Eric Raymond is wrong about the importance of 64 bit OSs

I usually find what Eric Raymond has to say interesting and entertaining, and I enjoyed 'The Cathedral and he Bazaar'. But in this recent interview, he talks about the importance of the transition from 32 to 64 bit OSs and how it creates a 'window of opportunity' to make the Linux desktop popular, that will only last until 2008.

Certainly all things being equal, 64 bits are better than 32 bits, but I personally doubt whether or not your desktop OS supports 64 bits is going to make much difference at all. I agree there is a window of opportunity ahead, but for entirely different reasons:

  • The cost of hardware is dropping rapidly, while the cost of MicroSoft software is not. It doesn't make sense to spend 300 Euros or less on a computer, and then 500 Euros for software to run on it. It also means that PC makers like Dell can't differentiate their offerings on CPU speed/memory size/cheapness anymore. PC makers will need to increasingly rely on good aesthetic design and tight integration with system software like Apple does today. If the PC market becomes like the iPod or mobile phone market, why would I want a badly designed noisy unreliable and hard to set up black box in my living room or workplace? You need more control over you software to integrate in the way Apple does, and you can get that by using Free Software.
  • We can assume every personal computer is always attached to the internet. There will be no stand alone desktop apps, and users will expect networked based features such as collaboration to be built into their software from the ground up. Why do I want to spend hundreds of dollars on word processor/spreadsheet PC software, when there are free ad supported web based alternatives with better collaboration facilities such as Writely?
  • Every device will have 3D graphics and 3D will be ubiquitous - we can assume it's always there. What if MicroSoft just use 3D for 'eye candy' and don't take the opportunity to change the way we interact with computers for the first time in 30 years? The desktop could soon be obsolete, except in Vista perhaps, and replaced by 3D 'shared worlds' like Croquet.
  • Unlike, the 32 to 64 bits transition, these three factors together will be highly disruptive in that MicroSoft won't be able to move easily to the new world without losing its revenue from their old stand alone shrink wrapped PC software cash cows. Here's what ESR has to say anyway:

    Q: Speaking of which, there have been multiple delays in the introduction of Microsoft’s Vista operating system. It sounds like you’ve got a big window of opportunity there that can’t last forever. How could you best exploit it?
    A: That’s right; my friend Rob Landley and I have done an analysis which we’re going to publish very shortly suggesting that there is a critical window of vulnerability for changing the dominant operating system. And that is probably going to close in 2008.
    Q: Wow.
    A: The reason we think that they will close them is because we’ve looked back at the history of the industry and we’ve seen that basically the only time that a dominant platform gets toppled is when the hardware platform changes out from under it, and the biggest driver of hardware platform changing out from under it is best with the changes. We saw one change, the 8- to 16-bit transition.
    Q: Yes.
    A: Another at the 16-to-32 bit transition, which was masked a little bit, because in that transition Microsoft succeeded in maintaining its incumbency, but they did it with a different software suite. And then there is a 32-bit to 64-bit transition going on now, which I think is going to be our best window for a long time to achieve majority market share, but the hardware trend curves indicate that the 64-bit transition will probably be over sometime in 2008, and that means that the market’s going to be making its collective decision about the dominant 64-bit operating system probably before that.
    Q: And will multimedia be one of the killer apps of a 64-bit desktop world?
    A: I think good support for multimedia is. It’s not a sufficient condition for 64-bit desktop dominance; there have to be other pieces in place as well. But I think it’s a necessary condition.


    I agree that ESR's argumentation is not completely right. If we have a look at the term disruptive technology, it does simply not apply to the ongoing transition to 64bits OSen.
    First, it's backwards compatible, so it's not all that disruptive. You can just keep running your ancient MS Office 95 version, even on a shiny new 64bits CPU.
    Second, the mere fact that it's 64bit has little advantage for the normal user, 64bits adds quite some overhead, and as long as you don't do long calculations or have extreme amounts of data, it doesn't buy you all that much.

    That said, let's take over the world!

    By sebas at Thu, 08/31/2006 - 16:37

    I don't agree if you talk about the amd64 one. That one is faster than the ia32 ones. Eight more int registers, some more fp/sse2 registers, fastcall-like calling convention (ie. less push/pop instr.) and no doubt other improvements. And no, not all pointers are 64bit, CPU makes smart use of 32bit relative pointers, and as such, still limiting app.'s address space to 4Gb.
    You might be right on the ia64.

    By koos vriezen at Thu, 08/31/2006 - 17:30

    Imagine you are selling very cheap computers to non-technical people, would the above description persuade them to buy a Linux based machine over a Windows one because Linux made better use of these great 64 bit hardware features? How would you explain the importance of 'eight more int registers' to someone who was thinking about whether or not the nice pink and white machine would match the decor of their living room better than the white one with the blue lights?

    By Richard Dale at Fri, 09/01/2006 - 17:42

    That was a response to the 64bit FUD that 64bit doesn't bring anything for the desktop, which is incorrect for the x64 arch.
    Selling points for non-geeks are of course not the number of registers on the CPU :-)

    By koos vriezen at Sat, 09/02/2006 - 07:26

    "Selling points for non-geeks are of course not the number of registers on the CPU"

    Well don't just tease us! What is it that is compelling and disruptive about 64 bit CPUs that even non-technical people would appreciate? Or are they just a bit faster, and moving to a 64 bit OS is just an evolutionary improvement as opposed to a disruptive change that MicroSoft won't be able to easily keep up with. It isn't that they don't bring anything at all to the desktop, just that it isn't as important as other factors in my opinion.

    By Richard Dale at Sat, 09/02/2006 - 12:32

    It was not my intention to comment on your blog, but here some thoughts then.
    I can now only think of the performance per watt or just the fact that these new generation runs much cooler and might then advertice their silence for living rooms or so, the fact that 32bit software don't benefit from this new arch and that PC's w/ 1Gb memory are already near the end of 32bit reach (forgetting the upper memory tricks).
    It's disruptive because one needs a new system and we finally might see the move to the living room for the masses. If this will change anything in OS land, one can only guess IMO. As we now already see, web, mm and games are dominating usage and linux does do it quite well in these area's. The consumer will decide our faith ..

    By koos vriezen at Sat, 09/02/2006 - 16:53

    Eric S. Raymond is wrong about everything else too. ;-)

    Wait -- is he saying that "2008 will be the Year Of The Linux Desktop!" Ooh!


    By Scott Wheeler at Thu, 08/31/2006 - 17:15

    The real deal is that in this perception, 2007 is *not* the year of the Linux desktop. That'd be the first "not year of the Linux Desktop" in something like 5 years. We can spend the time off hacking, my vote for aKademy 2007 goes to the Bahamas proposal.
    On a more serious note, for KDE, 2008 will be rocking, by the way. Hardware accelerated desktop (for eye-candy!) will be common, KDE4 has become a mature platform, application developers have caugt up and are now using that wealth of a new technology and Reiserfs is only two years away from being merged into the Linux kernel.

    The snowball effect (that "building on the shoulders of giants" thing) will have created a significant gap between Free Software and proprietary software. The proprietary world will have a real hard time to catch up by reinventing the wheel over and over.

    Good times to come!

    By sebas at Thu, 08/31/2006 - 17:28

    3D will be ubiquitous in the near future. Unfortunately, it will require proprietary drivers. No problem if you're using a Microsoft or Apple desktop. Big problem if you're using KDE or Gnome. Linux may have poorly supported and out of date proprietary drivers, but only if you're "lucky" enough to have the right hardware. And Linux isn't the only Free Software operating system. BSD and Solaris are getting screwed over, and tongue wagging jeers from the Linux side isn't a solution.

    3D is great, but until its usuable with 100% platform agnostic Free Software, it shouldn't be a requirement for Free Software desktops. Much more work needs to be spent on coaxing complete specs out of video card manufacturers. I shouldn't have to buy a ten year old card just to get usable OpenGL.

    End rant...

    By David Johnson at Fri, 09/01/2006 - 17:03

    So what do you think about Intel's recent announcement that they will be releasing their OpenGL drivers as Free Software?

    I think this is a disruptive event in that if Intel succeeds, the other vendors of proprietary 3D graphics cards will have little choice but to go along with what the main player has done. If the other vendors don't release Free Software drivers, just avoid their products.

    If MicroSoft's products don't make compelling uses of 3D graphics capabilities, then the graphics cards makers should collaborate with the more innovative Free Software community in order to help sell more of their graphics cards. I'm pretty sure Intel isn't a charity, and some pretty hard business thinking went into their decision.

    By Richard Dale at Fri, 09/01/2006 - 17:35