Linked by Thom Holwerda on Wed 12th Dec 2012 19:37 UTC
Windows "As early as mid-December, consumers will be able to go to retail stores in the United States and Australia to purchase a Surface with Windows RT. Additional availability will be added in a number of countries in the coming months." Sales might indeed benefit from, you know, allowing the world to actually buy your halo product. Us Dutch won't be getting the Nexus 4 and 10 either.
Thread beginning with comment 544871
To view parent comment, click here.
To read all comments associated with this story, please click here.
Morgan
Member since:
2005-06-29

I've heard people dish on Windows 8/RT for this reason, and they contrast it with Apple's smooth transition from PPC to x86 back in 2005.

The difference is that Apple invested the time and money in building a universal binary system. They had done a platform jump before with 68k to PPC, and they knew what to do to keep their users happy.

Microsoft, on the other hand, has been building on the same monolithic base from day one. Instead of doing the universal binary thing, they decided to push two incompatible architectures in parallel. This causes fragmentation and confusion, and I think it was the worst thing they could have done.

Reply Parent Score: 1

tylerdurden Member since:
2009-03-17

Microsoft, on the other hand, has been building on the same monolithic base from day one.


Yes and no. The NT kernel was designed/built with portability in mind from the get go. Microsoft never got their act together (with regards to portability) when it comes to the user/application space, however.

Reply Parent Score: 2

henderson101 Member since:
2006-05-30

Originally, yes - not since NT4. NT4 took the UI out of user space and put it in the Kernel, and so in doing, made portability much more involved. Up to NT3.51, it was designed so that every subsystem could be replaced easily - this is where a lot of the "NT is highly portable" comes from. e.g. NT hasn't run on PowerPC, DEC etc, *since* NT 4 was released. Any later port would have been a larger undertaking than "just recompiling the kernel and creating a new HAL".

Reply Parent Score: 2

galvanash Member since:
2006-01-25

The difference is that Apple invested the time and money in building a universal binary system. They had done a platform jump before with 68k to PPC, and they knew what to do to keep their users happy.


I don't think that is fair at all... In both of Apple's platform changes, the chosen path for backward compatibility was emulation. Emulating 68k on PPC (or PPC on x86) is a totally different proposition than emulating x86 on ARM. In the first case you are doing the emulation on hardware that is dramatically faster and more capable than what the original binaries were targeted against. Also, other aspects of the systems had improved as well - memory becomes cheaper so you generally have much more of it to work with, and more memory bandwidth as well. it is quite a different matter to go the other way around - ARM systems are both much slower and generally have much less memory than a typical x86 machine.

Point being, assuming Microsoft would have given ARM equal footing to x86, the real problem is not allowing new software to target both platforms equally - which is what something like universal binaries solves... The problem is supported existing x86 software on ARM

How do you do that when a typical ARM system performs like an x86 system from 10 years ago? In many areas, i.e. floating point, the performance delta can be as much as 2 or 3 orders of magnitude. There might be a handful of modern x86 apps for windows that would perform adequately on ARM through emulation, but most would simply be completely unusable.

Apple never even tried to solve this problem. Do you see iPads running OSX apps? You giving Apple credit for doing something they never did, and criticizing Microsoft for doing exactly what Apple did - which is pursue two separate parallel platforms.

Reply Parent Score: 3

Morgan Member since:
2005-06-29

The focus of my comment was on the PPC to x86 shift; I only referenced the 68k to PPC transition to show Apple's prior experience with the process.

Yes, there was PPC emulation for legacy apps, but it was never intended to be the solution because it was simply untenable for most practical applications. Apple fully intended to use universal binaries (so-called "fat binaries" that contained executables compiled on both PPC and x86) from the start. They had been building and testing x86 versions of OS X and core applications from the very first release of the new OS. There was extensive documentation and support in the ADP for using fat binaries, whereas Rosetta was deemphasized and rightly considered legacy.

But you're right, I should have clarified all of that.

Reply Parent Score: 2

zima Member since:
2005-07-06

I've heard people dish on Windows 8/RT for this reason, and they contrast it with Apple's smooth transition from PPC to x86 back in 2005.
The difference is that Apple invested the time and money in building a universal binary system.

That's a bit unfair. The difference is more nuanced... PPC->x86 of Apple was within the same kind of devices (PPC desktops & laptops -> x86 desktops and laptops). The universal binaries don't encompass Apple mobile devices.

And that's the jump MS is doing, desktops & laptops -> mobile.

Reply Parent Score: 2

Nelson Member since:
2005-11-29

I've heard people dish on Windows 8/RT for this reason, and they contrast it with Apple's smooth transition from PPC to x86 back in 2005.


I've heard people around the tech circles talk about this, but it's mostly repeating a talking point someone else has fed them.


The difference is that Apple invested the time and money in building a universal binary system. They had done a platform jump before with 68k to PPC, and they knew what to do to keep their users happy.


Microsoft has managed this transition better or the same as Apple. Windows Store apps support x86 and ARM seamlessly. The appropriate architecture is pushed down to you (in the case of C++), or it is JIT'ed on the fly in the case of .NET then NGEN'ed at a later time by a service. For JavaScript it is AOT'd iirc.

So the fact that my Windows Store app works on x86 and ARM without me even thinking about it is a huge achievement. Plenty of Surface RT users are enjoying it right now.


Microsoft, on the other hand, has been building on the same monolithic base from day one. Instead of doing the universal binary thing, they decided to push two incompatible architectures in parallel. This causes fragmentation and confusion, and I think it was the worst thing they could have done.


I don't really think you're in any position to know about the architectural characteristics of Windows 8. Windows RT in fact exhibits very comparable performance to Windows 8. I've seen no major performance differences between the two when it comes to writing Windows Store apps.

Now, the part you're likely complaining about is Desktop apps. It wasn't so much an kernel or OS design limitation (because the aforementioned Windows Store apps achieve compat w/ both architectures) but a conscious decision to move forward with the platform.

Desktop support is there for legacy reasons, but it is inevitable that eventually Windows Store apps will be the way to write Windows applications for the foreseeable future.

Reply Parent Score: 2

Morgan Member since:
2005-06-29

Now, the part you're likely complaining about is Desktop apps. It wasn't so much an kernel or OS design limitation (because the aforementioned Windows Store apps achieve compat w/ both architectures) but a conscious decision to move forward with the platform.


Sorry, I should have specified that.

Desktop support is there for legacy reasons, but it is inevitable that eventually Windows Store apps will be the way to write Windows applications for the foreseeable future.


I suspect this is the way OS X will end up too, if Apple continues on their current path.

I really don't like this push towards a tightly controlled, closed userspace on the desktop. It works well on mobile devices but in my opinion it doesn't scale well on business machines and power user workstations. But then, I'm from an older generation; I grew up with machines that were largely open to experimentation. Today's young engineers are brought up in a consumer oriented culture and are probably more accepting of an OS they can't hack around on.

Reply Parent Score: 2