Categories

OS turning circles: Questioning Windows' maneuverability

[Updated with Mac OS versions. See footnote 3.]

I’m glad Windows 8 is named the way it is. With Windows 7 Microsoft went to a numbering system which is much more rational than the mixed naming of the past. The number 8 actually corresponds to the actual sequential number of major versions of Windows released to date.

Windows proper actually did not start with what was called “Windows 1.0″. Windows actually started in April 1992 when Windows 3.1 was released. It was the first Windows which was an operating environment onto itself, apart from DOS. It was followed by Windows 95 (which we can call “2”), Windows 98 (“3″), Windows 2000 (“4″), Windows XP (“5″), Windows Vista (“6″), Windows 7 and now Windows 8.

Given this nomenclature and the dates of general availability of said versions, we can derive a measure of the frequency of upgrades. For example Windows “2” followed about 41 months after “1” and “3” took 34 months after “2”. If we continue this for all the versions, and assume “8” will launch by October next year, we can plot the cycle times of new Windows versions.

To make the story more interesting I added the same data for other OS platforms. OS X, iOS and Android have version numbers which correspond to the sequential order in which they were released. I am assuming that the numbering system (1.0, 2.0, 3.0 etc.) are meaningful and that major releases are given a new integer value.[1]

I also added Windows Mobile which suffered from the same confused numbering that Windows suffered until the current Windows Phone 7. To create a map to integer values I used the following: Pocket PC 2000 is “1”, Pocket PC 2002 is “2”, WinMo 2003 is “3”, WinMo 2003 SE is “4”, WinMo 5 is “5”, WinMo 6 is “6” and Windows Phone 7 is “7”. There is no clarity when and what Windows Phone 8 will be so that is left off the chart.

The chart above does give a clue that Windows is a bit slower (more months between releases) than alternatives. However, the point I want to convey is that this cycle time can be a strategic advantage. A faster upgrade cycle means more rapid adoption of improvements, not to mention more revenues. Rather than a line chart, the better way to visualize it is as a turning circle.

The analogy is from aerial combat where a plane that can “turn inside” an opponent can get into a firing position. The tightness of turning circles also means that a competitor can iterate more quickly and thus move up the trajectory of improvements necessary to remain competitive more rapidly. To that end, I re-plotted the data above with radial axes.

The chart clearly shows how among desktops, OS X has been tighter than Windows[3]. Furthermore, mobile OS’s have been, typically, more agile than desktop OS’s. The exception being Windows Mobile which had significant architectural and business model challenges. It will be interesting to watch the radius of Android. So far it seems to be tight, but as tablet functionality has been added, the speed of change has slowed slightly.

However, the big contrast is between mobile and desktop. Microsoft tries with Windows 8 to merge the two. But these are different ecosystems. Buyers, OEMs, users, all have different behavioral patterns (jobs-to-be-done) for mobile OS vs. desktop/portables. Will the cycle time for Windows 9 and Windows 10 tighten up? Can Microsoft operate its OS division at this new cycle time?

What is even more bad news is that the most important customers for Microsoft, enterprises, actually upgrade every other version. The Windows “radius” is nominally averaging about 35 months but large account adoption is closer to 60 months.

The contrast is then striking: Consumerized devices with over-the-air updates on a 12 month cycle are five times more agile than a traditional corporate Windows desktop. Another way to look at this is that for every change in a corporate desktop environment, the average user will change their device experience five times. Although Microsoft might find comfort in Enterprises’ leisurely pace of change[2], those are the wrong customers to keep happy going forward.

The implications are left as an exercise to the reader, but I would argue that along with the asymmetry in business model around pricing that I highlighted in the last Critical Path podcast, Microsoft has a significant operational asymmetry with devices as a new class of computers.

Notes:

  1. The dates of general availability of various versions of operating systems.
    Version Windows OS X Windows Mobile iOS Android
    1 4/6/92 3/24/01 4/19/00 6/29/07 10/21/08
    2 8/24/95 9/25/01 10/1/01 7/11/08 10/26/09
    3 6/25/98 8/24/02 6/23/03 6/17/09 2/22/11
    4 2/17/00 10/24/03 3/24/04 6/21/10
    5 10/25/01 4/29/05 5/9/05 10/1/11
    6 11/30/06 10/26/07 2/12/07
    7 10/22/09 8/28/09 11/8/10
    8 11/1/12 7/20/11

     

  2. While at Nokia’s Enterprise Solutions I was witness to this difference between the cycle time of devices and the pace of adoption of Enterprises. The fundamental disconnect was that the sales cycle for a “solution” was longer than the shelf life of a device upon which that solution was predicated. Needless to say, Nokia Enterprise Solutions did not last long. Nor do we see today any significant Enterprise-focused mobile device businesses (RIM itself moved to consumerize its products years ago.)
  3. I added Mac OS data to the second chart to also illustrate that no company is immune from spinning out of control. Mac OS clearly went spun out after version 6 with increasing delays leading to its demise and replacement. You can also note that its replacement (OS X) was able to straighten-up and fly right. The lesson may be perhaps that ditching and starting over is a viable option.
  • David Emery

    The Turning Cycles plot visualizes the OODA Loop decision cycle from Col John Boyd (see http://en.wikipedia.org/wiki/OODA_loop and http://www.fastcompany.com/magazine/59/pilot.html )

    What I think needs to be studied is how the -ecosystem- impacts a company’s ability to implement this agility. Generating lots of products that don’t have the rest of the ecosystem could well be another variation on a death spiral. One of the strategic decisions attributed to Steve Jobs was substantially reducing and focusing the number of products/product lines from Apple. One can argue that part of RIM’s problem is too many unfocused products.

    To me the big revelation when I first heard about Boyd’s OODA methodology is the inclusion of the ‘Orient’ step. This is where I think the “art” exists, it’s the ability to rapidly extract signal from noise, apply more filters, to get enough signal to make a decision (within acceptable risk.)

  • Anonymous

    Interesting visualization. The cost for desktop upgrades is typically much higher than mobile OSes (typically free, unless new hardware is a requirement), so the longer time between releases may increase the perceived value in the eyes of the consumer and thus allow for higher price point.

    While reading about agile software development, the writers promoted frequent releases for a variety of reasons. The most interesting for this discussion is that sitting on unreleased software is like sitting on inventory. You have created something with value, and if you don’t release it, it’s like inventory / work-in-progress that is sitting in a warehouse collecting dust. Think of Toyotas lean manufacturing compared to GM or Ford. The lean mentality flows through the whole organization and provides very real financial and strategic benefit.

  • http://www.digitalglen.com John Blackburn

    That radius chart—and the aerial combat analogy—is among the finest charts I’ve seen in years. Excellent!

  • Anonymous

    I was expecting Android to show a bunch of releases until I remembered that they have tons of new phones not new software. I’m curious if in mobile those two can be separated. Apple’s plan seems to be to release software faster than hardware and make old hardware do new things (iPhone 3GS, still on sale, started with iOS 3 and will be supported by iOS 5).

    • http://twitter.com/qka qka

      Also, Mr. Dediu is counting integer releases, where Android counts fractional releases as “big things”, with their fanciful sweet tooth names. Android 2.0. 2.1, 2.2, etc. And each of those fractional releases is not compatible with its predecessor.

      • Anonymous

        I think this is accurate. Version numbers are somewhat arbitrary. In Mac OS X world, a 0.1 bump in version number is a major update.

      • http://www.asymco.com Horace Dediu

        We can always argue that fractional releases are significant. Many have argued that OS X releases are less “meaningful” than Windows releases. Unfortunately I cannot second guess the decisions being made on the meaning of release numbers. I have to trust that the teams deciding on these numbers do impart some meaning to major releases.

      • bk

        Not to mention that service packs on Windows were sometimes close to a Major release, for example, windows xp sp2. Might be interesting to map these releases to sales figures.

      • http://www.intomobile.com/ Stefan Constantinescu

        Totally agree. Lots of companies don’t adopt a certain version of Windows until they issue the first service pack, which as “bk” pointed out, can sometimes re-imagine what the initial product is all together. I’d argue that Windows XP Service Pack 2 compared to Windows XP is what Mac OS X 10.6 Snow Leopard was to Mac OS X 10.5 Leopard.

  • Anonymous

    I’m not sure it’s that big of an issue. Corps love keeping OSes for ever, since it is cheaper. Let’s face it, OSes are not that important, they’re here to launch apps, drive peripherals, and supply a graphical shells. I really don’t know what 7 lets me do that 2k didn’t. Consumers OTOH are very aware of the eye-candy part, and love bleeding edge peripherals and software.

    Anyhoooo, Corps have been relying on old versions of Windows forever (witness the stampede each time a version reaches End-of-Support). They can keep doing so whatever MS’s OS release rythm. MS just have to keep supporting Windows versions for 5-10 years like they’re doing now, happily, for a price. It just means MS will have a few more versions on support at the same time. A drop in the ocean, I’m sure.

    • Andrew

      Back in the day … when computing was editing a spreadsheet on a network drive … the OS didn’t matter so much. It’s a “little” different now, with all the new devices and web applications.

      • Anonymous

        Mmmm…
        1- given that one of web applications’ selling point is that they abstract the app one step further from the machine (Hardware > OS > Browser > App, instead of Hardware > OS > App), I find that statement strange.
        2- As for devices, indeed. Drivers are one of the main reason for upgrading an OS. Or not upgrading it actually, when driver architecture changes break drivers for old stuff that’s still in use. Those architecture changes don’t happen a lot, but my guess would be that for Corps, it’s more of a drag backwards than a push forward, on the whole. Most often, drivers work cross-version (Vista+7…) and are not really a factor.

    • http://www.asymco.com Horace Dediu

      What you describe is the notion that ancient OS’s are “good enough” for corporations. This seems to be the case and I made a point of saying this in my podcast. The point where Windows became good enough was in about 2000 when it finally reached a point of stability.

      But the premise of modern, mobile operating systems is that they compete on new bases: productivity or enhancement to jobs to be done while mobile. If companies decide that work processes tied to a desktop are the only ones they need to worry about in perpetuity then perhaps mobility is not valuable. Otherwise the changes that mobility has brought to consumers might be cause for re-thinking the desktop paradigm.

      • http://twitter.com/jackschofield Jack Schofield

        > The point where Windows became good enough was in
        > about 2000 when it finally reached a point of stability.

        This is wrong. Windows NT was extremely stable, certainly from NT 4.0 onwards, where people started using it as a server operating system.

        > Otherwise the changes that mobility has brought to consumers
        > might be cause for re-thinking the desktop paradigm.

        And that rethinking might involve using a single operating system that does both jobs well, and uses familiar languages and development tools, rather than adding one or more incompatible operating systems.

      • http://www.facebook.com/profile.php?id=100001578926066 Coward Anonim

        WNT was “stable” only in comparison to W95/98. In fact it had to be restarted at least once a month. This “stability” persisted (with small improvements) until W2003 SP2 (+few patches) when MS finally plugged all significant memory leaks, so system could easily survive 12 months (only if you did not apply any more critical patches, which forced you to reboot any way).

      • http://riverlaw.myopenid.com/ riverlaw

        I supported NT 5 for many different customers and disagree with your definition of stable. 2000 on the other had was choice. Maybe NT had better reliability during 4 with certain hardware, but I did not see with 5 across a variety of hardware.

      • http://www.asymco.com Horace Dediu

        The discussion is, of course, referring to desktops vis-a-vis mobile and therefore not server operating systems. Server operating systems were reliable in the 1960s era of mainframes.

  • https://launchpad.net/~mpt mpt

    The conclusion may be correct, but I think this is the first Asymco post I’ve seen where the data is far too incoherent to support the conclusion.

    According to Wikipedia, the first version of Windows that was an operating environment to itself was Windows 95 (though it still depended on DOS code for some things). But even if it was Windows 3.1 as claimed, that wouldn’t justify the conclusion that that was when Windows “started”. Windows 1.0, 2.0, and 3.0 all had real APIs that application developers targeted, and (except where Microsoft applications shipped with a Windows 1.x/2.x runtime) users therefore needed to upgrade to those versions if they wanted to use those applications.

    And that’s what you’re grasping at here: the speed at which operating systems improve — for example, the rate at which compelling applications start depending on new APIs — and the speed at which users upgrade their OS as a result. And that just can’t be reduced to integers. Plenty of data on upgrade speed is available (on the Android Developer site, and in Web stats, for example), but how much successive versions improve is practically impossible to measure numerically.

    For example, the table classifies Mac OS X 10.6 (“enhancements and refinements”, remember?) as a major version. But a year after its release, what proportion of applications had compelling new versions that worked on 10.6 but not 10.5? Certainly it was much less of a jump than 10.4 to 10.5, or 10.6 to 10.7.

    Similarly, Ubuntu (which I work on) has a major new version, like clockwork, every six months. But that doesn’t mean we have a tighter turning circle than all the platforms listed here. Because successive Ubuntu versions are usually much more similar to each other than successive versions of other systems.

  • marko

    A turning circle chart, maneuverability, and an aerial combat analogy?! All in a context that nails the concept. You have outdone yourself on this one… that’s some next-level analysis!!

    • sh

      ….lucky he didn’t draw it as major releases per year then isn’t it!

  • S. Ben Melhuish

    I think Windows Phone 7 “Mango” is equivalent to iOS 2.0, meaning that for these purposes we should consider it to be version 8. If sustained to version 3, it would be a dramatic shift in the OS’s orbit.

  • Canucker

    I think Horace is subtly suggesting another dichotomy here. That is with Windows 8 Microsoft is caught between the glacial progress of the desktop OS beloved of corporate standardization/support and the rapidly evolving mobile OS ecosystem. Since The iPad (other tablets are irrelevant to Microsoft) is on 12-14 month OS cycle, Windows tablets will either wave as they are lapped by their competition, or will have to expend enormous energy to drag their desktop counterparts kicking and screaming into frequent upgrade cycles. This will be a tricky tightrope to traverse. Of course, the mobile OS’s are likely to slow their cycles as they mature (vis a vis OSX), but the inherent differences between the tablet/mobile and desktop/laptop markets are not as easily blended as Microsoft is pretending in Metro/Windows 8.

    • Sacto Joe

      I’m convinced MS has no intention of dragging “their desktop counterparts kicking and screaming into frequent upgrade cycles”. Rather, their intent is rather more subtle. They intend to build two types of tablets, one x86, 64-bit based (Intel Atom) and one ARM based. The ARM tablet will only run the new touchscreen system (Metro), but the x86/64 will run both the touchscreen system and Windows 7. An “overlay” will then let you access your Windows 7 software via the touchscreen. The x86/64 system will cost a couple of hundred bucks more than the ARM-based system, but allows you to access your legacy Windos software on your tablet. In this way, they hope to “leverage” the huge advantage in the number of installed Windows apps that they have over iOS apps.

      The two problems with this approach are (1) a competitive Atom-based tablet is probably a year off, and in the meantime Apple is ramping up production like there’s no tomorrow, making full use of its first adopter advantage, and making it very difficult for MS to catch up on production; and (2) this is a “kludge” approach to accessing Windows 7 software on a tablet, since it doesn’t optimize the interface for tablet use. If it’s too much of a kludge on the tablet form factor, people won’t spend the extra money for it. And if enough Atom-based tablets aren’t sold, it won’t matter how fast they can ramp up production.

      Time will tell.

      • Canucker

        The problem with that analysis (which I agree with) is that Microsoft cannot fork the ARM tablet OS from the x86 OS. That would quickly deteriorate and make Android fragmentation look simple. iOS development is not tied to OSX development in terms of features and releases (they are clearly connected in other ways). Microsoft has declared their mantra as the best of both worlds. My feeling is that they may deliver the sum of all evils by trying to marry these self-declared singletons.

      • Anonymous

        They’re not going to fork it. Metro apps are built on HTML, CSS, and Javascript, and they’ll run on either ARM or x86 just fine.

    • http://twitter.com/grzegorz_maj Grzegorz Maj

      You have no idea what you are writing. Windows 8 is not evolutionary change, so number of months to release is almost meaningless. Apple plan to merge OSX and iOS have to be disrupted if they want to compete with Win8. If not, their next OS is obsolete. Win8 also negates app number advantage, because to achieve the same level of innovation(as is in Win 8) every app has to be rewritten using API that do not exist yet. And I do not believe that Apple has developer resources to create any(API). Praised here lower Apple’s R&D cost compared to revenue than Microsoft now can be real roadblock.

      • http://riverlaw.myopenid.com/ riverlaw

        Maj, I agree that win8 looks to be a big change, worthy of the point upgrade. I strongly disagree win8 will cause any sort of reaction from apple. Mac OS will die out over the years and iOS will take its place. Win8 does not help any app numbers against apple. To say this is crazy. People are going to have to create metro apps to be compatible with arm by and large. While iOS will keep raking them in. It may already be to the point where iOS has more relevant and cheaper software.

      • http://twitter.com/grzegorz_maj Grzegorz Maj

        If Apple do not react, its OSes will be perceived as outdated compared to Win8. Metro apps are easier to design, since only typography and spaces are important. they look cleaner, and they have consistent, easy to familiarize layout. Most important feature are contracts, it works like web app apis – one app can use functionality of other app. iOS and Android apps do not have this.

  • Rob Scott

    How often are they going to offer upgrades? How are they going to charge for them? Is it still possible to even charge for mobile devices upgrades? Maybe Metro upgrades will be free and Windows upgrades will be paid for, but how much? OS X is only $29.00. How are they going to package Office, and how much are they going to charge? Unbundle and charge $19? How are they going to anniversary their sales let alone grow the doing that? How much are they going to charge OEMs? $45 is too much for tablets?

    This is a huge disruption of Microsofts business. If Windows 8 succeed Microsoft will be in big trouble.

  • Pingback: Windows 8: BFD — Big Forking Decision | Monday Note()

  • Jonathan Polley

    Your numbering of Windows versions is incorrect. The non-NT versions of Windows (1.0-ME) all required DOS to be installed first. The numbering of Windows 7 is actually related to the Windows NT version. The version that we all know of as Windows is version 4.0. Windows XP is 5.0 and Vista is 6 (I believe that the Vista kernel reports itself as 5.1, but this is marketing).

    • Anonymous

      I agree, except even that’s failing to account for win 2k which was originally going to be Windows NT 5. Possibly Microsoft is trying to make us all suppress the memory of Vista.

      The fact is that Windows 7 wasn’t called ‘7’ because it makes logical sense, it was done to match up with the versioning in their mobile space where they were going from win-mo 6.

      At any rate this error is significant because it reduces one of the big gaps in the MS release timeline. Either the jump was from NT 4.0 to Win 2K ( the enterprise upgrade path ), or it was from Win 98 to XP (the consumer upgrade path). Technically consumers also had ME in the middle but almost nobody actually used that – certainly not willingly.

      • Michal Tomlein

        It’s not failing to account for that. Windows 2000 was in fact NT version 5.0. XP was NT 5.1, 2003 Server and XP 64-bit were 5.2. Vista was 6.0 and Windows 7 is actually 6.1.

        You can check in the About Windows dialogue.

        Not that it matters.

    • http://www.asymco.com Horace Dediu

      I did not say that Windows began when it was independent of DOS. I wrote that it was a distinct environment. That means from a user’s point of view this was the beginning of “using Windows”. But that is neither here nor there. The process for deciding which point number corresponds to which version is simply to take 8 and count backwards to 1 with a significant release for each number. That’s how I came to 3.1. It happens to match nicely to what many called the beginning of Windows–at least I remember it as such.

      • http://www.metamatt.com/ Matt Ginzton

        Windows 3.1 is significant mostly because it’s the first version of Windows that really took off in the market, so the first version that most people ever used. That’s the reason it stands out, not because it’s independent of DOS, not because it’s a distinct environment. It’s just the first version to succeed enough that people remember.

        (There are a bunch of wrong factoids in this thread about the subsequent Windows versions in Microsoft’s own versioning. Windows 98 was 4.0, Windows NT was also 4.0, Win98 and WinME are forgettable minor versions, Windows 2000 is 5.0, Windows XP is actually internally seen as a minor version 5.1, Vista is 6.0, and “Windows 7″ is, amusingly, 6.1.)

        Anyway, even though it’s arbitrary I don’t disagree with Horace’s remapping here — and it gives a nice result in that truly excellent turning circle chart at the end — but there is one deeper point to argue about starting the Windows timeline at Windows 3.1. Namely, backwards compatibility.

        By Windows 95 (version “2” in this way of telling the story), Microsoft already had a huge backwards compatibility burden, with DOS apps and older Windows apps (mostly written for an earlier 16-bit architecture). Later versions based on Windows NT retained compatibility with those. By now, 16-bit apps no longer run on modern Windows versions but still, Microsoft’s backwards compatibility window is something like 15 years and is much longer than anyone else’s (OS X 10.7 won’t run apps from 10.3, largely because there was a processor architecture change in the intervening time, but hey, that’s only about 5 years ago).

        Point being — and this is already well known, but worth reiterating in the context of this article — Microsoft historically likes to retain backwards compatibility even if it slows down their Windows release cycle; Apple likes to aggressively move forward even if it means cutting ties to things not that far in the past. I think this is an intentional choice that bears directly on those turning circles.

      • PF

        16 bit versions of Windows:

        Windows 3.0 that was the first “big” version of Windows that got a lot of popularity and hype, released May of 1990. It was hyped as the “Macintosh-like” version of Windows due to the nicer icons, even though it was still a split Program Manager / File Manager.

        Windows 4.0 was renamed to Windows 95 by the marketing team in early 1995 if I remember right, and that was the one super-hyped as being very close looking to a Macintosh, with a big introduction featuring Jay Leno, the Rolling Stones “Start Me Up’ song, etc.

        NT-based versions:

        Windows NT 3.1 Jul 1993
        Windows NT 3.5 Sep 1994
        Windows NT 3.5.1 May 1995
        Windows NT 4.0 Jul 1996
        Windows 2000 (5.0) Feb 2000
        Windows XP (5.1) Oct 2001

        http://en.wikipedia.org/wiki/Windows_nt#Releases
        or
        http://en.wikipedia.org/wiki/Microsoft_windows#Timeline_of_releases

      • http://www.smartlevers.com Rukesh Patel

        It’s worth noting that though Apple likes to move forward aggressively, it has managed the transitions with the least pain to its users (Motorola to PPC to Intel, MacOS to OS X, 16- to 32- to 64-bit, PC to post-PC).

        Whereas, despite the slower release cycle, Microsoft has created confusion and headaches for its users in their present (maintaining different consumer/enterprise versions for long, often with different UIs, 32-bit to 64-bit, requiring not only different binaries, but often with applications also priced differently for their 32/64-bit versions).

        With Steve’s absence at Apple, Microsoft briefly led in the transition to 32-bit in the mid-90’s but have since trailed in bringing the consumer masses to 64-bit computing in a seamless way.

        And now they’re trailing way behind in bringing their users to the mobile, multi-touch, post-PC world. Windows 8 and its ‘no compromise’ approach looks like an invitation to major headaches for its users having to manually negotiate between the X86/ARM, legacy/Metro, IE +/- Flash worlds.

        It’s like trying to build a car, a plane, and a submarine all-in-one. Great concept for a Bond movie. ILM’s special effects can certainly pull it off. Microsoft, on the other hand, would need a massive ‘reality distortion field’ to make it stick, despite its marketshare.

      • http://twitter.com/jackschofield Jack Schofield

        ROFLMAO

    • http://markk-b.sitesled.com Mark Kéy-Balchin

      No, Windows XP reports itself as 5.1. Windows Vista reports itself as version 6.

  • http://www.lazyprogrammers.com Eugene Kim

    Would it make sense for MS to do what Linux is doing by simply accepting the slow adoption rating of Enterprise users and split the consumer based OS with Enterprise following albeit more slowly and stable? And where’d Windows ME go? /jk

    • http://www.asymco.com Horace Dediu

      Me was not a significant release. It was a stopgap one year crash effort to bridge 98 before XP would ship. It shipped the same year as Windows 2000. See “History of Windows” on Wikipedia. http://en.wikipedia.org/wiki/History_of_Windows

  • Anonymous

    I haven’t heard enough about the new Windows 8 pricing schemes. Considering a competitor in Android is available for free, and the market leader sells their product cheaper than anyone else, what is MS going to price Windows 8 at for the tablet makers? Will they sacrifice profit margins so that their HW partners can price their tablets decently? Will the additional tablet sales make up for the lost profits on desktops/laptops? Or will they price the ARM and Intel editions differently?

    • unhinged

      Microsoft has traditionally protected its own profits while allowing its hardware partners to absorb the cost of making hardware cheaper and cheaper for the consumer. Looking at the mobile space, MS is doing the same thing (licensing fees from HTC, etc) and I really don’t see them changing this approach. If anything, they are trying to position Android as being more expensive than Windows Phone 7 because of the former’s patent issues.

    • unhinged

      Microsoft has traditionally protected its own profits while allowing its hardware partners to absorb the cost of making hardware cheaper and cheaper for the consumer. Looking at the mobile space, MS is doing the same thing (licensing fees from HTC, etc) and I really don’t see them changing this approach. If anything, they are trying to position Android as being more expensive than Windows Phone 7 because of the former’s patent issues.

  • http://twitter.com/adriancjr Adrian Constantin

    Even if Microsoft decided that it is worth making their enterprise unhappy by releasing the OS more often, changing the release cycle of the OS is not easy. I think that the processes, software design and cost structure of the OS development are linked closely to the release length. Linux kernel releases changed from very long release cycles to 4 releases per year, but that did not change the Debian distribution cycle dramatically. It needed a whole new team and new ways of working to adopt a faster release cycle for Ubuntu. For Nokia it took about 3 years to learn how to make the release cycle faster (see pre-S^3 vs. Anna and Belle). When they finally pulled it off, it was too late. Maemo/Meego did not even have the chance to learn it before it got axed.

    The difference between long release cycles and short release cycles is like the difference between walking and running. You will use different muscles, different body posture, different ways of training and possibly even a different diet. Running fast is not just a faster walk.

  • https://launchpad.net/~mpt mpt

    The conclusion may be correct, but I think this is the first Asymco post I’ve seen where the data are too feeble to support the conclusion.

    Even if Windows 3.1 was the first version that appeared to be a distinct environment (and I’m pretty sure I remember PCs starting up automatically into 3.0), that wouldn’t be relevant. Windows 1.0, 2.0, and 3.0 all had APIs that application developers targeted, and (except where Microsoft applications shipped with a Windows 1.x/2.x runtime) users therefore needed to upgrade to those versions if they wanted to use those applications.

    And that’s part of what you’re trying to measure here. The speed at which an operating system, as used in the world, improves — or challenges competitors — depends not just on (1) frequency of new versions, but also on (2) how quickly users upgrade to each version (e.g. problems with handset vendors neglecting to issue Android updates), (3) how much changes in each version, and (4) how quickly compelling applications or peripherals start requiring the new version.

    You have, here, data for the first of those. Data for the second is obtainable, though patchy (on the Android Developer site, in WWDC keynotes, and in Web stats, for example). But the third and fourth are practically impossible to measure numerically. And they certainly can’t be reduced to integers.

    For example, the table classifies Mac OS X 10.6 (“enhancements and refinements”, remember?) as a major version. But a year after its release, what proportion of applications had new versions that worked on 10.6 but not 10.5? Certainly it was much less of a jump than 10.4 to 10.5, or 10.6 to 10.7.

    Similarly, Ubuntu (which I work on) has a major new version, like clockwork, every six months. But that doesn’t mean we have a tighter turning circle than all the platforms listed here. Because successive Ubuntu versions are usually much more similar to each other than successive versions of other systems.

    It may be true that, as you replied to qka, you “cannot second guess the decisions being made on the meaning of release numbers”. But for these charts to be meaningful, I think you would need to.

    • Kizedek

      “For example, the table classifies Mac OS X 10.6 (“enhancements and refinements”, remember?) as a major version. But a year after its release, what proportion of applications had new versions that worked on 10.6 but not 10.5? Certainly it was much less of a jump than 10.4 to 10.5, or 10.6 to 10.7.”

      On the contrary, 10.6 Snow Leopard is much more significant than you make out. One might think the was little change in their applications if you were already running relatively new Mac hardware, but for those with PPC Macs (I run one, still going strong at about 10 years of age), 10.5 Leopard was the end of the line. Many, many of my most used apps require an Inte Mac, and this shift to more modern, Intel-only apps coincides with 10.6.

      Apple generally understates its products, and referring to 10.6 as mere “enhancements & refinements” was intentional. It reduced the tendency for fickle persons to be disappointed at the lack of “eye-candy” and sensational features accompanying this major update, by which they might normally make their decision. Reference all the arguments about adding “lipstick to a pig” regarding Vista and other Windows updates… Updates that are hyped ad nauseum by MS a year or years in advance, only to fall short of expectations. The very reason that many posts right here are noting that XP is still good enough for most Windows users. No real advance has been made.

      These mere enhancements in 10.6 actually paved the way for 10.7 and future updates. Some real underlying changes were made. For example, a large portion of the OS and core apps were made fully 64-bit.

      Apple both under hyped it and lowered the usual price, so there was little excuse to upgrade if one could. Indeed, the last couple of versions of OS X have seen great, early adoption rates, with a huge percentage of the installed base upgrading. Of course, 10.7 requires prior updating to 10.6, unlike earlier versions — so that right there belies your statement that 10.6 was insignificant.

      All this, just highlights another way in which Apple’s approach is so different than MS’. I think that there is justified fear that Metro is simply more lipstick on a pig.

      • Kizedek

        Regarding updated applications: if you are using MS and Adobe as a measure, don’t. Mac users interested in real productivity from their apps knew exactly what they were getting from Snow Leopard 10.6. Modern Mac app developers are running rings around Adobe and MS software, in the same way that OS X and iOS are running rings around Windows. I have moved from PhotoShop to Pixelmator due to the developer’s commitment to the future and their faster upgrade cycle.

      • https://launchpad.net/~mpt mpt

        Dropping PowerPC may have been an example of “maneuverability” in the technical sense, but not obviously in the strategic sense of improving or challenging competitors. To use an example from the distant past, Netscape 4.0 to 4.07 could run on Windows 3.1. Later 4.x versions dropped that ability, running only on Windows 95 or later, but Netscape 4.x nonetheless became one of the most infamous examples of unmaneuverability in the history of software.

        To return to the four factors I listed, Apple’s move (3) improved Mac OS X technically by making it smaller and faster, but conversely (2) reduced the speed at which people could upgrade (since many more of them than usual needed to buy new Macs to do so).

        Vista isn’t relevant to this point, and neither is the requirement to use 10.6 to upgrade to 10.7 — except that the latter again reduces the speed at which people can upgrade.

  • Eric

    Horace, wouldn’t you think starting Windows at 3.1 but Mac OS at X is a somewhat unclear comparison? You’re talking about the entirety of the Windows evolution with version 10 of the Mac OS. While I don’t think it would be less oranges-to-apples if you started from Windows 2000, or on the flip-side if you counted Mac OS major version increments instead, or in addition to, the OS X increments, the current method seems flawed in its omission of all Mac releases before OS X as well as Windows service packs.

    That the Windows release dates start in ’92 but the Mac ones begin in ’01 belies the problem.

    Eric

    • unhinged

      There was such a significant change between Mac OS 8/8.5/9 and Mac OS X that I think Horace’s use of the latter is both accurate and acceptable. They really are/were two separate technologies; “Classic” Mac OS was originally a single-user system created for much more primitive hardware and reached the point where it could not grow to incorporate new technologies (see the issues Apple had in the mid-90s) whereas Mac OS X is a modified version of NeXTStep.

      Should NeXTStep be included in the numbers? Given the number of users it had before being repurposed as Mac OS X, I don’t think there would be any value in showing the early dev cycles, similar to the pre v3.1 versions of Windows.

      • http://twitter.com/jackschofield Jack Schofield

        > They really are/were two separate technologies

        Windows (DOS-based) and Windows NT were also two separate technologies.

        One major difference is that Microsoft just started the move a lot earlier than Apple (which got bogged down with a version of Mac OS it couldn’t finish and therefore bought in something to replace it). Indeed, at the time, Apple (specifically, Ellen Hancock) discussed licensing NT from Microsoft.

        > similar to the pre v3.1 versions of Windows.

        Windows 3.0 sold more than 10 million copies, which is comparable to some versions of Mac OS and Mac OS X. It’s not an order of magnitude different.

        Windows 3.0 and Windows 95 were major upgrades. Windows 3.1 was a minor upgrade.

    • Anonymous

      This is how things match up between Apple and Microsoft systems:

      • command line PC system — Apple DOS (1977) and PC-DOS (1981)

      • command line PC system with optional graphical shell — Apple ProDOS (1986) and MS-DOS with Windows 1.0 (1985)

      • graphical PC system with bitmap graphics in the kernel — the original Mac (1984) and Windows NT 3.1 (1993)

      • PDA system — the original Newton (1993) and Windows CE 1.0 (1996)

      • enterprise graphical Unix system — NeXT (1988) or A/UX (1988) and Cygwin on Windows NT (1995)

      • graphical PC system with 2D/3D graphics — Mac OS X v10.0 (2001) and Windows Vista (2007)

      • PC system with HTML5 aka WHATWG — Mac OS X v10.3 (2003) and Windows 8 (?)

      • desktop class mobile system — iOS v1.0 (2007) and Windows 8 (?)

      • touch PC system — [?] (?) and Windows 8 (?)

      Notice how Windows 8 is all of the last 3, that is its all-in-one-go catch-up nature showing. Microsoft is attempting to catch up with HTML5, iOS, and possibly the touch Mac if it arrives before Windows 8.

      Notice the frenzy Microsoft went through to answer Mac OS X v10.0 with Vista, throwing out an “XP 2″ release entirely in 2003 and starting over and making something “lickable.”

      Notice how IE9 is Safari 1, a download for the current system that ships with a nonstandard IE; and IE10 is Safari 2, the default browser on a Web standard system that features standardized Web code not only in the browser, but also within native applications.

      Notice how a graphical shell extends the life of the Apple II and MS-DOS command line PC’s with Apple II GS and Windows 1.0, during the early days of graphical computing. If you look at screenshots, you see the same non-square pixel graphics, they are contemporaries.

      Notice how the Mac and NT 3.1 are both ignored by the respective command line communities. You could still buy a new Apple II 8 years after the Mac shipped. You could still buy a new version of MS-DOS with Windows shell 8 years after NT shipped.

      Notice how Microsoft’s first DOS runs on the IBM system that was built to replace the Apple II (with VisiCalc) systems that IBM’s mainframe customers were all crazy about. That is Microsoft making their bones as an Apple cloner. They are famous for that deal. They found themselves with that deal.

      So you see Microsoft tracking between around 5 and 10 years behind Apple the whole way. That is why Paul Thurrott was knocking full-face touch phones with soft keyboards and mysteriously missing Copy/Paste functionality in 2007 and embracing them whole-heartedly in 2010.

      • deV

        Considering you got 10 likes, I just have to say that what you said….

        – PC system with HTML5 aka WHATWG — Mac OS X v10.3 (2003) and Windows 8 (?)

        …is all wrong. The working group WHATWG didn’t even begin drafing the many-years-in-development HTML5 standard until 2004…and they’re not done yet. Apple must time warp huh? Is this like that thing where 34% of iPhone 4 owners believe they actually have a 4G phone?

        http://caniuse.com/#eras=now&cats=HTML5,unoff&nodetails=1

        It really makes sense to look at browsers because over half of Windows users (aka over 40% of all computer users in general) do not use the built in browser. As far as browsers, Google leads the way in compliance, with Firefox close behind, then Safari. As they should considering Google’s Ian Hickson is the author of the HTML5 standard.

        Cygwin is not a GUI either, and there are really too many errors to continue.

    • http://www.asymco.com Horace Dediu

      I would consider adding Mac OS (System 1 -7 and Mac OS 8-9) as a separate thread but OS X is a fundamentally different OS. As different as Windows CE is from Windows.

      • http://twitter.com/jackschofield Jack Schofield

        And as different as Windows (DOS-based) is from Windows NT (New Technology) though you appear to be astonishingly ignorant of the facts… or in denial.

        Yes, I did already make this point…..

      • http://www.asymco.com Horace Dediu

        Unlike OS X, NT-based operating systems maintained a consistent backward (binary) compatibility with DOS-based operating systems.

        Microsoft went to great lengths to transition their users smoothly between two architecturally different OS threads. Their intention and investment and insistence was that “Windows” was cohesive and consistent and cost-effective given prior investments in software. This is the core of the Windows brand. A brand which Microsoft protects dearly. Furthermore, the time frames for Windows are often so long exactly because of the effort to preserve this compatibility. Had Microsoft indeed forked Windows (as you suggest they did with NT) then their agility would have improved (which it did not).

        In contrast, Apple made a clean break between Mac OS and OS X, both in branding and in architecture.

        This is the asymmetry of the two approaches–preserve compatibility at the cost of cycle time vs. orphaning at the benefit of rapid iteration–both of which have benefits. And penalties.

      • http://twitter.com/jackschofield Jack Schofield

        > Microsoft went to great lengths to transition their users
        > smoothly between two architecturally different OS threads.

        Sure, but they were different operating systems with different development schedules and release dates. This is the *fact* that you seem oddly unwilling to face…

        > Had Microsoft indeed forked Windows (as you suggest
        > they did with NT)

        I never even remotely suggested any such thing.

        > Windows are often so long exactly because of the effort
        > to preserve this compatibility.

        That’s an unsupported assertion for which I don’t think you have any evidence. Of course, if you have, then you are welcome to provide it.

        Meanwhile you continue to ignore real issues that do affect the time scales of Windows releases, which I have already pointed out. These include:

        (1) the long beta-testing schedules required because of the size and lack of homogeneity of the Windows hardware base;

        (2) the need to involve many thousands of software and hardware partners in preparing to move to a new operating system;

        (3) the long testing procedures for in-house apps and the cost to businesses of rolling out new versions of Windows, which means they are not going to do it every year. Many of them won’t even do it every three years.

      • http://twitter.com/jackschofield Jack Schofield

        > Microsoft went to great lengths to transition their users
        > smoothly between two architecturally different OS threads.

        Sure, but they were different operating systems with different development schedules and release dates. This is the *fact* that you seem oddly unwilling to face…

        > Had Microsoft indeed forked Windows (as you suggest
        > they did with NT)

        I never even remotely suggested any such thing.

        > Windows are often so long exactly because of the effort
        > to preserve this compatibility.

        That’s an unsupported assertion for which I don’t think you have any evidence. Of course, if you have, then you are welcome to provide it.

        Meanwhile you continue to ignore real issues that do affect the time scales of Windows releases, which I have already pointed out. These include:

        (1) the long beta-testing schedules required because of the size and lack of homogeneity of the Windows hardware base;

        (2) the need to involve many thousands of software and hardware partners in preparing to move to a new operating system;

        (3) the long testing procedures for in-house apps and the cost to businesses of rolling out new versions of Windows, which means they are not going to do it every year. Many of them won’t even do it every three years.

  • http://twitter.com/elizabethrendon Elizabeth Rendon

    This is very interesting article comparing the OS cycle amongst the OS’s: Microsoft Windows 3 thru 7, Apple, and Andriod. One thing that I highlight from this article is that competitors: Google Android, Apple and Microsoft learn from each other. Such Microsoft and Apple have the older OS’s, Google just recently came out with Android and they [Google] made their OS better because they learn from their competitors. They do not get enough input from the users, but in contrast they compare their products from their competitors. Microsoft has longer agile OS because Microsoft’s OS enviroment is the first “onto itself” apart from DOS. I disagree that Microsoft is loosing customers because “enterprise” new version releases but because competitor’s products have improved. The one thing other competitors cannot take away from Microsoft is the user experience heredity. Microsoft is the first to open the doors to user experience UI, and moreover developer’s API experience.

  • http://twitter.com/jackschofield Jack Schofield

    > Windows actually started in April 1992 when Windows 3.1 was released.
    > It was the first Windows which was an operating environment onto itself,
    > apart from DOS.

    This is wrong. The first popular version of Windows was Windows 3, which came out in 1990. The first version that was integrated with DOS (but still based on DOS) was Windows 95. The first version that was an entirely different operating system was Windows NT 3.1 in 1993.

    Your numbering system is totally crazy, exactly like confusing Mac OS 6 and OS 7 with versions of Mac OS X (which started with NextStep, if we’re being picky). There were separate release cycles for consuser and business verisons of Windows until Microsoft managed to shift consumers off the DOS version with XP.

    Generally, Windows was on a 3-year release cycle (both versions) but minor upgrades were often done in the same sort of timescales as Mac OS X upgrades, which are generally far more incremental. Examples include NT3.1 to 3.5 to 4.0 and Windows 98 to 98SE to Me. In fact, like Intel, Microsoft tends to alternate major and minor updates….

    There are two other things you should have taken into account. The first is the size of the market, which governs the speed at which it can be upgraded. (There are more than a billion Windows PCs in the world, compared with 55 million Macs.) The second is the amount of testing and ecosystem-development that has to go into a release of Windows, which can involve 12-18 months in beta. (Microsoft can’t develop in secret then chuck it over the wall with a Stevenote.) After release, large companies typically spend 18 months testing in house apps before they start a roll-out. The idea that they would upgrade to annual releases is silly.

    • http://www.ratdiary.com spragued

      I wish I could “Like” this comment a hundred times. It’s far more valuable than the post it accompanies.

      • http://twitter.com/jackschofield Jack Schofield

        Thanks ;-)

    • Alan

      “The idea that they would upgrade to annual releases is silly.” – This is really the crux of the argument I think. iOS and Android will certainly maintain a release cycle close to a year while the desktop OS cycles are longer. So MS is lumping it’s tablet OS in with the desktop OS and not with the mobile OS. This is the asymmetry that is intriguing – even with a compelling tablet offering can MS change it’s methods enough to compete in the current tablet market, or will the market change by force-of-will of MS?

      • http://twitter.com/jackschofield Jack Schofield

        Yes, that’s an extremely good point, and none of us knows the answer. However, there are a couple of pointers.

        First, the underlying operating system is an improved Windows 7, which is arguably far more powerful and more capable than iOS or what’s available via Android (where you don’t get all the raw power of Linux). So it seems to me that Windows 8 will not be in the position where it needs rapid updates: it’s already mature, and it’s not starting from a low base.

        Second, the Metro UI is infinitely (and very simply) malleable. Anything you could write as a web app you could write as a Metro app (using HTML, CSS, DOM, and JavaScript), though you can also write Metro apps in C/C++/C#/VB etc. In other words, Microsoft could update Metro simply by releasing apps. As far as I can see, you wouldn’t actually need to update what you might call the “Metro OS” unless you were changing WinRT, and you could do that by updating .Net (since ultimately WinRT is just another instance of the CLR).

        I don’t think the average Windows user sees a new version of .Net as an operating system upgrade, though it is.

    • http://twitter.com/asaunders Alec Saunders

      And I can clearly remember writing code for applications running on Windows 1.03, circa 1986.

      • http://www.facebook.com/georgecreedle George Creedle

        DFX Composium was a 1.03 application. IIRC, windows didn’t overlap and there was no multitasking.

    • Anonymous

      Funnily enough, though, NT does start at 3.1, so his math still works.

      Starting at DOS 1.0 makes sense, or at Windows 1.0, or at NT 3.1.

      I disagree that Microsoft inherently can’t do yearly releases. I think if they can’t, then what good are they? That is slow today. It is only a question of good management. The mistake you are making is thinking Windows has to be big and monolithic simply because Microsoft built it that way.

      OS X is split into many tiny projects. Most of Mac OS v10.7.1 and v10.7.2 will be the same. But the key is, by shipping v10.7.2, Apple is asserting that whatever you have in there, that is the latest of everything. And by shipping every quarter, they are showing that should someone create a commercial malware for Mac OS, it will have a shelf life of only one quarter, because a new version that is immune to that exploit will be shipped by then.

      The above level of responsiveness is the standard in tablets, iPods, and high-end PC’s, where 3 out of 4 devices is an Apple device. Those are the markets Windows 8 has to conquer, because tablets are replacing the low-end PC’s that Windows rules today, and Consumer Electronics is so much bigger than PC’s. New software every 18 months is irrelevant. Who will remember? Software has to be alive today. Fresh.

      • http://twitter.com/jackschofield Jack Schofield

        > I disagree that Microsoft inherently can’t do yearly releases.

        I didn’t say they couldn’t. The problems are more to do with the kind of extensive testing needed to support a mixed infrastructure of more than a billion PCs, and the rate of adoption by companies. Neither of these is a problem faced by Mac OS X or iOS.

        > The mistake you are making is thinking Windows has to be big
        > and monolithic simply because Microsoft built it that way.

        The mistake you are making is to think I said this when I said no such thing. However, you are welcome to explain prcisely how monolithic Windows is compared to Mac OS X. Please give five or six examples.

        > it will have a shelf life of only one quarter, because a new version that
        > is immune to that exploit will be shipped by then.

        I don’t understand what you are trying to claim here. According to NetMarketShare. Mac OS X 10.5 still has a makret share of 1.17% which compares with 3.46% for 10.6. So, roughly 25% of the Mac user base is still on a version launched somewhat more than one quarter ago.

        Mac fanboyism ignored. You’re making a lot of assumptions when you don’t know what Microsoft is going to do.

      • Kizedek

        “> it will have a shelf life of only one quarter, because a new version that
        > is immune to that exploit will be shipped by then.

        I don’t understand what you are trying to claim here. According to NetMarketShare. Mac OS X 10.5 still has a makret share of 1.17% which compares with 3.46% for 10.6. So, roughly 25% of the Mac user base is still on a version launched somewhat more than one quarter ago.”

        I think what he is saying is that Apple manages a fairly rapid and orderly update cycle: a major release about every 18 months and a healthy number of 10.x.x updates in between, say at least one per quarter. If there happens to be something that Apple needs to address urgently, like a security risk, then it is slipped in there and you know where you stand. Most Mac users do indeed avail themselves of these 10.x.x updates — because it is simple to take a few minutes and let Software Update run and update your system over the internet.

        With others copying Apple in various ways, one wonders why this aggressive commitment to its users and to details and infrastructure is not one of the things replicated.

        MS seems to be at once both more talkative than Apple concerning it’s future roadmap, and less predictable at the same time. That companies stick to XP and refuse the latest “innovation” to come out of Redmond not only says something about how companies need to take their time to test and roll out updates, but it also says something about how MS largely got those changes wrong — despite the extra time it took them to finish the updates….

        Who knows what direction MS is going in now? The Enterprise apparently thinks Metro is not the direction for them. I guess we’ll find out when and if MS gets there sometime late next year.

        Oh, the reason that there are a number of Macs still running 10.5 is that there are a large number of PowerPC Macs still in daily use. I have a dual G4 PowerMac still going strong after about 10 years.

    • http://www.asymco.com Horace Dediu

      I am not confusing Mac OS 6 and OS 7 with Mac OS X. The data in the chart in the first footnote should confirm that I am referring to versions of OS X.

      You are right about the constraints on Windows. That’s exactly the problem I am trying to highlight. The constraints are there because Microsoft cultivated them for decades. Those constraints are also the core value proposition.

      • http://twitter.com/jackschofield Jack Schofield

        I didn’t say you were confusting Mac OS 6 and OS 7 with Mac OS X. I said you were confusing DOS-based versions of Windows with NT-based versions, which is the same kind of mistake. As I pointed out, this is wrong. Please try to read more carefully.

  • Anthony Starks
  • http://twitter.com/cdelrosso christian del rosso

    I really like the last paragraph of your post Horace.
    The difference in the traditional development cycles between the desktop and the mobile OSes versions creates significant challenges. Especially, when the two are supposed to be somehow merged as in the case of Windows 8.
    And those challenges are not only technical. As you highlighted, there are problems with the Enterprises sales cycles, operational, and I would add cultural within the same organization.
    Is the merging of the OSes (desktop and mobile) the right way to go? Ideally yes, but it is not an easy path.

  • Chris K

    David, I get the gist of your comment, but FYI just note that the “turning circle” chart is not really a visualization of the OODA loop concept. The OODA loop is traditionally spoken of as a decision-making tool or framework, whereas the turning radius charts are traditionally used to measure comparative fighters in terms of turning performance when studying adversaries. I think Horace was trying to measure company performance in issuing OS releases. The OODA loop concept doesn’t really fit with the idea of the comparative performance chart.

    Obscure info, I know! But just wanted to clarify the aviation terms. I flew and instructed fighters for a while.

  • Chris K

    David, I get the gist of your comment, but FYI just note that the “turning circle” chart is not really a visualization of the OODA loop concept. The OODA loop is traditionally spoken of as a decision-making tool or framework, whereas the turning radius charts are traditionally used to measure comparative fighters in terms of turning performance when studying adversaries. I think Horace was trying to measure company performance in issuing OS releases. The OODA loop concept doesn’t really fit with the idea of the comparative performance chart.

    Obscure info, I know! But just wanted to clarify the aviation terms. I flew and instructed fighters for a while.

  • Chris K

    David, I get the gist of your comment, but FYI just note that the “turning circle” chart is not really a visualization of the OODA loop concept. The OODA loop is traditionally spoken of as a decision-making tool or framework, whereas the turning radius charts are traditionally used to measure comparative fighters in terms of turning performance when studying adversaries. I think Horace was trying to measure company performance in issuing OS releases. The OODA loop concept doesn’t really fit with the idea of the comparative performance chart.

    Obscure info, I know! But just wanted to clarify the aviation terms. I flew and instructed fighters for a while.

  • Chris K

    Horace – excellent analogy. I never thought I’d see a common fighter analysis applied to the tech startup world. Yes, your chart fits pretty well in line with what fighter pilots look at when sizing up aspects of adversary turn performance. It can reveal weaknesses/strengths and suggest particular tactics you can use against them. Nicely done!

  • Chris K

    Horace – excellent analogy. I never thought I’d see a common fighter analysis applied to the tech startup world. Yes, your chart fits pretty well in line with what fighter pilots look at when sizing up aspects of adversary turn performance. It can reveal weaknesses/strengths and suggest particular tactics you can use against them. Nicely done!

  • Guest

    If this logic was compelling, Linux would be the leader. I think your analysis breaks down on numerous points. First, comparing releases as if they’re equivalent. They’re not. Second, assuming that “mobile” OSes won’t slow down as they mature. They will. The alternative for MS would be to have three OSes, WP7, WP8 and some tablet OS. Hard to see how managing three different code bases would speed agility. If ARM/Windows succeeds, which is currently tbd, presumably MS can innovate there far more rapidly than on Windows/INTC, since much less backwards testing is required.

    I think MS has an agility problem. I just don’t think this analysis is all that valid. And I guess we should point out that MS is still under sanctions with respect to what it can and can’t include in its OS. That’s not the case for Apple or Google, at least not yet.

    • Ian Ollmann

      At least in this country, the mobile phone upgrade cadence is roughly timed with the 2 year carrier contract. It is hard to imagine customers being excited about OS update cycles slower than that. It would lead to customers getting the “new phone – same OS” experience. I don’t think that would go over well.

      To the extent that Windows8 is intended to be both a mobile and desktop OS, Microsoft will have to reconcile the expected upgrade cycle for both. If it spends 5 years delivering Windows9 for mobile, that is a problem.

    • http://www.asymco.com Horace Dediu

      Linux, in the form of Android looks pretty good on this basis. I also noted how one mobile OS in particular (Windows Mobile) did slow down. I should also include Symbian and MacOS for completeness. However, that is not the whole story.

      Agility alone is not a sufficient determinant for success. It needs to be applied to a problem that needs fixing (a trajectory of improvement along a dimension meaningful to the majority of users.) A rapidly iterating OS which solves problems that few consider important does not make for a successful business.

    • Kizedek

      “Second, assuming that “mobile” OSes won’t slow down as they mature. They will.”

      But this is exactly the assumption that MS is banking on — that at a certain point, an OS (desktop or mobile) is mature and good enough. This is the MO of MS, this is what Horace is modeling and questioning.

      If you look at the history of Windows and think that is somehow natural, then you are going to think Horace is assuming too much. But Apple has delivered major desktop updates every 18 months, and major mobile updates every year. Apple has a 20-year plan, apparently! Apple motoriously holds features back until either the feature is ready, the hardware technology is ready, the ecosystem is ready, or until the user is ready! Apple doesn’t talk about its plans. With what confidence can you predict when OS X or iOS will have matured, peaked, or reached the end of their roadmap, or whatever? The agile turns of Apple are probably mapped out already. This is the strategic game that Apple is playing.

    • Anonymous

      A solution only matures if you keep the problem set the same. What is the guarantee that the problem set in mobile will stay stable long enough for the OSes to reach maturity?

  • http://twitter.com/j0el Joel Berman

    First off there are no good explanations of OS numbering. What makes a major release or a minor release is up to some optimization of upgrade revenue opportunities, new hardware support, wording in licenses and support agreements, and marketing.

    I think versioning has all changed. People will not be camping outside a store to get the next version of Apple or Microsoft OS. It will not be a revenue event. The big capability changes will occur on the cloud and as HTML5/CSS3 and their children advance, the browser becomes the important element.

    So your analysis is more interesting as history, but I do not believe it will be a predictor of the future. And changes will be much more incremental

    • http://riverlaw.myopenid.com/ riverlaw

      Or, no one is camping because you can download it on day one at home! I expect window to there too. I know that was not your main point. :)

    • http://www.asymco.com Horace Dediu

      Indeed. And what does that say about the sustainability (maneuverability) of the Windows franchise? Will Microsoft continue to obtain $x per device/PC/server/tablet/slate/phone?

  • Anonymous

    Would it be possible to align the turning circles so that the ends of the arrows are aligned? So instead of aligning the starting points, align the end points? I think that would make the current situation clearer. Not that it is difficult to read from the chart now, but I think it is more interesting to compare how the different OSes are doing right now as opposed to their respective points in history.

    Another interesting factor would be to look at the point updates because both iOS and especially Android have introduced many major new features outside the integer versioning. It also reflects the difference between desktop OS development cycles and mobile OS development cycles

  • Anonymous

    I’m not an Android developer (and neither a Windows dev) so can anybody explain what is the difference between Android services and Windows 8 contracts? I was under the impression that Android services are a way for one app to provide functionality for other apps to use, for example system wide custom keyboards. iOS does not have anything like that, though there are ways apps can communicate and work together. But it is quite rudimentary and I would welcome something more sophisticated and granular.

    • Kizedek

      I think many would say that Apple’ decision to sandbox apps is more a weighed decision than rudimentary. My understanding is that letting apps “provide functionality for other apps to use” contributes to the malware and security issues in the systems that allow it.

      It also means that Google and MS have less incentive to provide great APIs and services themselves. I think Apple typically introduces hundreds or thousands of new APIs each year, and finds secure ways to implement the sharing of it’s own core services such as location, contacts, copy and paste, etc. (BTW, copy and paste was notably absent from Windows Phone 7 first version, too, wasn’t it?)

      • Anonymous

        Of course Apple has made a conscious decision in how it implements inter-app communications, and not just based on security but also on the resource costs, UI and UX uniformity etc. But, the irony is that the current solution (URL schemes in conjunction with the clipboard) provides a very poor UX for anything happening between apps. It is not an easy task to solve, and Apple clearly has done the bare minimum there. Yes, URL schemes are very powerful and you could, in theory, build insanely complex inter-app interactions. But from the user point of view, apps jumping in and out of view is hardly the best possible solution.

        In my opinion, the current solution is very un-Apple: workable, quite clumsy and very unrefined. I’d expect more from Apple, that’s all. Maybe they are working on something more sophisticated, perhaps Windows 8 contracts will prompt them to do more, or perhaps that’s a part of iOS Apple thinks is less important than other areas.

  • Virgil Palanciuc

    Horace, your choice of major versions for Android is questionable, to say the least. I’d argue that each new “letter” is a major release, regardless of the version numbering. Thus, the major releases were Cupcake, Donut, Eclair, Froyo, Honeycomb, and now IcecreamSandwitch (to follow). Eclair & Froyo were 2.1 and 2.2 respectively, but you can’t skip them as “major releases”. Just look at how Google advertises their versions… it’s wrong to look just at the first version digit.

    • Virgil Palanciuc

      I meant “at the version’s first digit”, of course (can’t edit my post :( ).

      • http://www.facebook.com/valibirzoi Vali Birzoi

        Actually, to be even more confusing, Android version 3 (Honeycomb) is only for tablets. After Froyo, in Google’s smarphone OS environement comes Gingerbread, which is delivered as a new operating system, though it’s release number is 2.3 (Froyo is 2.2)…
        Probably (and hopefully) the two OSes will “meet” again in version 4 (Icecream Sandwich).

      • Virgil Palanciuc

        Forgot about Gingerbread.
        Anyway, it doesn’t really matter that Honeycomb is only for tablets. It’s hard to dispute that (a) it is Android (b) it’s a major version.
        BTW, there was a similar situation for a while with iOS – the iPhone & iPad versions of the OS were not in sync (wiht iPhone being more recent, in Apple’s case, if my memory serves me right).

      • Kizedek

        If by “similar situation” you mean the features of iPhone and iPad were not in the sync, then you would be correct. There wasn’t really a different version of iOS.

        iPad came out mid-cycle, between major releases of iOS (between 3.0 and 4.0, I think). Apple clearly said that iPad will not get certain features released to iPhone in a couple of the mid-cycle point updates. iPad would, however get them come 4.0 ….making 4.0, tada, a major release.

    • http://www.asymco.com Horace Dediu

      My choice of major versions is no more questionable than Google’s choice of major versions.

    • http://www.asymco.com Horace Dediu

      My choice of major versions is no more questionable than Google’s choice of major versions.

      • Virgil Palanciuc

        Which proves… what?
        Are you trying to say “I agree I compare apples with oranges, but it’s all Google’s fault!” Because if it’s so, I find your answer strange, to say the least.

      • http://www.asymco.com Horace Dediu

        I set the definition of “major version” to be the same as the definition of “major version” that the creators of the software chose. If you don’t agree that their choice is valid, there is nothing I can do to change it. I defer to their judgement. To the extent that judgement differs between companies that is in itself a matter of judgement. To that I defer to the reader.

      • Virgil Palanciuc

        BTW, I find it extra strange that you refuse to update the “major version” counting of Android, given that it completely supports your theory!

        There is little dispute, I hope, that Android started by being massively inferior to iOS, but caught up (now, while it may be considered to be ‘behind’ in a few areas, it is on par in most areas and even ahead in others). So, if Android started by being inferior, iterated slower, and is now roughly on par (if not ahead) iOS – theres something wrong, right? Either the iteration speed is irrelevant, or the way you counted “iteration speed” is flawed.

      • http://www.asymco.com Horace Dediu

        I would be especially reluctant to change the counting if doing so supports my theory.

      • Rob Scott

        Android is not disadvantaged here, its iteration speed appears fast. What is the problem. Its Google who decides that this is version 1, 2, 3 etc not us.

      • http://profiles.google.com/te.richardson Tim Richardson

        I think you got it right. Also with the Win XP service packs, which added scant new functionality in most cases (ok, sometimes they seemed to include a new version of IE). Haven’t been a Mac user long enough to know, but Snow Leopard to Lion was a big move, like Vista to Win 7.
        I don’t think you mentioned it in the article, and it’s tangential, but there interesting hardware changes in Apple’s development, which have no parallel.
        As far as turning circles go, nothing is as agile as Linux which provides multiple, rapidly evolving desktop experiences on a vast range of hardware, but neither Linux’s agility or Apple’s seems to have done much for market share gains. But Microsoft is now in a morass of complexity, and they’ve never been good at branding. They have almost no hope of explaining “Windows” when it’s one thing on mobile phones, another on a tablets and a third on PCs (presumably MS will bump the Phone os to “Windows 8″.). The brand is pretty dead, I would have thought. What value does it bring to the platforms where MS has no market share?

      • Rob Scott

        Apple has out grown Windows based PCs for five straight years. I would say their strategy is definately working.

      • deV

        Yet you’re ok, with calling all of Apple’s minor “point” releases as major ones. 10.1, 10.2, etc. Mac OS has been on version “X” (10) for 10 years!

        You’ve ignored several releases of Windows and even got the product line wrong (2000, XP, Vista, 7, 8 are all NT-line, not 95/98 line. They don’t even use the same code base.) You also choose to use Apple’s desired named major releases instead of the minor point number increases while not applying that to anyone else, including Android. Pathetic.

      • http://www.asymco.com Horace Dediu

        This has been discussed at length in the other comments here. I will summarize it again as a distinction between the way the companies choose to communicate significant change (i.e. the meaningfulness of that change). The decision on whether a version of Windows or OS X or Mac OS is significant enough is a marketing decision as well as a technical one. The marketing decision includes the branding of the release and how to communicate to users the value of upgrading to it. Note that meaningful updates come with a cost in license fees and costs associated with the challenges of migration.

        So the framework for deciding significance in a release is to ask why a marketer had chosen to place meaning in the brand and a price point in the license.

      • deV

        “Note that meaningful updates come with a cost in license fees and costs associated with the challenges of migration.”

        Google doesn’t charge for licensing. So by your definition they never have a “meaningful update”, ever.

        Also note that just because you haven’t done your homework on the Android front, doesn’t mean you can just make assumptions and have them be true. The Android API has changed with each of these named releases. When you create an app for Android, you have to choose an API version to target, and they match the Eclair, Froyo, Gingerbread, Honeycomb, etc releases. You also have to set a “minSdkVersion” if you want to use features that only exist in a newer version, which also hides your app from the market for phones that are running older versions and are not compatible.

        If you look at the spec sheets for Android phones they do list these versions. And throughout the Android ecosystem, these named releases are used everywhere. Just because you haven’t paid attention doesn’t mean you can write up whatever analysis you wish and have it actually be relevant without doing your homework first.

      • deV

        As far as Windows, I mentioned before that you did not get the versions right as far as how the code base has evolved from Windows NT 3.5 to Windows 8.

        You also ignored:
        – Server versions such as Windows Server 2003
        – 64-bit products (keep in mind one of Apple’s main features for Tiger or some such was moving to a 64-bit code base)
        – Windows Media Center editions (2002, 2004, and 2005), which also use a code base which was updated beyond XP (some of which came from Server 2003)
        – and all service packs for all of the above versions, as unsexy as they are.

  • Pingback: Sharer » Blog Archive » 就系统周期问题,提问微软Windows()

  • Anonymous

    > Another way to look at this is that for every change in a corporate
    > desktop environment, the average user will change their device
    > experience five times.

    That explains the consumerization of I-T pretty handily.

    • Dave Gude

      This is the part where Horace lost me. Seemingly there is data to support the statement that enterprises generally upgrade every other version. But what data is there to refute the idea that consumers probably do the exact same thing? I am typing this on my PowerBook G4 running 10.5 Leopard.

      Love your blog and podcast, Horace!

      Cheers,

      Dave

      • http://www.asymco.com Horace Dediu

        The point is about “devices” vs. traditional computers. The lifespan of a device is about two years. Mobility is hard wearing on hardware.

  • Guest

    If you did a true comparison, you would list Windows XP, XP SP1, SP2, SP 3 as seperate releases..
    There are more improvements, fixes and enhancements in free service packs than there are in OS X 10.x -> 10.x+1

    • Kizedek

      Well, this is something that is debated isn’t it?

      Horace mentions taking his cue from the designations the companies themselves are using, and one way to do that is to see what they charge for, and what they don’t, regardless of number of new features.

      With MS, Service Packs are free. With Apple, 10.x.x is free update from 10.x

      By all accounts, SP2 made XP just about usable. Furthermore, people still swear by it and brag about DOWNGRADING their new Windows machines to XP SP2, for heaven’s sake!

      On the Apple side, you do have those cautious folks who wait for 10.x.2 before they take the plunge and upgrade from 10.(x-1) to 10.x …but this usually comes out a month or two after 10.x, — not a year or two.

      Spin it however you want, but Apple is on a different model, plan and timeframe than the other OS companies.
      A) The turning radius of MS is far greater.
      B) consumers expect a shorter cycle between mobile OS upgrades
      C) MS is combining desktop and mobile OSs, yet MS must take B into account. With MSs current model and track record, it does not bode well.

      This certainly does look like an interesting piece of analysis, and I think Horace has put his finger on something significant.

      • John H. Frantz

        I seem to remember people in droves who bought Windows Vista machines downgrading them to XP.

        On another note, I would say that Windows 7 is today closer to Mac OSX in development (some would even say better than – depends on the usage) than it was 10 or 20 years ago. That would imply that Windows is developing faster and catching up.

        One of the things that Windows has done well is terminal services that is being able to log into a remote windows machine and either start a new session (windows server) or take over a session e.g. your windows computer at work from home. This is done in a way that requires a layer of software that other OS’s don’t seem to have. VNC solutions are not in the same league.

  • Pingback: WINDOWS 8: MICROSOFT AT A FORK IN THE ROAD » WINDOWS, MICROSOFT, INTEL, MICROSOFTS, METRO, OFFICE » TECH MAGAZINE()

  • systemsplanet

    distance between version numbers is meaningless. a better measure is new Apis
    per month.

    if you did the same analysis on humans
    using age as version, you’d conclude
    all humans are equal because they
    have one new version a year. a better
    measure is what did they accomplish each
    year

    • http://www.asymco.com Horace Dediu

      The assumption (stated) is that point releases are meaningful. From what I know of the software business there are collections of changes to the code which are significant enough that they form a departure from a previous thread of development. Rather than measuring age, this is more about measuring major life events. Like graduating from school, getting a job, getting married, perhaps having children. These milestones are significant pivots in a life and are marked ceremoniously, much like software releases.

  • Anonymous

    The Horacle,

    I suspect I am fairly typical of your readers, in that on a daily basis I use a Windows computer at work and a Mac computer at home. The former is bought by the company, the latter of course by me, and decisions about upgrades, and therefore the relative “manoeuvrability” of each, are likewise either corporate or personal.

    As a user, my experience of the Mac is that with every upgrade, it “just keeps getting better”, while on the Windows side it varies from “more or less painful” to “we need to go back to the old version”. The fact that Mac updates have recently transitioned from “between free and $128″ to “between free and $28″ has of course helped, while the cost of Windows updates, whatever it was, has sometimes stalled my company’s update cycle. The “free’ updates are still painful, in that they take time and may or may not work. Windows updates are still dependent on my company’s IT.

    I believe a similar differential affects the upgrade cycle of mobile devices. Apple has moved control over software updates from the service provider to the user, not just on the computer but also on the phone, while control of both Windows and Android mobile devices remains in the hands of service providers.

    This means that the “manoeuvrability” of each OS is a theoretical advantage or disadvantage, but one that ultimately depends on the service provider. If they decide not to manoeuver, then there is nothing that the device or OS maker can do.

    So there is a clear advantage to Apple’s OS here. Microsoft has chosen to combine the OS of its computer and handheld tablets but not its phones with Windows 8, while Google has an OS that combines its handheld tablets and phones.

    Google has already shown that such a combination does not work harmoniously. We have yet to see whether the combination proposed by Microsoft for Windows 8 will succeed in working with both tablets and computers.

    So, in my view, “Windows’ manoeuverability” is an illusion.

    Both Microsoft and Google have chosen to encumber their mobile devices, both phones and tablets, with an OS that does not work with their intended hardware, both phones and tablets.

    • http://www.asymco.com Horace Dediu

      You are very much on track on the *why* Apple is more maneuverable. Integration is a big part of it. (That and architectural elegance). Integration leads to the right operational structure and also to the right business model which keeps the turning tight.

    • http://www.asymco.com Horace Dediu

      You are very much on track on the *why* Apple is more maneuverable. Integration is a big part of it. (That and architectural elegance). Integration leads to the right operational structure and also to the right business model which keeps the turning tight.

    • http://www.asymco.com Horace Dediu

      You are very much on track on the *why* Apple is more maneuverable. Integration is a big part of it. (That and architectural elegance). Integration leads to the right operational structure and also to the right business model which keeps the turning tight.

      • Dan

        Hmmm… in my magnificent opinion, the reason Apple tends to be more maneuverable is that Macs generally don’t do much of anything. There is no real business model involved since Macs can’t integrate with or connect to anything useful (if you don’t believe me, try connecting one to SQL Server, Oracle or an Exchange server); there really isn’t much of anything in the way of useful software available for it (unless you’re a hippie) so basically, upgrades tend to consist of some little tweaks to the user interface which don’t affect the software you’re running since it’s mostly proprietary anyway.

      • Alan

        @Dan, congratulations on coming out of your 10 year coma.

        Some things have changed while you’ve been away – Exchange support is built into OS X (as well as iOS), Oracle 10g is an OS X app, and several options will connect (though not host) a SQL Server.

        My congratulations to your primary doctors, but the jury is out on whether you need more treatment.

      • Rustyspamola

        Dan your “magnificent opinion” is a fairly lame troll.

      • deV

        There’s plenty of reasons not to like Apple. But none of them are given in your post.

  • Sander van der Wal

    Should be interesting to add Symbian and Maemo/MeeGo. Symbian for instance went from 1 (Epoc ) to 9.1 pretty fast, and then slowed considerably.

  • Max

    The desktop OS is more complex than mobile one. When desktops were simpler (MS-DOS?) they had shorter cycles too.

    • http://www.asymco.com Horace Dediu

      I don’t have a way to test this hypothesis. How do you measure complexity?

      • http://can-turtles-fly.blogspot.com Sivaram Velauthapillai

        Although not perfect, lines of code is a rough proxy for complexity (or at least person-hours usually required to produce software).

      • http://www.asymco.com Horace Dediu

        But the lines of code of iOS are probably in the same order of magnitude of a desktop OS of a few years ago. Certainly the power of the devices is on that level. You can see that here: http://www.asymco.com/2011/03/14/on-feeds-and-speeds/

      • deV

        The latest iOS for iPhone 4 and iPad is roughly 700 MB uncompressed, as installed: http://theiphonewiki.com/wiki/index.php?title=Firmware

        The Windows 7 install DVDs are 3.0-4.7 GB, which is a gross underestimate, because install DVDs are highly compressed.

        Anyway, they are not on the same order of magnitude of complexity. Regardless of these basic figures, iOS only targets a few hardware options, has a single method of acquiring software, supports only one graphics framework, the list goes on…

  • Pingback: OS nimbleness plotted, desktop OSes can’t keep up with mobile | CD DISK()

  • Pingback: The tell-tale signs suggesting a platform’s demise | asymco()

  • Ephllc

    Evolution from Windows 98 to Windows 2000 did not quite happen that way. Windows 98 moved on to become Windows ME and then reached the end of product line basically end of MSDOS based OS. Windows 2000 evolved from NT 3.1 (1993), NT 3.5 (1994), NT 3.51 (1995), NT 4.0 (1996) and Windwos 2000 and XP plus the NT OS based major server editions release of Windows 2003, Windows 2008.

    So the Windows release gap based on Windows numbering is invalid. The Windows releases were based on two different OS product lines. Accurate Windows release gap should have been the release gap between NT 4.0 and Windows 2000 and not bwtween Windows 98 and Windows 2000.

    • deV

      This. To expand on that a bit, here’s the full list of versions of the Windows NT product line and code base that evolved into Windows 7 and 8 (not the completely distinct 95/98/ME line): http://en.wikipedia.org/wiki/Windows_NT#Releases

      Note that there are 8 version numbers in the 5.x series. The whole list for the Windows NT line alone includes 18 distinct products spanning 8 different numbers.

      Version numbers are kind of silly anyway. In the browser market, Google Chrome is up to “major release” 14.0 within 3 years.

  • Tevye

    Users shouldn’t have to care about versions or upgrades…

  • Tevye

    Users shouldn’t have to care about versions or upgrades…

  • Jondmichael

    While the idea is nice, this is a chart of non-equivalent items measured using poorly specified increments.
    Each OS serves a different function: they cannot replace one another. If one is comparing Android to Windows NT then why not throw in console systems as well? Even Apple and Windows have evolved along different lines, with Apple operating in a closed hardware environment and Windows in an open hardware environment.
    Version increments are misleading and non-comparable: a set of comparable changes isn’t identified, and changes are bundled differently by different companies, based on their particular system and needs. Many significant changes common as third-party apps, especially on the more open Windows system that relies on third party providers.

    I like this blog, but this analysis missed too many basic facts. It’s misleading.

    • Anonymous

      > this analysis missed too many basic facts. It’s misleading.

      I respectfully disagree.

      While the underlying operating systems are not equivalent (nobody’s suggesting that Mac OS Classic is a plausible substitute for Android), the engineering and product management processes are. The idea of turning circles, particularly in the OODA/combat sense, is a useful metaphor here in trying to understand why some product pipelines are more readily able to adapt and respond to the world around them. The disparity between release cycle times of the classes of (internally equivalent) desktop systems and (internally equivalent) mobile systems is clear in these diagrams, and is corroborated by community perception.

      > why not throw in console systems as well?

      This would have been an interesting inclusion. The long, slow, coordinated releases typical of that market would have stood out very clearly, I suspect.

      > Apple and Windows have evolved along different lines

      Is this then correlated with Apple’s relatively narrower turning circle? If correlated, might it be partially causal or at least enabling? That strikes me as a potentially interesting avenue of inquiry, highlighted by this analysis.

      > Version increments are misleading and non-comparable

      It’s true that version numbers mean different things to different companies, but Horace has clearly tried to think through carefully and maximise comparability. e.g. using point releases for OSX but major version numbers for Windows.

      > Many significant changes common as third-party apps

      Are any of them so fundamental and so widely deployed as to materially affect the mass market’s use and perception of the operating system? Personally, I can’t imagine that an update to Office or Firefox would cause me to less desire an operating system refresh.

  • Jumpcable

    Apple’s turning radius would look a lot different if they had to worry about enterprise support. Consumer tech is Apple’s focus. While they dabbled in the enterprise with Xserve, it was never meant to be a market where they would make the bulk of their money. So, operating systems aside, you have one company that is working at the speed of the consumer market (Apple 1 – 2 years- just look how fast they are releasing new mobile devices), and you have the other that has had focus on both for a very long time (Windows with their home based OSes and then their enterprise focus with Server, Datacenter, Exchange, SCCM….it goes on and on – a market that usually is on a 4 – 5 year cycle). Hell, I’m not sure if any one has mentioned in the comments that Apple also controls the hardware. They can make the line swing any which way they want – I’m pretty sure if they drew it off a cliff, many would follow without question.