Sunday, October 03, 2004

Microsoft Titanic, Meet Mr Open Source Iceberg

Summary (of linked article)
So it's begun. Microsoft is admitting in its financial filings that Linux and open source are eating into its revenues. Worse for them, they say, is that what they're seeing is the tip of the iceberg.

[my comments]
Microsoft could start from scratch to build a new OS that doesn't have all the problems found in the Windows architecture. Otherwise Linux will just keep getting better and MS will just be slapping band aides on Windows in the meantime.

For a completely new OS they could adopt the mySQL dual licensing model from the very outset and call it "open source" so that they'd be in step with the ideological politics of the times. (mySQL certainly seems to have the smartest business revenue model of any of the open source products.)

A new kernel at the lowest level could be designed with concepts that promote security and rock stable software. The .NET abstraction could be rolled on top of it (like Longhorn is doing). However, it would have no ability to run old Windows software - just new managed code software. (Any manner of support for the old software is nothing but a security and stability quagmire - which is the foundational problem of Windows to begin with and is primarily why Windows is loosing business in the server market place.)

Obviously it would be primarily a 64-bit friendly OS from the outset but could be scaled down to 32-bit for the embedded marketplace. (By the time this OS would surface in the marketplace, 64-bit would be predominant in desktop/laptop machines, in addition to servers. So 32-bit will be cell phones and control processors.)

The enterprise flavor would be the first natively cluster-aware OS so that the very OS layer itself would make simple the activity of writing enterprise applications that can scale and be highly available in a grid/cluster configuration. Hence this new OS, in its enterprise flaover, would incorporate all the features found in something like JBoss: hands-off transparent clustering, transactional distributed caching with replication (JBossCache), transparent ORM persistence (Hibernate), easy declarative distributed transactions, etc. Messaging would be the substrate for all communications. Even RPC, where provided, would be layered on top of messaging. Effectively it would be a fully grid-ready OS in its enterprise incarnation.

The whole Longhorn thing was such a major disappointment for its lack of imagination and vision. It didn't go near far enough in addressing OS fundamentals, but concentrated almost solely on end-user experience stuff. That has been Microsoft's problem all along from the very beginning with Windows. They deal mostly in eye candy while practically ignoring the termite eaten wood underneath.

Linux itself is based on OS ideas that are 30 years or more old. To write modern enterprise software Linux has to have Java slapped on it. The ancient notion of restricted user-mode vs. privileged kernel-mode is not an adequate concept for promoting security and stability. Native instruction code execution is too dangerous to permit in user-written software. A modern OS design should permit only managed code execution of any user-written software - including device drivers (special memory management features could be provided to deal with heap garbage collection vs safe interrupt handling).

Microsoft could out-do Linux in coming up with a modern OS. They have the money in the bank to fund reinventing themselves. The question is do they have the imagination and the spunk? They're known for being fierce competitors, but are they really? Longhorn was not a very "fierce" response at all to their predicament. Microsoft needs to start by firing a lot of VPs and axing a lot of looser projects. They need to trim themselves back down to a lean machine. They need to cease out of control sprawl that is expansion for the sake of expansion. They need to make it known again that they're first and foremost a software development company - and one of the very best at it. They need to make it clear to the marketplace that they're going to deal once and for all with being known for having a shoddy OS that's dangerous to business. Stock holders are going to start getting restless with Microsoft's listlessness if they fail to demonstrate bold action at this crucial juncture of company history.