In-Depth
Vista: One Year Later
After a rough start, will things finally align for Microsoft's flagship OS?
It was November 2006 and the final production bits of the long-awaited Windows Vista operating system were finally available for download. For Microsoft and the Windows faithful, it was a critical month.
After all, Vista was running late. Very, very late. Work on the OS had begun soon after the launch of Windows XP in 2001, but the project known as "Longhorn" was delayed by false starts, troubled development and unfocused management. Five years later, Vista delivered only partially on the early promise of Longhorn.
Now, a year after the "official" Vista launch in January 2007, there are critical lessons for developers about its troubled development and rollout. Those lessons offer a glimpse into the future of the Windows client and can provide useful insight for development shops.
Perhaps the most significant decision: Should development managers start tuning their applications for Vista-specific clients, or have the Internet and broad frameworks like .NET 3.0 made the question largely irrelevant? We look at where Vista has been and where it's going to find out.
Vista's Rough Ride
As the leading component of a two-part, client-and-server OS release, Vista emerged as a slow-motion train wreck that ultimately yielded a large-scale reorganization of the Windows Platform Group and ended the careers of several key Microsoft executives. The question is: Why?
The consensus from observers -- including current and former Microsoft staffers, analysts and partners -- was that Vista suffered from a combination of sky-high expectations and a classic case of too many cooks in the kitchen.
As the Vista project lead, Jim Allchin fielded a lot of input from the top of the org chart, according to both current and former Microsoft insiders.
"Gates and Ballmer had a field day putting everything in there but the kitchen sink," says one former Microsoft exec, referring to Microsoft Chairman Bill Gates and CEO Steve Ballmer. "Who at Microsoft says no to Bill?" he asks.
Pressured to design in new features, the team found itself working with a moving target as specs and features changed based on competitive pressures and executive whims, the sources say.
Overly Ambitious
It didn't help that Microsoft built up expectations at the Professional Developers Conference (PDC) in 2003. The proposed WinFS relational file system was heralded as a sea change in the file-management infrastructure. Three years later, it was gone -- pulled by Microsoft just weeks after it had promised a new WinFS beta build.
It was a blow to developers, in particular, who came out of the PDC in 2003 believing that WinFS would be a strategic underpinning of the Longhorn OS.
"The thing that delayed Vista was the extremely ambitious scope of the initial project," says Paul DeGroot, analyst with Directions on Microsoft, the Kirkland, Wash.-based research firm. He ticks off the now-infamous "pillars" of Longhorn: The aforementioned WinFS file system; "Indigo," the communications subsystem; and the "Avalon" presentation layer.
"These are all very ambitious things, which appeared highly conceptual when started and basically dropped off or were unrealized in a timely way," DeGroot says.
Alex St. John, CEO of WildTangent Inc. and a former Microsoft vice president, says the Vista development team failed to keep driver stability and application compatibility.
"Operating systems take a lot of work. I wasn't there, but at the top level you have to get [these issues] right before you can even dream of having a new feature," St. John says. "You can't remove any value. You have to start with the level of stability and consistency you have now and you build on it. You don't add new features until you have that."
The years-long delay allowed Microsoft competitors like Apple Inc., Google Inc. and Salesforce.com Inc. to establish leads in important new markets.
One current Microsoft exec says a perception problem around Vista persists within the company to this day.
"No one here talks about Vista any more, and they haven't for a while. All the buzz is around what Ray [Ozzie] is doing in cloud computing or whatever they're calling it now," he says. Ozzie, the chief software architect for the company, is leading Microsoft's Software plus Services (S+S) initiative as a response to Internet-bound competitors.
Despite Vista's well-documented troubles, top management at Microsoft seems committed to establishing a hybrid app environment where cloud-based computing takes on a growing role. The unfolding roadmap for Windows Live Services and new tooling such as the Volta Visual Studio add-in (see ""Battle
Royale Brews over JavaScript Turf") are all evidence of that position.
The Vista People Mover
Vista took a human toll, as well. Some longtimers at Microsoft remain upset at what they perceive to be poor treatment of execs like Allchin and Windows Core Operating Systems Division VP Brian Valentine. "Both [Allchin] and Valentine were thrown to the wolves," says another former exec who knew both executives. "Valentine was one of the best guys they had and they treated him badly."
A 19-year Microsoft veteran, Valentine left Microsoft in September 2006 -- just two months ahead of the initial Vista rollout -- to join Amazon.com. Allchin was still formally co -- president of the Platforms and Services Division at launch time, but he was nowhere to be seen during the Vista launch events.
Allchin quietly retired as Vista shipped, giving way to Steven Sinofsky, the former head of Office development who's widely credited for keeping Office releases on target. As senior VP for the Windows and Windows Live engineering group, Sinofsky quickly got to work addressing the problems that plagued Vista development.
Industry blogger Mary Jo Foley, author of the upcoming book "Microsoft 2.0: How Microsoft Plans to Stay Relevant in the Post -- Gates Era" (Wiley, 2008), says it's hard to gauge Sinofsky's impact on Windows development because he tends to tightly control information.
"Word was Sinofsky cut some of the fat out of the Windows client team when he took over in 2006. But I haven't heard a lot about what they're doing differently with SP1 and beyond [Windows 7]," writes Foley in an e -- mail interview.
"One promising sign is Windows 7 will be using the MinWin kernel, which sounds like it means Microsoft realizes the Windows client is bloated and still too full of dependencies and plans to do something about it," Foley concludes.
The MinWin kernel -- described in an October presentation by Microsoft Distinguished Engineer Eric Traut -- is an internal project that aims to enable a highly efficient and compact core for future versions of Windows. An audio file of the presentation can be downloaded here.
Microsoft, for its part, isn't ready to talk about what's beyond Vista. "We're focused on the value Windows Vista will bring to people today -- we're not yet giving official guidance about the future versions of Windows," says David Zipkin, Microsoft senior product manager for the Windows Client.
Saved by the Service Pack?
The good news for Microsoft is that Windows Vista Service Pack 1 (SP1) is plugging holes in the Vista OS right now. Microsoft dropped the SP1 release candidate on Dec. 5, 2007, and expects to ship the final SP1 bits in the first quarter. By all accounts, the service pack has improved matters.
Foley describes SP1 as the operating system Vista "should have been when Microsoft shipped it a year ago," citing improved networking performance, hibernate/shut-down functionality and device-driver support.
According to Microsoft, total sales of Vista topped 100 million in early January, with 96 percent of all retail PCs shipping with the new OS -- up from 77 percent last February. But despite the SP1 update, Zipkin doesn't expect an immediate uptick in enterprise deployments.
"We're still in the 'early adoption' phase, although we're starting to see more of the mainstream businesses begin their planning and deployment," Zipkin says.
Michael Cherry, senior analyst at Directions on Microsoft, believes SP1 should help sway IT organizations.
 |
"Microsoft has always gambled that the hardware would get in front of them. But this time around they were wrong. And wrong by a lot." |
| Michael Cherry, Senior Analyst, Directions on Microsoft. |
"In some ways Microsoft has trained customers not to upgrade until Service Pack 1," says Cherry, himself a Microsoft veteran. "So there are a certain number of people who have it almost as a policy to not deploy until that time."
For many IT organizations, the value of Vista SP1 boils down to one overriding issue.
"To me, the real issue with Vista has been compatibility," writes Andrew Brust, a Microsoft regional director and chief of new technology for twentysix New York, a Manhattan-based consultancy. "That's changing now, but given that we're a year post-RTM, the pace of change has been somewhat disappointing."
In November 2006, when Vista first shipped, there were 254 applications that carried the Certified for Windows Vista logo, according to Microsoft. Under SP1, that number rises to nearly 2,300. Device compatibility, another major sticking point for IT shops considering Vista, has shown major improvement under SP1. Today, some 74,000 Vista-specific device drivers are available, up from just 33,000 when Vista first arrived.
One thing the service pack likely can't fix is the conundrum posed by Vista's unique hardware requirements. Vista's most visible, user-facing feature -- the stunning Aero Glass user interface -- demands expensive, high-end graphics hardware and processing power.
"I think the hardware requirements are going to hold Vista back for a long time," says Cherry. "Microsoft has always gambled that the hardware would get in front of them. But this time around they were wrong. And wrong by a lot."
Delaying Tactic
Surprisingly for such an important OS release, Vista demands little specific attention from developers. Asked about the level of developer activity around Vista, Cherry is blunt.
"I've not seen any real [activity]. No, I'm not seeing it. There was an initial scramble to fix any significant app-compatibility problems you might have. That's been ongoing and it may in fact be already done," Cherry says. "That to me is different from actually writing an application to exploit Vista, which I'm not seeing happen."
One reason is that the latest versions of the .NET Framework, integrated into Vista, are freely available for Windows XP-based systems. That means developers don't have to target Vista at all to take advantage of key features like Windows Presentation Foundation (WPF) and Windows Communication Foundation (WCF).
"There's almost no such thing as 'coding for Vista clients.' WPF and the entire .NET 3.0/3.5 stack run on XP, making the only real Vista-specific development arena that of desktop gadgets," Brust says.
Dino Chiesa, director of .NET platform product management at Microsoft, begs to differ. He notes that Vista offers more than 7,000 new APIs for developers, including the Aero UI, and points out that Vista development can tap native OS features like the Vista Search APIs, RSS support and enhanced peer-to-peer APIs.
"We see a number of developers with extensive native code bases who want to keep their existing C or C++ applications, while extending them using functionality available from the .NET Framework, including WPF," Chiesa says. "These developers will be able to utilize both the Win32 APIs and the .NET Framework 3.0 from Windows Vista."
Foley thinks Windows XP SP3, due out in the first half of 2008, could delay Vista deployments for some time: "If you have XP SP3 running well, why move to Vista? Why not just wait until 2010 for Windows 7?"
So how should dev shops proceed? Cherry urges a conservative tack: "My advice would be you always are looking at the current version -- what I call the version-minus-one," he says. "What are the common set of features that are available in these versions, without anything fancy. And that limits what you write, because that's always going to give you the widest support."
Casting the .NET |
Because Microsoft made the .NET Framework 3.0 and 3.5 available for Windows XP, companies don't have to deploy Vista to take advantage of many of its most engaging features. Case in point is Fujitsu Transaction Solutions (FTS), a wholly owned subsidiary of Fujitsu Ltd. that provides point-of-sale software and solutions to major retailers. The company recently overhauled its GlobalSTORE point-of-sale application from Windows Forms to Windows Presentation Foundation (WPF).
The application runs on Windows XP-based client platforms, including the retail-specific Windows Embedded for Point of Service (POS) OS. The early adoption, says FTS Chief Software Architect Dave Killian, was vital to help FTS retail customers stay ahead of the curve.
"When you replace POS equipment, you're probably looking at a decade or more of runtime before they run another replacement round," Killian explains. "It's pretty important, when they make these decisions, that you can offer a solution that's pretty close to the state of the art at the time. WPF was directly in our targets."
The effort focused on reworking the brittle, forms-based screen interface and providing a more flexible, intuitive and engaging user experience. Killian and Senior Software Architect George Havens found WPF's XAML underpinnings to be a good fit.
"Our older platform was kind of a pre-XAML XAML," says Killian. "So our developers were already used to describing screens in XML anyway."
Working with Visual Studio (VS) 2005 Service Pack 1 and Expression Blend, Killian and Havens redesigned the application experience. The new tooling enabled a host of key advances, including an intuitive, task-based user interface with touch-screen controls and a scalable display. The new foundation also allows FTS clients to tweak the interfaces.
"One thing that WPF gave us is a toolset to allow our customers to go in and heavily modify the application from a look-and-feel perspective. We've architected our system such that this level of change doesn't really impact our other code," Havens says.
Killian and Havens did struggle with some aspects of the project. For instance, Havens says the ability to round-trip XAML code between Expression Blend and Visual Studio is limited.
"The one issue [Microsoft has] is the fact that the XAML code includes all of the hooks for the events. The problem with that is basically you have to send the entire set of screens back to them. And if I'm hooking up more things, having the events in the XAML code is not good," Havens says. "The abstraction for the event hookup needs to be separated out."
The dev team is also looking forward to updating the VS 2005 toolset, which Havens says is not optimal. "By March we'll be on VS 2008," he says.
In fact, the results of moving to .NET have been so compelling in the GlobalSTORE project that FTS is considering a move off ASP.NET for its Web-based in-store operations and administration software. Killian says his team may replace that infrastructure with a Silverlight 2.0-based application that leverages Windows Communication Foundation and Workflow Foundation.
-- Michael Desmond
|