Waffle is a weblog.
The author of Waffle, some guy in Sweden, also occasionally writes stmts.net.

Armed With Evidence to the Contrary

Gruber speculates that Windows 8 Metro-only ARM tablets would make for terrific iPad rivals, and that there’s no reason for Windows 8 ARM to run “desktop apps”. It would be an interesting bet that would make sense in many ways; there’s the chance for a clear start, finally.

Here’s why it’s not going to be that way.

The build-up to Windows 8 has focused on marketing a definition of “no compromises” that involves being all things to all people in a way that desktops, laptops and tablets all get both environments. Microsoft has set the expectation that Windows 8 will be both the desktop and Metro and that Windows 8 ARM hardware will be no better or worse than x64-architecture hardware. Windows 8 ARM is supposed to be able to do all that Windows 8 Intel can. Microsoft is interested in offering up old and new on both the “old” platform and the “new” platform and they see it as being beyond a feature but a necessity.

Microsoft’s position on the environment duality is that the desktop half won’t even be loaded until you use it initially. Metro apps run in a new environment with more rapacious process and power management and may not unto itself mean doom for battery usage. With this in mind, Microsoft gets the chance to tout the only alternative that remains fully a tablet OS as long as you only do tablety things and that can be talked into running Office when the need arises.

It’s not just that it makes sense through Microsoft’s constantly backwards-compatible-colored glasses, but that it makes a lot of sober business sense as well. It’s a trite example, but surely the enterprise that’s currently and somewhat shockingly the iPad’s biggest supporters wouldn’t mind an alternative that could also run Office and their old libraries and Windows programs? Office isn’t available for ARM right now, but probably will be within months of launching. Programs and libraries can mostly be recompiled and work for ARM; something like Excel is shock-full of hand-written assembly routines. Given the ramifications, there’s no reason why porting this isn’t the Office team’s highest priority from now until Windows 8 ships — split with starting the Metro versions, of course, but that’s done by complementary teams.

Windows 8, even with both environments, can still be used to power an iPad rival. I certainly didn’t expect to be writing that until just recently, but it’s true. Microsoft has made the right bet and made it look like the recumbent “let’s do nothing and hope we still win” position. Windows 8’s tablet chops are now on the non-ridiculous end of the scale and you could do something brilliant with Metro apps. As long as the market is seeded with lots of Metro apps, there won’t be very many reasons to run desktop apps. In such a situation, Windows 8 tablets will be able to compete based on the apps that will be running and not foiled by the power management tax of the desktop code that could be running.


  1. Strikes me that the answer to Gruber’s point about consumer confusion is Microsoft’s App Store, and some kind of ‘built for Windows 8’ to indicate that an application is CPU agnostic (i.e. same as Universal Binary was for OS X).

    And I agree with you – the killer feature for a Microsoft tablet is the ability to port legacy code, rather than complete rewrite – and you need to take into account how many companies have only done minimal work to get their applications onto .NET, if at all. (Most native software I use on Windows is either C++ based, or in some cases Delphi).

    (Plus there are a lot of firms out there looking for a route out from their Palm and Windows CE based applications, who for whatever reason are not keen on iOS/Apple. Many of them are dying, as their customers are replacing their systems with iOS ones, but there are plenty of niches out there, filled by ugly but effective stylus driven applications).

    By JulesLt · 2011.09.20 22:39

  2. Windows 8 ARM is supposed to be able to do all that Windows 8 Intel can.

    But it will never be able to, because it cannot run x32 or x64 code, and that is 100% of Windows apps today, and they will be the majority of apps long after Windows 8 ships also. You could almost hear Adobe’s sigh of relief when Apple dropped PowerPC and Creative Suite went Intel-only. They are not going to be ready to run Creative Suite on ARM any time soon. Same for Office and AutoCAD and so on. The latest thing in these kinds of apps is 64-bit, and ARM is not even 64-bit yet.

    It took Microsoft years to move MS Office from Mac PowerPC to Mac Intel, a simple PC CPU to PC CPU switch, and to newer and more powerful CPUs, and to Intel which they already had a version of Microsoft Office running on. ARM will be much, much harder.

    Apple also sold Mac OS X Tiger (Intel) as being a no compromises Mac OS X Tiger, but they built in Rosetta so that Tiger (Intel) could run PowerPC apps, and had Universal apps so that developers could push a tiny release with an extra Intel binary added to it. They made it so for the user it was 100% the same. The problem for Microsoft is if they sell Windows 8 as no compromises, but then a user buys a “Windows iPad,” just to run Dreamweaver (like the post-post-PC guy wants to do) and Dreamweaver won’t compromise by running on there. Even if there is a Windows 7 -like environment for it to run in, Dreamweaver won’t have the ARM compatibility.

    It’s a trite example, but surely the enterprise that’s currently and somewhat shockingly the iPad’s biggest supporters wouldn’t mind an alternative that could also run Office and their old libraries and Windows programs?

    No, definitely not. One of the biggest reasons the enterprise loves iPad is it is administered like a smartphone (cheap), not a PC (expensive). iPad doesn’t require the kind of I-T support and user training that a Windows desktop does.

    When iPad launched, I was working at a big multinational that was 98% Windows XP and Office 2003 and they had been rolling out Windows 7 and Office 2007 (or close to that) since the day both had shipped. They got stuck on the user training. They had to give each user so much training and they were not even as productive as on XP. Managers were dragging their feet, nobody wanted their teams to be switched over. And in the middle of all that, users started bringing in their own iPads, logging into Exchange, training themselves on the iPad, and moving themselves off XP. Their email and calendars and other data came with them, but they left behind a support and training and security nightmare on the PC. The iPad cost less to buy than just paying I-T hours to upgrade a Windows box, or just the user training for Windows 7.

    For years, Microsoft’s customers have been asking Microsoft to make the equivalent of an iPad with iWorks: almost instant setup, almost no training, highly reliable, very portable, and enables the user to do straightforward documents, spreadsheets, presentations, email, Web. Apple made it for them. The enterprise wants iPads, not PC’s. They wanted them all along. We already know this because the enterprise passed on Wintel tablets for $999, but can’t get enough of $699 iPads. The difference is the retail price of Windows itself. If it was worth it to them to have Windows in their tablets, they would have bought Windows tablets. There were over a decade of them to choose from.

    So I think what Microsoft is doing if they have traditional Windows apps on ARM is giving users 50% of what they think they want instead of 100% of what they really need. What users need is a cheap ARM tablet that only has Metro apps, has no viruses, and you can tap a button in Microsoft Store and install a simplified Microsoft Office for $30. If Microsoft doesn’t provide that, their customers will keep buying iPads and iWork and going crazy over it. I’ve heard so many old-school MS Office users raving about Keynote on iPad over the past 17 months, it is crazy. Their expectations of their business tools have changed. The last thing they want is the stuff they already left behind. In a year or 2 when Windows 8 ships, that will be even more so.

    By Hamranhansenhansen · 2011.09.21 01:51

  3. Please note that I’m not saying that Gruber’s idea of a Metro-only tablet isn’t a good idea. It makes for a more consolidated experience, and if you like living in Windows 7 on a touch device, you can do that today. But leaving the Desktop half of Windows 8 in isn’t just a liability, it’s also a cover-my-ass-portunity. Whether it’s a good idea in practice will depend on how many Metro apps there are available and in development on day one, and how many ARM desktop apps there are.

    By Jesper · 2011.09.21 07:12

  4. There are two massive hurdles that make “classic” apps running on ARM seem highly implausible.

    1. The Win32 API was never designed with power management in mind, so even if you could recompile the x86 apps, they would likely chew through the battery in no time.

    2. There is a massive amount of Intel assembler and architecture-specific code throughout Windows itself, not to mention Office. Porting this to ARM would be a massive task.

    I agree with Gruber – my bet is that ARM will be Metro/WinRT only.

    By GB · 2011.09.21 23:57

  5. GB: They already have ported much of it. How much we won’t know, but they have shown IE10 on stage at MIX11 this spring running on ARM. Porting this will not be a problem. Windows already runs on x86, x86-64 and Itanium; although they are all from Intel and although Itanium support is being dropped, there’s not some sort of magical Intel pixie dust that holds it together. NT is portable stuff, and Microsoft has managed to deliver most parts of Windows on top of it for many platforms.

    I won’t fight you on power management. You are probably right, since Windows CE, which is meant for this sort of thing, still requires a different kernel.

    The Microsoft of five years ago would have shipped it. The question is whether the Microsoft of today, capable of aiming for, producing and relishing something like Metro, will opt to ship it. Every principle that they have ever held tells them to “remain compatible”, even if Win32 on ARM is something new. Whatever they’ve learned recently tells them to take a step back and try to work it out.

    It doesn’t come down to skill. It comes down to will.

    (I don’t personally necessarily want them to ship Win32. I want them to bury it deep and never look back, although from several standpoints I am scared of Metro and what it means. But I am mindful enough to have observed that the universe optimizes for other terms.)

    By Jesper · 2011.09.23 18:38

Sorry, the comment form is closed at this time.