News

Why Longhorn Still Matters

• By Scott Bekker
• 04/01/2005

First Microsoft separated WinFS, Avalon and Indigo from Longhorn and last month the company announced Internet Explorer would join them in shipping before Longhorn does. Casual observers may shrug and say there's nothing much left of interest in Longhorn. But a more careful reading of the Microsoft tea leaves suggests Longhorn presents the company with a rare opportunity to address a fundamental problem: speed.

Competitive pressure forced Microsoft to upgrade Internet Explorer before it ships Windows Longhorn, as was the original plan. The pressure is broader and more subtle than the infamously quick transfer of a slice of usage share from Internet Explorer to the open source Firefox browser (see "Internet Explorer's Loss, Firefox' Gain," below). Microsoft had to fight concerns about the security of Internet Explorer and quell doubts about its commitment to its dominant Web browser. Those question marks leave Microsoft vulnerable to browser challenges from Firefox, AOL-Netscape, Opera and possibly Google or Yahoo.

The problem is that, until mid-February, Microsoft clung to Internet Explorer improvements as one of the last feature areas made public so far that distinguish the 2006 Longhorn release from Windows XP.

Internet Explorer 7.0 for Windows XP Service Pack 2 will strip even that vestige of identity from Longhorn. The WinFS file system is already gone—consigned to an uncertain future as a beta technology. The Avalon and Indigo subsystem pillars will both be available as add-ons for Windows XP and Windows Server 2003.

But the Intel Developer Forum (IDF) in March served as a reminder that there's a fourth pillar to Longhorn called "operating system fundamentals." It may not have a sexy moniker, but it's the only pillar left standing, and it could be the reason Longhorn still matters.

Internet Explorer's Loss, Firefox' Gain
Browser usage share statistics from Web analytics firm WebSideStory:
Date	IE Share	Firefox Share
Jun 4, '04	95.48%	3.53%*
Nov 5, '04	92.89%	3.03%
Dec 3, '04	91.80%	4.06%
Jan 14, '05	90.28%	4.95%
Feb 18, '05	89.85%	5.69%
* WebSideStory did not track the Firefox browser separately until October 2004. The June figure includes all Netscape and Mozilla-based browsers, including Firefox.

IDF underscored the growing momentum of the x86-64 architecture, which allows most x86-based apps to run unmodified against the much larger memory limits of 64-bit systems, and of multi-core processing, processors containing more than one processor core. Both AMD and Intel are introducing dual-core processors this year, and Microsoft expects dual-core processors to be mainstream by Longhorn's release next year. The x86-64 technology is on a similar production path.

"We believe … the next step in Moore's Law is really about the parallelization," Microsoft Group Vice President of Platforms Jim Allchin said at IDF. Parallelization refers to the idea of multiple tasks running in parallel to boost speed, a process that multi-core processors make far more feasible. "You take a laptop, a typical laptop, there are about 500 threads in it. You can imagine, if you had the extra processors, what you could actually do with it."

With Microsoft licensing by the socket, Intel charging by the whole processor instead of per core and memory prices continuing to fall, some of this new power could be had for the same price as current desktops when Longhorn ships.

When the new übersystems arrive, we'll quickly see killer apps in gaming driving new system purchases the same way Doom 3 and Half-Life 2 drove new PC buying last Christmas season. Wizards like John Carmack at id Software will no doubt create dizzying gaming engines that conjure every bit of the potential of the processing power and affordable memory.

On the business side, the hard thing to envision with these rapid advances is where it fits in if your users aren't doing oil and gas exploration or automobile design. Consider that the killer business application for this explosion of computing power could be Microsoft Office or any other desktop program that users can start working in immediately without any need to wait for a network connection.

In an increasingly connected world, there's still a lot to be said for fat clients. If the host is down, the Web app doesn't work. If the DSL line goes out, the whole branch office is on coffee break, again. In that light, Microsoft's self-interested business case for smart, fat clients that connect to what's available and cache or offer seamless offline productivity when there's no connection makes sense.

"We believe the next step in Moore’s Law is really about the parallelization." Jim Allchin Microsoft Group Vice President of Platforms

So, good for fat clients. In the 32-bit era, Microsoft just couldn't improve the speed of Windows and Office enough to offset the feature bloat that Microsoft executives determined was justified to satisfy the needs of its sprawling user base. Word 2003 on Windows XP Professional coughs and chokes to a start seemingly as slowly on an Intel Pentium 4 with 256MB of RAM as Word 97 did on Windows 98 SE on a Pentium with 32MB of RAM. If the fat clients ran more like leopards than cows, Windows and Office would have more customers who were also fans.

Imagine a computer that pops on with the speed of a TV. Imagine Word, Excel and Outlook popping open instantaneously. Imagine the productivity gain in your organization if, when your users hit their power buttons in the morning, they didn't need to get up and go for a cup of coffee to kill five minutes before they can log on.

That kind of speed is within reach. Multi-core processors and 64-bit chips should tear the roof off capacity, leaving Microsoft adequate space to continue to meet the requirements of customers without killing system performance. If no new killer 64-bit applications are on the horizon to suck up all those valuable new system resources, at least those resources can be used to gang-tackle the old killer applications into really fast, more productive, old killer applications.

With beta versions of Longhorn getting closer, Microsoft may have some astounding new features lined up, like a 3-D interface, that may somehow hog resources, especially in the next couple of years when system memory will remain too expensive to soar much beyond 32-bit limits. But Microsoft has a rare opportunity to focus on end users' need for speed. It will require focus and attention—and potentially a willingness to sacrifice eye candy for productivity.

Windows XP wasn't a great operating system because it had better photo management, better video capabilities or a better media player. Windows XP was a great operating system because it gave consumers and business users an easy-to-use OS that didn't lock up and demand a hard reboot at least twice a day (the comparison is to Windows 98/98 SE, not Windows 2000 Professional). In other words, it worked almost the way it was supposed to, although security problems have cut back on ease of use.

Microsoft has the opportunity to make Longhorn a great operating system by optimizing it to be a fast, easy to use OS that doesn't lock up and demand a hard reboot twice a day. The convergence of hardware gains mean Microsoft may be able to make Longhorn and the accompanying Office 12 extremely fast without the kind of extreme discipline that might cost market share and make it impossible for Microsoft to consider.