The long kiss goodbye for x86 desktop Windows

The transition from traditional Windows desktops to the Post-PC world, the ARM architecture and the Metro user interface is inevitable. But it won't a be quick one.
Written by Jason Perlow, Senior Contributing Writer


My two ZDNet colleagues, David Gewirtz and Steven J. Vaughn-Nichols have gotten themselves into a bit of a lover's quarrel. SJVN says that Windows 8 will be dead on arrival, and yet Gewirtz says that Windows 8 will matter for real work, as will Windows 9.

Seriously guys, get a room.

Both of these men have some interesting viewpoints to offer. Neither of them are correct, though -- the real answer lies somewhere in between.

Back in September, shortly after the completion of Microsoft's BUILD conference, I did some preliminary analysis of the huge pile of information that Redmond dumped on us that week. The conclusion that I came to was that Microsoft had handed down the death sentence for the Personal Computer as we know it today.

The crux of my argument at the time was that the new WinRT API's clearly represent the next phase in Windows' evolution and that there was no room on the ARM platform for porting legacy Win32 applications.

While Microsoft did not definitively say that at the time, I had inferred this simply because I knew what a difficult task it would be to move all of that legacy code over to the new ARM architecture.

Even if Microsoft had ported all the supporting libraries and APIs over, it would be a nightmare to try to support every developer to move their spaghetti Windows code over to the new platform.

No, a clean break was absolutely required.

If you read the Talkbacks in that September 2011 article I was derided by many of our readers for making such a bold statement, that none of the code would be portable with a simple re-compile.

I can sum up the bulk of their responses fairly neatly. "But look! There's a traditional Windows desktop on the ARM version! There's a video that proves it! Microsoft demonstrated Office on ARM! You're an idiot!"

But I knew that I was right.

Speculation continued on for months as to whether or not the "Blue" APIs would be avaliable on the traditional, non-Metro "Desktop" on the ARM version of Windows. But last week, Microsoft Windows group lead Steven Sinofsky put all of those arguments to rest in an 8,000 word manifesto which confirmed exactly what I had believed and knew all along. Here's the key takeaway:

Developers wishing to target WOA do so by writing applications for the WinRT (Windows APIs for building Metro style apps) using the new Visual Studio 11 tools in a variety of languages, including C#/VB/XAML and Jscript/ HTML5. Native code targeting WinRT is also supported using C and C++, which can be targeted across architectures and distributed through the Windows Store. WOA does not support running, emulating, or porting existing x86/64 desktop apps. Code that uses only system or OS services from WinRT can be used within an app and distributed through the Windows Store for both WOA and x86/64. Consumers obtain all software, including device drivers, through the Windows Store and Microsoft Update or Windows Update.

So, just in case you didn't understand that, if you are a developer and have a complex Win32 application, you are going to have to completely re-write it in the WinRT APIs if you want to target it to run natively on an ARM device. The "Desktop" environment exists in Windows 8 on ARM, but it is strictly reserved for Microsoft's use only, which includes their port of Office. NO WIN32 FOR YOU!

Nobody likes anyone who says "I told you so." Well, too bad. I told you so.

So what does this mean for the future success of Windows 8 and the Microsoft technology platform as a whole?

Well, the answer is not so cut and dry. In one sense, Windows 8 will be a failure. But in another, it will be a huge success.

From the perspective of the software developer who has written Win32 apps that run on Windows 7 today, it means that continuing work on Win32 and improving those products probably doesn't make a heck of a lot of sense.

At best, for the software developer, Win32 apps become a maintenance business. All efforts are now going to be focused on writing applications that are written in the WinRT APIs, so that they can run on the Metro interface across the two architectures.

From the perspective of the consumer and enterprise, an upgrade to Windows 8 in late 2012 or early 2013 means getting all new Windows software from the new Microsoft Windows Store (the "App Store") if they want to take advantage of the new Metro features on x86 PCs.

The likely outcome is that Windows 7 is going to remain the most popular version of desktop Windows until a significant number of applications are ported over to the new Metro/WinRT environment. So we can expect both consumer and corporate adoption of Windows 8 at least on the desktop to be slow. Very slow.

The bottom line is that there's very little incentive to move to Windows 8 at all if you're just going to be running the same Win32 apps you have today.

After all, many large enterprise customers only just completed their Windows 7 rollouts or are still in the process of doing them. They're not going to undertake another migration in 2013. Or 2014 or even 2015 for that matter.

Sure, they'll refresh the hardware if necessary, but Windows 7 is going to be king for a very long time.

Now, from the server perspective, that's an entirely different ball of wax. As I said in an earlier piece, I think Windows 8 Server is going to be a very big deal for corporations, because the value proposition is extremely strong.

Windows 8 Server is going to be a major upgrade for any organization looking to take advantage of server and desktop virtualization technology (VDI) and for building private cloud infrastructure.

Desktop Windows on the x86 platform is going to be around for a long time -- but in the form of consumers and enterprise sticking with their Windows 7 systems and Win32 apps until the value proposition of moving to WinRT-based apps is high enough to justify the upgrade.

That kind of transition could take 7 to 10 years, the very same type of long, drawn out upgrade cycle we saw from users moving from XP to Windows 7. So SJVN is correct that Windows 8 on x86 may not get a ton of traction.

But what about Windows 8 on tablets? Or on ARM-based thin-clients?

Well, that's where things get a lot more interesting.

We now know that Windows on ARM (WOA) is going to include a version of Microsoft Office pre-loaded. So while enterprise take-up on the desktop, x86 version of Windows 8 could indeed turn out to be slow, this may be the exact kind of thing the doctor ordered for Windows to gain traction in the enterprise as the tablet OS for "Getting Real Work Done" as Gewirtz has so eloquently put it.

And ARM-based desktops, which will essentially be maintenance-free appliances like the iPad, when paired with the enhanced RemoteFX/Remote Desktop Services capabilities of Windows Server 8, will significantly lower total cost of ownership of the desktop environment by allowing these devices to run legacy Win32 applications via VDI.

It may sound awfully Science-fictioney, but I've had this technology demonstrated to me by Microsoft and the user experience of running remote Win32 apps on the server looks amazingly seamless and there's no compromise for the end-user whatsoever. And I would expect the same of server-side WinRT applications as well, once the Windows ecosystem has moved to that programmatic model.

But the most important thing is that the balance of the ongoing maintenance burden for these ARM-based PCs will be transferred from the enterprise support staff to Microsoft and the OEMs, which is a big deal.

To quote Sinofsky's Windows on ARM manifesto again:

Partners will provide WOA PCs as integrated, end-to-end products that include hardware, firmware, and Windows on ARM software. Windows on ARM software will not be sold or distributed independent of a new WOA PC, just as you would expect from a consumer electronics device that relies on unique and integrated pairings of hardware and software. Over the useful lifetime of the PC, the provided software will be serviced and improved.

Personally, I'm looking forward to the day when enterprises can finally get the desktop support monkey off their back and concentrate on managing their shared private cloud infrastructure instead and leveraging as much server and virtualization technology as possible.

Will the transition to Windows 8 desktops be a quick and painless transition, or will it be a long kiss goodbye from Windows 7? Talk Back and Let Me Know.

Editorial standards