Windows Monoculture Myopia Revisited

round stic writes "eWeek magazine has an interesting look at the effects of the Windows monoculture on IT budgets, even as everyone agrees on the severity of the inherent security risks. The article contains interviews with Dan Geer and others who warned about the risks of the Windows monopoly three years ago. The article coincides with a piece in the Observer that suggests Vista is the end of the Microsoft monolith because of how complex the operating system has become."

Comments

Re:End of the monopoly...

[WhiteWolf666]

It would obviously be preferable to have a well-written universal OS, but that brings us around to the old saying: The best kind of government would be a benevolent dictator, but how many dictators stay benevolent?

It would be vastly better if we have well-written universal API layers. Like Java, C#/.NET/Mono, Qt, GTK, and other beautiful cross-platform toolkits.

Unfortunately, except for Java and C#, we don't have any toolkits that go "all the way" in being cross platform, with the possible exception of Win32 (WINE), but Wine is reverse engineered, not bottom-up designed, so there are limitations.

There's no reason for application interfaces to be deeply tied into the OS. Properly engineered, a user-space environment on Linux should be able to run Windows or OS X or whatever applications, and vice versa. The reason we do not have this is not because of engineering limitations, but because of vertical vendor lock in. Lately, this seems to be easing slightly.

I envision a future where applications come with API requirements, not OS requirements. "Requires GTK 2.42, OpenGL 3.0, and SDL. OpenAL 5 required for 3D audio." Software manufacturers would probably support particular "distributions" on the box ("Runs on OS 12.5, Mandriva 2012, and Windows Super-Next-Hubble-Viewpoint"), but like *current* binary software for Linux you shouldn't have many problems installing on the "wrong" distribution; with minor API-requirement caveats.

Think Python applications (these are often cross-platform). Think Java. Think C#. As CPUs get faster, we can put up with some of this overhead; and indeed, in some cases there is very little overhead (WINE does Win32 in userspace on Linux really quickly. Imagine if Microsoft gave up the OS business, but just started selling something like Wine. The "Windows" application layer for Linux, OS X, Unix, Solaris, whatever.

If you want an example of this environment, look at Linux, Solaris' Linux Application Environment, FreeBSD's Linux Application layer, and lxrun, the Linux application layer for (ick) SCO Unix. IIRC, AIX is also Linux compatible.

I think it can work; and giant commercial developers have no problem operating in this multisegmented space. Sure, there are a few more compatibilty bugs than in the Windows monoculture, but there's a greater diversity of applications and environments (from very small systems to giagantic systems), and if the commercial OS space was more competitive in the Desktop world (multiple vendors of multiple pedigree OSs) we would see these compatibility issues worked out quickly.