Why MinWin speculation matters
MinWin has re-emerged as a source of debate since Microsoft last week released a few more details regarding Windows 7. Previous discussion of the "MinWin" kernel by engineers related to the Windows 7 team would have led one to believe that MinWin was a fancy new kernel that would be released with Windows 7. Steven Sinofsky, a man well positioned to know what he is talking about, explained that we already have MinWin, as MinWin is little more than a stripped down version of the Vista kernel, one that is shared with Server 2008 and which will be improved in Windows 7 timeframes.
The status of MinWin in Microsoft's desktop and server products, however, is not the key reason people are so interested in the existence of a stripped-down Windows kernel (though it certainly matters, as it might hint at new levels of modularity). Rather, the interest in MinWin rests more on a realization that more and more embedded devices will share our lives in the future, causing desktop and servers to shrink as a percentage of the daily computing pie (though just as a percentage; it's not so much that we will use desktops and servers less, but that we will use a lot more embedded devices).
This means Microsoft's strategy vis a vis the embedded space is rather important. Right now, most of that strategy seems oriented around Windows CE (the base OS for Windows Mobile), an OS that was based on Windows-proper at some point in the past, but is clearly not Windows.
Contrast this with the strategy followed by some of Microsoft competitors. Apple has managed to fit a stripped-down version of their desktop operating system into both the iPhone and the iPod touch. This means that the exact same skills associated with desktop OS development work applies to iPhone and (eventually) iPod work, which makes the APIs unique to Mac development more interesting than their desktop market share alone might merit. The wisdom of a common codebase only becomes more apparent as the iPod "halo" effect reinforces Mac market share.
Linux might not be lighting the world on fire in the desktop space, but it has a strong showing in the server space (though Windows is ahead there), and even stronger presence in the embedded space. Finding Linux in everything from Television Set-Top Boxes to network routers is common, and those embedded chops are being brought to bear on the mobile phone market. Big companies like Google, through its Android initiative, back Linux as an embedded platform, and some analysts think 23% of the mobile phone space could be Linux-powered by 2013. I believe that is very possible, as Linux is a commanding force in embedded development.
Microsoft's biggest competitors, in other words, have had strong success turning a single OS platform into something that spans device categories. Many people wonder when Microsoft plans to do the same, and MinWin seems the best path by which to make that happen.
Would a MinWin-based embedded strategy supplant Windows CE? I don't think so, as Windows CE is suited to resource-constrained environments in ways MinWin might not be...at least in the near term. Code compatibility that stretches from the desktop and server space to embedded consumer devices, however, would be a pretty powerful competitive differentiator...particularly given the market share of desktop and server Windows.