Is Windows "Midori" doomed to failure?

Microsoft's cloud computing strategy will fail if it seeks to discard the NT code base prematurely

Last week, I took Microsoft to task for floating yet another batch of unrealistic tidbits about Windows "Midori," the rumored future version that will encompass a complete rewrite of the core OS plus a new, distributed application development model for linking together disparate devices of varying form factors (a.k.a Windows "Mouthful"). I objected to this vision for a couple of reasons:

  1. It involves a clean break from the current Windows architecture. As fellow newshounds will attest, the calls for Microsoft to perform just such a move -- to abandon Windows as we know it and start from scratch -- have been increasing of late. It seems everyone who's anyone is opining that Windows needs a fresh start – everyone, that is, but the poor souls who have to support Windows in the real world.

    To them, the idea of junking a proven architecture for something that's based on a kind of pie-in-the-sky thinking normally reserved for academic PowerPoint decks is terrifying. Enterprise IT just got through the whole Longhorn Reset mess (and the resulting hairball of false promises and subtle incompatibilities called Vista). The thought that Microsoft might be back at the OS crack pipe, ready to dream up new ways to make our lives miserable, is simply too much to bear.

    IT wants evolution, not revolution. Show us a road map that preserves our current investments while allowing us to embrace new computing paradigms in a practical, nondisruptive fashion, and we'll sign on in droves. But rip & replace? With something only a CompSci professor could love? That's so '80s!

  2. It's unnecessary. When people call for a clean break, they usually cite Windows' perceived lack of security or DLL Hell or some similar straw man. What these arguments have in common is a focus on the User Mode of Windows, the place where applications run and (sometimes) conflict with each other. The other half of Windows –- the Kernel Mode components –- rarely get mentioned, mostly because those calling for a fresh start (typically the Mac-using media blowhard types) don't have a clue what goes on underneath the GUI.

    The truth is that, from an architectural standpoint, there's nothing wrong with the New Technology (NT) kernel that has been powering all non-DOS-based versions of Windows for more than 15 years. When it was originally conceived back in the early '90s, the NT kernel's developers -– led by the legendary Dave Cutler –- designed enough modularity and flexibility to allow for significant evolution without requiring a total rewrite. As a result, Microsoft has been able to expand its NT-based OS table to encompass everything from tiny embedded systems to massive server farms, all using a single kernel implementation that –- with the exception of Vista –- has provided seamless backward compatibility since before many of its present-day administrators graduated from kindergarten.

    My problem with "Midori" is that it advocates dropping the NT kernel in favor of something newer and lighter, perhaps based on the much-ballyhooed Singularity OS project being conducted by Microsoft Research. But again, the arguments in favor of Singularity are focused in the wrong place. As the infamous MinWin demo showed, when stripped to its core, NT is remarkably lithe. Yet peeling the layers away from Windows to make it more agile is a much less drastic proposition than starting with a completely new kernel, which would likely also entail rewriting the device driver model that has become a de facto standard throughout the hardware and software industries. In other words, if it ain't broke, don't fix it.

Prediction: Windows "Midori," with its fresh-start architecture and pipe-dream compute model, will never see the light of day. What will happen is that the myriad concepts associated with "Midori" will trickle down into staid old Windows and be implemented as User Mode extensions that further integrate the OS into the cloud (you can see some of this today with Live Mesh). In the meantime, the NT kernel will trudge along, perhaps getting a few tweaks here and there, but otherwise preserving the DNA of one of Mr. Cutler's most enduring professional achievements.

Copyright © 2008 IDG Communications, Inc.

How to choose a low-code development platform