inset
Longhorn Components on Windows Roadmap
Sep. 6, 2004

A new OS roadmap announced in Sept. 2004 indicates that several features planned for the next Windows client, code-named Longhorn, will be released for current versions of Windows as well. In particular, the WinFX API set, Avalon graphics system, and Indigo messaging system planned for Longhorn will also work on the current OS generation. However, the WinFS file system that was planned for Longhorn has been postponed. These shifts dilute the impact of the Longhorn OS release, but they could attract more corporations to current Windows versions and encourage more developers to evaluate the strategic WinFX API set.

(See the illustration "Windows Roadmap, Sept. 2004".)

Longhorn Splitting and Shrinking

Announced in 2002 and initially planned for delivery in 2005, the Longhorn OS release has now been substantially reduced in scope, to the benefit of current OSs and Microsoft's platform strategy as a whole.

New APIs for Graphics, Messaging, Files

Initially, Longhorn promised three major new features:

"Avalon." A new Windows graphics technology and alternative to the current Windows (Graphical Device Interface (GDI) and User systems, Avalon improves display quality by exploiting 3D graphics hardware, and it delivers a new user interface technology called XAML (Extensible Application Markup Language, pronounced "Zamel") that enables developers to create application user interfaces with less code than current technologies. Avalon also supports a new printing architecture and Document Services technology that enables applications to generate output in fixed format documents (analogous to Adobe PDF files) that produce predictable results on any supported printer.

"Indigo." An application-to-application messaging system analogous to the existing Microsoft Message Queuing (MSMQ) and Distributed Component Object Model (DCOM) systems, Indigo uses multivendor standards such as XML to encode messages and the WS-* family of Web services technologies to secure messages and ensure reliable delivery.

WinFS. Using an integrated database engine (derived from SQL Server) on top of the current Windows file system, WinFS enables applications and users to label files with custom properties (e.g., geographic coordinates where a digital photo was taken) to simplify search and organization. WinFS also provides standard property schemas for commonly used data types, such as contacts, calendar items, and e-mail, to simplify sharing these data types among applications and enable users to do some simple tasks (e.g., changing a contact's primary phone number) directly from the Windows shell.

All of these features were to be accessed on Longhorn through a new set of APIs called WinFX. Unlike Microsoft's existing Win32 APIs, WinFX is designed to support applications that are loaded and run under Microsoft's .NET Framework, which relieves application developers of such housekeeping tasks as memory management to make them more productive and reduce errors. The .NET Framework also is the primary target of Microsoft's developer tools, so WinFX could reduce the "tools gap" between the time Microsoft adds a feature to the API and the time its tools support the feature. Finally, WinFX is better organized and less cumbersome than the over more than 15,000 Win32 APIs in use today, enabling developers to find the APIs they require more quickly.

Current OSs Now Get Avalon, Indigo, WinFX

The Avalon and WinFS features were to be exclusive to Longhorn, while Microsoft had always planned to ship Indigo for Windows XP and Windows Server 2003 in addition to Longhorn. Under the new roadmap, in contrast, Microsoft plans to create versions of Avalon, Indigo, and WinFX for redistribution onto Windows XP and Windows Server 2003. Microsoft has not said how it will package the new features for current OSs. Most likely, it will take the form of a 2006 "WinFX feature pack," a Windows service pack that also delivers the new features.

However, Avalon is likely to be significantly scaled back for these older systems. As originally planned, Avalon relied on a new Longhorn device driver model that will not be present on current OS versions. "Aero," a Windows desktop user interface demonstrated for Longhorn that used Avalon, is unlikely to be implemented on current systems.

Microsoft still plans to release a Longhorn OS for the client in 2006 and for the server six to 12 months later. In effect, Longhorn will have the WinFX feature pack preinstalled and will deliver additional improvements, such as a new generation of the Automated Deployment Service (ADS) to support creation and deployment of OS images.

Cairo Still a Mirage

Under the new roadmap, the WinFS file system will not ship in the WinFX feature pack or the Longhorn OS release.

Originally proposed in 1994 for "Cairo," a subsequently abandoned release of Windows, WinFS is slated to arrive in a beta release in 2006 and for production "shortly thereafter," according to Microsoft. However, the company has not set a date for production release. If Microsoft decides to hold production release of WinFS for the next major Windows release after Longhorn, WinFS is unlikely to appear before 2008.

Changing the "Competitive Dynamic"

Overall, the Longhorn OS release will be less of a technical leap than Microsoft originally planned. For Microsoft, this might seem embarrassing, but it has many advantages.

The primary advantage is strategic: the move increases the number of computers that could potentially run WinFX and benefit from Avalon and Indigo. It also improves the odds that Microsoft can deliver these technologies on time, which in turn makes these technologies more attractive to application developers, including Microsoft product groups, and speeds the day when Windows applications can showcase the technologies. The result, Microsoft hopes, will be a new base of desktop applications that can't be easily replicated on competing platforms, such as Linux, and a new burst of developer momentum that could carry the .NET Framework past Java once and for all.

The decision could also benefit Microsoft's immediate business prospects by triggering migrations to Windows XP and Windows Server 2003 among organizations that were waiting for Longhorn. And if Avalon applications prove truly compelling, they could encourage new PC sales as consumers and businesses upgrade to the latest graphics hardware. However, as Chief Software Architect Bill Gates noted at the 2004 Financial Analyst meeting, the goal of Longhorn was less to bring in direct revenue than to change the "competitive dynamic." Bringing Longhorn technologies to more computers, more quickly, will strengthen that effect.

Next Client Update in 2006

Given these changes, the roadmap for the Windows client OS looks as follows:

WinFX for Windows XP. A feature pack or other form of software release will deliver Avalon and other WinFX APIs for Windows XP. Microsoft has not said how it will license this technology, nor has it indicated how application developers who use the technology will be able to ensure it is present on target computers. Estimated release date: 2006.

Windows Longhorn Client. As noted, this release will deliver preinstalled WinFX technologies and will deliver as-yet undefined improvements to OS fundamentals, such as an updated installation technology for OS images. Estimated release date: 2006.

As for existing versions, Windows XP (released in Sept. 2001) is scheduled to exit Mainstream support two years after Windows Longhorn Client ships, which implies full support through 2008 under the current schedule and policies. (See the chart "Past Windows Client Versions".) Windows XP Service Pack 2 (SP2) is the most recent service pack and has features and testing requirements similar to an intermediate OS release.

Microsoft has not indicated whether it will release another service pack for Windows 2000, which therefore may never get the security improvements delivered in Windows XP SP2.

Windows 98, 98 SE, and Me will see no development effort other than security hotfixes as Microsoft focuses on more recent Windows versions.

Windows Server: The Service Pack Squeeze

Microsoft says the Longhorn announcement does not affect the roadmap for Windows Server. However, recent events raise doubts about that roadmap. Specifically, the next major service pack for the server, Windows Server 2003 Service Pack 1 (SP1), has slipped from the second half of 2004 to the first half of 2005, leading to some tight deadlines for the Windows team.

Service Pack 1 Next on Horizon

The Windows Server roadmap currently looks as follows:

Windows Server 2003 SP1. In addition to completely tested fixes for known security vulnerabilities and bugs, this service pack will contain some security improvements also delivered in Windows XP SP2 (such as a more secure remote procedure call service) and a role-based security lockdown wizard. Windows Server 2003 SP1 will likely become the new baseline OS shipped to the channel. Estimated release date: first half of 2005.

Windows Server 2003 R2 update. This release (R2 is still only a code name) will deliver Windows Server 2003 with SP1 preinstalled (slipstreamed), together with some new features. The new features will simplify remote access to applications running on Windows Server, for remote users and business-to-business application integration. The release will also include new features for managing branch offices from a central site.

In addition to new features, R2 will package various Windows Server feature packs for tasks such as portal hosting and Unix emulation and will likely deliver the .NET Framework 2.0. Customers who do not have upgrade rights on Windows Server will need to buy new server licenses, which will include the R2 features. Microsoft intends to retire support for R2 at the same time as Windows Server 2003. Estimated release date: Late 2005 or early 2006.

WinFX for Windows Server 2003. Like the corresponding feature pack for Windows XP, this feature pack will bring Avalon and possibly Indigo to Windows Server 2003. It's possible that some client-oriented features in Avalon will not be present or enabled by default. For example, few servers have 3D graphics accelerators or benefit from their use, so features that target such hardware might not be available. Estimated release date: 2006.

Windows Longhorn Server. This major server release will be built on the same code base as the Longhorn client and will support WinFX, although some client-oriented features (e.g., 3D graphics user interface features) might not be present on the server or might not be enabled by default.

A single code base for both client and server will simplify maintenance for both Microsoft and customers. For example, it will enable Microsoft to release Windows service packs that apply to both client and server, something it can't do today with Windows XP and Windows Server 2003.

Windows Longhorn Server is also currently slated to deliver new systems management technologies being developed as part of the Dynamic Systems Initiative (DSI). Among other things, these technologies will enable applications to be monitored and configured using a System Definition Model (SDM) created during application development and installed with the application itself. However, it's not clear at this point how much SDM support will actually ship in Windows Longhorn Server. Estimated release date: Six to 12 months after Windows Longhorn Client (probably 2007).

As for existing Windows Server versions, NT 4.0 is approaching its final support dates, and Mainstream support for Windows 2000 Server ends in 2005. (See the chart "Past Windows Server Versions".)

A Squeeze on the Team?

The revised roadmap will put unprecedented strains on the Windows team. It calls for the team to deliver Windows Server 2003 R2, WinFX feature packs for Windows XP and Windows Server 2003, and the Longhorn client OS in 2006, followed by WinFX for Windows Server 2003 and Longhorn Server six to 12 months later. This schedule would stress resources, such as test centers, that are shared across the Windows client and server teams—a possible factor in Microsoft's recently announced decision to hire expertise in automated testing. However, automation is not enough—key developers in the Windows core team must often be consulted to fix problems found in testing, and their work cannot easily be automated.

The most likely outcome is that Windows Server R2 will be eliminated as a separate release, and some of its features put into the Windows Longhorn Server release in 2007. This would eliminate a large number of build and test cycles and the resulting strain on the Windows team. Another, more remote possibility: the company could ship R2 but substantially delay Windows Longhorn Server, giving the Windows Server team time to integrate a production version of WinFS.

New Roadmap Could Spark Reevaluation

The new roadmap suggests possible strategy shifts for organizations that have not yet migrated to Windows Server 2003 and Windows XP, and for software developers who were underwhelmed by Longhorn or who have delayed evaluating it.

Organizations: A Reason to Get Current

Organizations on Windows 2000 or earlier should seriously consider moving to Windows XP SP2 on clients and to Windows Server 2003 SP1 on servers after it becomes available.

Computers running these versions will benefit from most planned improvements to Windows through 2008 without requiring a full OS upgrade. Both Windows XP and Windows Server 2003 deliver significantly stronger security than earlier versions, and each has performance and administration improvements that will pay off on at least some computers. Windows XP and Windows Server 2003 will have paid support and free security hotfixes at least through 2013. They will have Mainstream support at least through 2008, although license sales and substantial software updates from Microsoft could end well before then (as has been the case with Windows 2000).

Note, however, that organizations might still have to buy licenses to get future improvements, even if those improvements apply to existing computers. Microsoft has already said that Windows Server 2003 R2 will require new licenses or upgrade rights. It has not said whether it will charge for the WinFX feature for Windows XP and Windows Server 2003, or whether it will allow developers to redistribute a run-time version of WinFX with their applications.

Developers: A Second Look at Longhorn Technology?

Application developers should reevaluate Avalon, the WinFX APIs, and the .NET Framework, given the change in release strategy.

Under the latest roadmap, WinFX will reach many more customers than it would if it were limited to Longhorn. Furthermore, Microsoft's decision to cut out WinFS makes it more likely that WinFX will ship as planned in 2006, giving developers who want to use WinFX a clearer target date. This date will also encourage Microsoft's own application developers, especially the Office organization, to demonstrate the value of WinFX in new application versions, which in turn could raise user interest in WinFX applications.

In particular, ISVs who deliver client-side applications should revisit WinFX. As a Longhorn-only feature, WinFX looked like a hard sell to such ISVs, which benefit most from a broader market. Those ISVs can now include the Windows XP base in their calculations. They will also want to evaluate planned improvements to the Visual C++ programming language that will make WinFX’s strengths more accessible to developers. (See "ISVs Focus of C++ Roadmap" on page 17 of the Sept. 2004 Update.)

The time is not yet ripe to evaluate WinFX, however, because its design is not yet complete and its redistribution model is undefined. As noted above, some Avalon features depend on a driver model and graphics hardware support that were to be delivered in the Longhorn OS: these features will have to be scaled back on Windows XP and Windows Server 2003. Furthermore, Microsoft has not said how it will license WinFX and any system components that it requires to run. They could be redistributable with applications at no charge, as the .NET Framework is today, or they could be extra-cost items like Windows Server 2003 R2. Licensing could thus have a major effect on developers' cost-benefit calculations. Microsoft will have to finish the WinFX design and clarify its licensing before application developers can sensibly evaluate it for their own work.

Resources

For a convenient reference to enterprise software releases and retirements, see the Directions on Microsoft Enterprise Software Roadmap. At the time this article was published, the latest version of the Roadmap report was dated June 2004.

A preview of Longhorn features as originally planned appeared in "Longhorn Preview Focuses on Developers" on page 3 of the Dec. 2003 Update.

The WinFX APIs are analyzed in "WinFX: The .NET Framework Takes Center Stage" on page 18 of the Dec. 2003 Update.

The Indigo messaging and Web services technology was the topic of "Indigo to Aid Web Services Development" on page 22 of the June 2004 Update.

Microsoft's technical site for Longhorn is msdn.microsoft.com/longhorn.

For an overview of Microsoft's product life-cycle support policies, see support.microsoft.com/default.aspx?scid=fh%3Ben-us%3Blifecycle.

Support life-cycle dates by product are listed at support.microsoft.com/default.aspx?scid=fh;[ln];complifeport.

Supported service packs and their release dates appear at support.microsoft.com/default.aspx?scid=fh;[ln];lifesupsps.

Service pack release histories for many products are at support.microsoft.com/default.aspx?scid=FH;[LN];sp&.