they didn’t have powerful enough hardware to run the Aero user interface with its glass-like transparency effects.

In April 2004, Business Week acquired some internal Microsoft emails that detailed yet another new schedule for Longhorn, which was delayed, again, this time to 2006. And to meet that date, Microsoft would trim some features, including some key storage technologies that allowed WinFS to work with network-based file stores.

“I think we all recognize that we need to turn the corner on Longhorn,” Microsoft vice president Joe Peterson wrote to employees in a March 19 email. “We are going to focus on doing fewer things, and doing them well … “It’s fair to say we’ve had very ambitious dreams that were very large for Longhorn. Some of the things that we do in the big dream, we are going to do in the Blackcomb time frame.” Microsoft would also rework the next version of Microsoft Office, which was originally designed to only work on Longhorn, to also work with current Windows versions as well.

Microsoft went into damage control mode, with lead product manager Greg Sullivan handed to the press like a sacrificial lamb.

“Microsoft is not cutting back on our vision for Longhorn,” he told various media outlets. “We’re determining the work that will enable us to deliver that vision and figuring out what is not core. We don’t have any specific details to share about the current plan or which features are in or out.” But he did admit that the first Longhorn beta had now slipped to early 2005.

“What we told developers at PDC is the essence of Longhorn,” Sullivan said, responding to the obvious complaint that Microsoft had overpromised and would now underdeliver. “We are now determining what the areas are where we can do some shaping around the edges so we do get the product in the hands of customers.”

In May, Microsoft hosted yet another Windows Hardware Engineering Conference (WinHEC) in Seattle, and Bill Gates only mentioned Longhorn once in his keynote address. Oddly, he chose to focus on WinFS.

“That’s a key breakthrough in Longhorn,” he said. “Longhorn is a lot about the fundamentals, the reliability, security, ease of setting things up. It’s a lot about visualization, using the latest graphics, and new interaction techniques. And it’s a lot about these storage breakthroughs, it’s called the WinFS file system that’s built-in there. [Note that WinFS was a file system.] We’re making great progress on this, this is the year that we’ll get really a pretty fully capable version of that out in the hands of developers to give us feedback, and really understand, did we get this breakthrough advance in storage, did we get all the pieces we need there so that software developers and users will see what we want out of that?”

Predictably, Windows lead Jim Allchin had more to say about Longhorn in his own keynote address, and he verified the new 2005 date for Beta 1. He reiterated that Microsoft had switched gears on whether there would even be a Longhorn Server, noting that “we’re building it in sync [with the Longhorn client]. That was a change. Perhaps a year ago that wasn’t our plan, but it is our plan today.”

“Now, what is Longhorn really about?” he asked rhetorically. “We are thinking about it in terms of doing … Now, we are making a huge step in terms of developer productivity in Longhorn. There are several different new pillars that are coming in Longhorn. First is, there’s a new storage system called WinFS. Second, we have a new communications system that is based on web services, called Indigo, which lets you connect out to other systems in a much more unified way. Third is a new graphics capability which we call Avalon, which takes full advantage of the GPU, and manages that. Now, all those APIs are managed, and we call the collection WinFX.”

“Last is about people, the design we’re using is a person-driven design,” he continued. “What do I mean by that? I mean, we have personas of individual people that we’re locked on to. We know about what these people do in their daily lives, we know their personal-professional interaction, we know the environment that they’re in. This is true whether they’re an IT professional dealing with a particular type of servers, or they’re someone who stays at home and they’re a parent dealing with their children.”

Allchin then launched into a series of Longhorn demos, using a new build, 4074, that was handed out to testers as part of a new series of Longhorn Technical Previews that would help it get to Beta 1. He focused, in particular, on how much easier it would be to migrate users from Windows XP to Longhorn. But the most notable event in the demo was when the videogame Quake wouldn’t run properly alongside some playing videos in yet another warning sign.

“I don’t know why it’s not coming up, but Quake is running in the background,” he started. “Ah, it’s finally coming up. This machine is very slow for some reason. Eventually, I think that will come up. We’ll maybe have to come back to it.”

At the show, Microsoft revealed that Longhorn would now have a two-tiered user interface, instead of three, and that Aero would require DirectX 9 graphics with a screen resolution of at least 1024 x 768 (XGA), compared to 640 x 480 (VGA) for today’s Windows versions. Aero Glass would require a DirectX 9-compliant 3D graphics processor with at least 64 MB of RAM, though 128-256 MB of RAM would be recommended.

As for Longhorn build 4074, it came to be known generally as “the WinHEC 2004 build,” though Microsoft described it as the first of a series of Longhorn Community Technical Previews.

“These Community Technical Previews are not Beta releases of Longhorn,” Microsoft explained.
These are early builds that developers will find useful to evaluate, but they are not functionally complete nor do they meet the quality bar of a Beta release. You will likely see both progress and regressions as you look at the functionality in these builds. This is to be expected, but as we advance with the project you should see solid progress toward the vision we outlined at the 2003 Professional Developers Conference … [But] when you review this Community Technical Preview, you will find it looks like the initial preview made available last year. The teams at Microsoft are continuing to focus on the developer functionality at this stage of the project, and consequently, the current user experience with the operating system is unchanged.”

Build 4074 didn’t appear to offer much in the way of new features, though it came with a new and temporary Jade user interface that emulated Aero for low-end displays. But enthusiasts discovered about a month later that Aero glass effects could be enabled, making this build a lot more interesting.

And then things got weird.

There were no more Community Technical Previews to come, which was concerning. And when Microsoft hosted its annual Financial Analysts Meeting in late July, Jim Allchin wasn’t on hand to discuss Longhorn. Instead, senior vice president Will Poole provided the Windows presentation, which focused on momentum and not the future.

“Longhorn, of course, is a very important effort for the company,” he said at the conclusion of his talk. “It’s something that we are looking at both from a business perspective as well as a consumer perspective. It is the next big release of Windows, and for businesses, we really have made sure that we make this thing manageable and deployable. One of the number one things that we hear from our business customers is that we need to make it easier for them to move the advanced technologies out into their businesses … Longhorn is going to take us a long way toward helping realize more of the opportunity we have in both business and consumer markets over the next couple of years.”

Poole then reiterated a 2005 date for Beta 1 as part of the same roadmap slide we’d seen earlier at WinHEC.

The press noticed the lack of specifics and news on the Longhorn front and started asking questions. Some wondered if the ongoing delays could be attributed to Microsoft’s renewed focus on security, which had triggered two major new updates to existing platforms in Windows XP Service Pack 2 and Windows Server 2003 SP1. But we would later learn that Microsoft had reached the breaking point with Longhorn. This project was too big, too unwieldy, and too out of control for Microsoft’s established software development processes. It was teetering on the edge of disaster.

“Longhorn is a pig and I don’t see any solution to this problem,” Jim Allchin had written with characteristic credibility earlier that year in an internal memo that was later made public as part of a court case. “If we are to rise to the challenge of Linux and Apple, we need to start taking the lessons of ‘scenario, simple, fast’ to heart.”

Among the changes he suggested was “a simple, fast storage system” to replace the bloated and inefficient WinFS that Microsoft continued to promote publicly despite its many issues. It was part of a “Plan B” he was already formulating for Longhorn, a reset of the project that would result in a less far-reaching but more realistic product.

By late summer, Allchin finally admitted defeat. He told Bill Gates in a private meeting that “it’s not going to work.” Longhorn wasn’t just running late, it was impossible.

Microsoft, he said, would need to start over.