2009/04/27

Mircosoft can't write O/S code

This cartoon lampoons Windows 7.0 Beta. Eerie.

It underlines for me that Microsoft is crap at writing Operating Systems code.
O/S's need to be correct, secure, robust (resilient to errors internal & hardware) first and foremost. Only after that look to features and 'performance'.


How come the Great Unwashed, the home & business users and the media seem blinded in the same way the fanboi's are?

Wikipedia's "Time Line of Operating Systems" underscores this point - compare Microsoft's offerings since 1999: Win-2000, Win-XP, Longhorn: abandoned, Vista and 'soon', Win-7.

Compare to Annual or Biennial shipping of new releases OS/X, AIX, Solaris, HP/UX, Ubuntu, Red-Hat & Fedora, SuSE, ...

Microsoft is the exception, not the rule. Everyone else in their class does better.

Apologists for MSFT may add Win-98 SE, Win-ME, Win-server 2003/8, Media Centre, Home Server, Small Business Server and even Win-CE - plus the many 'Service Packs'.
This is all confabulation. The code differs mainly in licensing restrictions and add-ons.

The 10,000 strong Microsoft 'team' put around 25,000 man-years into Longhorn before Jim Allchin decided to call it quits and start again with what became Vista...
Jim Allchin, who had overall responsibility for the development and delivery of Windows, explained how development of Longhorn was "crashing into the ground" due in large part to the haphazard methods by which features were introduced and integrated into the core of the operating system, without a clear focus on an end-product.
Producing commercial grade O/S's require discipline, design and defined processes - Software Engineering 101 - to pump out regular updates & upgrades. The professional challenge isn't in producing a good piece of code just once, but like the Space Shuttle - doing it again and again to Schedule, Cost, and near flawlessly.

That's also basic Project Management: Plan, Schedule, Control.
Decide what you're going to do, plan how & when you're going to execute the project, manage deviations/problems/surprises during execution.

Operating Systems are vitally important - they sit under every other software layer.
They set the upper-bound for Security, Reliability, Usability and Performance of all applications on the platform. Nothing can be more robust than the layers it sits atop.

Hardware Engineers might take issue. It's long been understood how to fabricate ultra-reliable hardware systems from redundant components. Nobody has paralleled that work in the O/S and Applications arena, so my assertions stand.

Don't get my criticism wrong. Microsoft and Windows have been important in creating the world of commodity computing as we know it. I'd even rate them good at producing some Applications, such as Office. (But the software does suffer from 'featurism' , another challenge.)

The first Microsoft O/S's were adequate and, supported with unrivalled marketing, brought in the age of ubiquitous, mass market computing.
Microsoft have leveraged their natural monopoly (read 'proprietory lock-in') better than any company I know of, in any field.

Their Marketing is an example to be emulated, unfortunately their production of O/S software isn't.

No comments: