Programming Windows: Reset (Premium)

On August 27, 2004, Microsoft group vice president Jim Allchin told employees that the company still planned to ship Longhorn to customers in calendar year 2006. But to meet that schedule, it would make a major change: WinFS, the new Windows storage subsystem, would ship sometime after Longhorn, so it was no longer part of the product. Even more dramatically, Microsoft would make the other two major Longhorn technological pillars, the Avalon presentation layer and the Indigo communications system, available to users running Windows XP and Windows Server 2003.

“Allowing developers worldwide to target this existing installed base will create huge new opportunities for them and enable exciting new experiences for hundreds of millions of PC users,” Allchin explained. But left unsaid was a shocking reality from which the Windows team was still reeling: Longhorn as originally envisioned would not come to market and Microsoft had “reset” development of the product using Windows Server 2003 with Service Pack 1 (SP1) as its new base.

“We will not cut corners on product excellence,” Mr. Allchin’s email continued. “Our powerful vision is intact; our shipment plan changes will let customers get access to parts of the vision faster. With the decisions we are announcing today, I believe we are on a strong path forward to deliver an awesome Longhorn product that will provide incredible value to our customers, partners, developers, and shareholders.”

As one of the few people at Microsoft who fully understood the problems with Longhorn, it’s unlikely Allchin believed this. But what choice did he have? All Microsoft could do was start over and take ownership of its problems. And knowing that the Allchin email would leak, Microsoft took the additional step of publicly announcing the changes, again without mentioning that it had, in fact, scrapped all of its previous work and started over.

“We’ve heard loud and clear from customers that they want improved productivity, easier deployment, increased reliability, and enhanced security, as well as the many innovations we’ve been working on,” a quote attributed to Mr. Allchin reads. “We’ve had to make some trade-offs to deliver the features corporate customers, consumers, and OEMs are asking for in a reasonable time frame. Our long-term vision for the Windows platform remains the same.”

The press reacted predictably, noting that Microsoft had originally promised to release Longhorn in 2004. And while Microsoft tried to blame the distraction of Windows XP Service Pack 2 (SP2) on the delays, few believed that a company of that size couldn’t have advanced both products simultaneously.

But this led to some tense moments during a news cycle in which Microsoft was trying to spin the delay as a positive. For starters, some critics began referring to the product as “Shorthorn.”

And when CNET suggested, correctly, that Microsoft would have “rewrite” Longhorn, Bill Gates lied, telling the publication that “there’s no rewrite going on here … The plan we have does give up WinFS shipping with Longhorn. And so if you want my basic assessment here, the glass is three-quarters full.”

“My goal is to have Longhorn the highest-quality OS we’ve ever shipped,” Jim Allchin told the same publication. “At one level, you could say, ‘I’ve had enough,’ and so we’re on a path to drive up the quality level.” It was a much more honest assessment, though Allchin bristled at the “Shorthorn” name.

“The Longhorn mess points to a recurring problem at Microsoft: hype often gets way ahead of the company’s computer science,” Business Week opined. “With Longhorn, as with other past products, Microsoft talked up a technology it couldn’t quite deliver.”

“Longhorn is no longer the ‘big bang’ update Microsoft had promoted,” IDG News wrote.

“No Microsoft product has ever been delayed as much as Longhorn, and as its ship date slipped from 2004 to 2005 to 2006, it became less exciting to users and more the object of ridicule,” I wrote of the news. “According to Lead Product Manager Greg Sullivan, the changes came about because of an internal review of where Longhorn was and where it was going.”

“Basically, coming off of Windows XP SP2, it was a logical time to take stock and evaluate the Longhorn project,” Sullivan told me. “We needed to establish priorities, and see where we were. As a company, we learned a lot from SP2, and in conversations with our partners around that, and at the PDC with the developer community, we had a good handle on what they were expecting. We mapped out a very ambitious plan for Longhorn at the PDC, and we are absolutely on track to deliver on it. How we will get there, however has changed a little bit.

Many were confused by Microsoft’s decision to bring Avalon and Indigo to Windows XP and Server 2003. Did this mean that the vaunted Longhorn user interface would also make its way down level? No, I was told: Windows XP would still look like Windows XP after Avalon is installed; but XP users would be able to take advantage of Avalon-based applications and services. To get the advanced Longhorn user interface, you would still need to upgrade to Longhorn.

How the removal of WinFS would impact Longhorn was another source of confusion. Did this mean that Longhorn would not include advanced search functionality?

“Longhorn will still feature a very rich search experience,” Sullivan answered. “It’s kind of a mistake to equate local search with WinFS. They are different platform elements. So local search will still happen, and we will still deliver a very compelling full text search in Longhorn, alongside new shell capabilities. This will provide some semblance of the ‘Find My Stuff’ experience. But it won’t be the full relational store with deep integration and platform exposure via APIs. For end users, however, it will be the same. We will offer a very compelling user experience for local search inside of Longhorn.”

Perhaps not surprisingly, things went dark on the Longhorn front for the next few months. Microsoft talked up some new driver work at the Intel Developer Forum (IDF) in September, but there was little else in the way of hard news.

And then the leaks started.

In October, a source told me that Microsoft was planning to ship the long-awaited Longhorn Beta 1 release in February 2005.

“According to internal documentation I’ve reviewed recently, Microsoft is planning an aggressive schedule for its oft-delayed Longhorn successor to Windows XP and will ship the long-awaited Beta 1 release of that operating system on February 16,” I wrote. “According to the latest release schedule, Microsoft will begin locking down the code for Beta 1 in early December 2004 and then enter the so-called escrow phase in early January.”

And later that month, I was told that Microsoft planned to dramatically expand the number of Longhorn SKUs, or product editions. There would be at least two offerings for consumers, then called Longhorn Home and Longhorn Media Center Edition, at least two Business offerings, then called Business and Tablet, and a Starter Edition. And aside from Starter, which would be 32-bit only, each would ship with both 32-bit and 64-bit versions of the installer CD.

“There will also likely be an ‘uber’ SKU that bridges consumer and business lines and is the superset of all SKUs,” the source told me. I was unable to publish any information about the SKUs at the time because my source was one of only 25 people at Microsoft aware of these changes at the time.

That source also told me that the Itanium version of Longhorn was “dead.” But there was good news, too: after a rough few months, the Windows team was building new versions of Longhorn every day again. And because of ongoing componentization efforts, they even experimented with a version of Longhorn that had no Explorer shell and was highly reliable.

(Historical sidenote: this would later surface as Server Core in Longhorn Server.)

In November, Microsoft issued a first Community Technical Preview (CTP) version of Avalon for Windows XP and Server 2003, showing developers what they could expect from this technology on existing, pre-Longhorn systems. And Don Box and Chris Anderson, who had done such a terrific job of showing off Avalon to developers at PDC 2003, published a similar but shorter video demonstrating Avalon for XP/2003.

But 2004 ended without any new external builds. And, perhaps not surprisingly, that was related to more quality issues, and so by January 2005, the schedule had changed yet again. Now, a reliable source provided me with a fairly complete Longhorn rollout schedule, albeit one that was doomed, like the others before it:

• Longhorn Beta 1 code complete March 16, 2005
• Longhorn Beta 1 internal release and domain rollout: April 2005
• Longhorn Beta 2 (and product) code complete: July 1, 2005
• Longhorn Beta 2 internal release: Q3 2005
• Longhorn Beta 2 public release: September 2005 (PDC 2005)
• Longhorn Release Candidate 0 (RC0) internal release: Q4 2005
• Longhorn Release Candidate 1 (RC1) internal release: March 2006 Longhorn release to manufacturing (RTM): May 2006

I also used this opportunity to publish the SKU details I mention above, with a few changes: Longhorn Media Center Edition was now called Longhorn Premium, and Microsoft was considering a Small Business Edition. This news was widely republished by other tech news organizations, and in one infamous example, an analyst claimed that Microsoft would never ship so many versions of Longhorn in an attempt to refute my report. Suffice to say that my source was well-placed, and I had no doubts. Also, I was later proven right on the SKU expansion.

In February, Jim Allchin chatted with Microsoft’s MVP community on what was coincidentally the original date for the Longhorn Beta 1 release. Predictably, no one asked him any difficult questions, and Mr. Allchin didn’t provide any new information on any topic. I’m surprised no one asked him about Beta 1 since that release, along with the rest of the Longhorn roadmap, had slipped yet again, as expected.

In early March, I receive an internal presentation and reported that Microsoft now expected to deliver Longhorn Beta 1 by mid-2005, with the final version expected by mid-2006.

“In order to build excitement about Longhorn, Microsoft will disclose information about this next-generation operating system in stages,” I wrote. “First, it will use the Windows Hardware Engineering (WinHEC) conference to prepare the market, and discuss the ‘essence of Longhorn’ as it now stands. WinHEC attendees will receive a pre-Beta 1 build of the operating system.”

“Then, Microsoft will utilize a disclosure approach it calls ‘rolling thunder,’ which will build up to a crescendo by the Longhorn launch. The company hopes to position Longhorn as a major, must-have upgrade for both business and consumer customers. It will build excitement with enthusiast consumers starting with Beta 2.” Beta 2 was now expected in the second half of 2005.

I also revealed some of the key features that Longhorn would include, some of which hadn’t been disclosed previously. These included a least a privileged user account (LUA) that evolved into User Account Control (UAC), a new hot patching capability for applying updates without needing to reboot, and a new instant-on feature that would see Longhorn PCs resume from Standby in two seconds or less. It would feature anti-spyware/malware capabilities (but not anti-virus, curiously), a new file system encryption feature, and a Fast Search feature that would work locally, across the network, and via the Internet.

And I revealed Longhorn’s hardware requirements:

Desktop CPU: 3 GHz Intel Pentium 4 processor with Hyper-Threading Technology 530 (or higher) or 3 GHz Intel Xeon processor with 2 MB L2 cache, or AMD Athlon 64, Sempron, or Opteron 100, 200, or 800 processor, single or dual-core versions.

Mobile CPU: 1.86 GHz Intel Pentium M processor 750 (or higher), or AMD Turion 64 Mobile Technology, Mobile Sempron, or Mobile Athlon 64 processor.

RAM: 512 MB of RAM or more, all platforms.

In April, ahead of WinHEC, Microsoft came through on its “rolling thunder” strategy and sent Jim Allchin around the United States to discuss Longhorn with the press.

He told CRN that Beta 1 was now expected in “early summer” and that attendees could expect a new build; it would be the first external build since 4074 from one year earlier. He claimed it would be “dramatically different” from the first preview but that it would still not include the new Aero user interface. He hinted that Beta 2 would arrive at PDC 2005 later that year.

PC World asked Allchin about the SKU information I had published, but he declined to discuss marketing. “Final naming for Longhorn, which is a code name, has not been decided yet,” Allchin said. But the final release date had slipped yet again, to in December 2006.

That week, Microsoft provided a WinHEC pre-briefing in which it told me that it would announce the general availability of what were now called Windows XP Professional x64 Edition and Windows Server 2003 x64 Edition. And that Bill Gates would promote the increased security, greater reliability, and faster performance enabled by 64-bit computing, which would become mainstream in Longhorn. As expected, attendees would receive a “developer preview” of Longhorn.

Perhaps predictably, that build leaked the night before WinHEC started on April 24, just as the PDC 2003 build had leaked a day early six months before. It was build 5048.

(Historical sidenote: Previous build numbers had been in the low 4000s. The change was related to the reset: when Microsoft restarted Longhorn development based on Windows Server 2003 in August 2004, it pushed the build numbers forward.)

More problematically, Longhorn build 5048 was a disaster.

“Longhorn build 5048 actually represents a usability back-step from last year’s build 4074,” I wrote at the time. “That’s because some features, like the Sidebar and the new system-wide Contacts utility, are missing in action in 5048. There are reasons for these omissions. None of them are particularly good.”

“I have to be honest here. After a year without a single new Longhorn build and very little concrete information about what was going on with the project, I had high expectations for build 5048. And a pre-WinHEC briefing with the software giant did nothing to assuage those hopes. Plus, I’ve seen advanced Longhorn UI work and I knew how cool this thing was going to be.”

“Build 5048 communicates none of that. And that’s a shame, because Microsoft had a chance to ramp up the momentum of a product that, quite frankly, could use a little momentum. Longhorn build 5048 is a disappointment.”

Longhorn build 5048 communicated itself as Windows 2006 Professional and Windows version 5.5, which was interesting but in no way predicted how Microsoft would later market the product. Worse, Microsoft found itself at the center of another Longhorn-related controversy, and this was one they should have predicted and worked proactively to prevent.

If it’s not clear why this was disappointing, remember what Microsoft originally promised for Longhorn:

And then the waiting game was on, again.

In late June, Microsoft announced that it would support a technology called Real Simple Syndication (RSS) in Longhorn. RSS is an XML data format that’s used to publish web content to which people can subscribe, and it appeared to solve a problem that Microsoft and Netscape had tried to solve a decade earlier with so-called “push” technologies: rather than force people to manually navigate to web sites to see when content has been updated, they could subscribe to that content and receive notifications when updates were made. As part of this push, it would update Internet Explorer 7 to include RSS subscription capabilities in addition to integrating RSS directly into Longhorn.

“Like web searching, RSS is an additive, not a replacement, for web browsing,” Microsoft director of product management Gary Schare told me ahead of the announcement. “Web searching doesn’t replace browsing, and neither does [RSS] subscribing. But it will develop into a key way for people to use the web.”

In late June, a source told me that Microsoft was secretly working to improve the Longhorn shell via something called Project M. Project M was headed up by Chris Jones, now vice president of the Windows shell, who reported directly to senior vice president Will Poole. Intriguingly, Hillel Cooperman of PDC 2003 Longhorn demo fame was now working for Jones. “This is very secret,” I was told, “with the changes expected in Beta 2 or RC0.” My understanding is that Project M turned into the last-minute glass icon additions that Microsoft made right before shipping the final version of the product in late 2006.

Microsoft missed its Beta 1 release date—of course it did—but beta invites started going out to testers in early July. “We are pleased to offer you an early preview of Windows, Code-Name ‘Longhorn,’ by extending this invitation to join the Longhorn beta program,” the invitation noted. “Longhorn promises to be the most secure and intuitive Windows release to date.”

And then things got weird. And to understand this weirdness, one needs to understand how Microsoft was organizing its Longhorn builds and build numbers. As noted earlier, all post-reset Longhorn builds were in the 5000 range. But Beta 1 builds were supposed to be in the 5100s and Beta 2 builds—which Microsoft was working on concurrently at that time—were in the 5200s. And so it was a bit weird when Longhorn build 5203 leaked in early July ahead of a pre-Beta 1 build, 5098.

But whatever: build 5098 was a significant step forward from build 5048, and it righted many of the wrongs of the past by simply including the Aero glass user interface by default, at least for those with hardware graphics. “Windows feature a polished, glass-like translucent look,and the window buttons light up as you mouse over them,” I wrote. “One nice touch: the Recycle Bin visually appears to fill up with little crumpled balls of paper as you throw items out.”

There were lots of enhancements in this build. Icons and other on-screen elements could scale to sizes never before seen in Windows. Virtual folders were thoroughly integrated into the file system. Key user interfaces like Control Panel were significantly updated.

There were new experiences, like Games, and a new browser, Internet Explorer 7. There was even a new Windows Backup program, the first time this NT-era tech had been updated.

Beta 1 also included our first peek at User Account Control (UAC), a controversial new security technology that was formerly called Least Privileged User (LUA).

“In Beta 1 it is off by default, but you can enable it with a shortcut in the Start menu,” I wrote. “I’m told it will be on for good in future builds. If my experience with Beta 1 is any indication, many Windows users are going to find this change very difficult, and much more aggravating than any of the security changes Microsoft added to Windows XP SP2. Maybe it will get simpler over time.”

Maybe. But Microsoft would soon face yet another Longhorn controversy: In late July, it would announce to the world that it had finally settled on the final name of the product.

And it was terrible.