Stable Windows

Reading about Unix all nught, and then reading UnixFunctionalityVsWindowsFunctionalityDiscussion, something clicked, and a question arose: Could Windows be as stable as Unix if it was open source? Or would it just be hacked into Unix? -- AluoNowu

RedHerring. NT-based versions of Windows are as stable as open-source Unix, give or take a subjective argument. Also, non-open source versions of Unix are also known for their stability. The issue isn't development method, it's kernel architecture, and that's solved in the NT versions of Windows.

Not entirely true. The graphics device context is still managed at the kernel level, which introduces stability issues and affects performance. Were the source available for perusal, a few people who really get the knickers in a bind about that might come up with a a different implementation, much like the various VM/filesystem/device implementations that make Linux kernel development interesting. -- StevenNewton

Early versions of NT had graphics handled outside the kernel. The result was slow, unresponsive GUI applications (much like Unix), so Microsoft moved it into the kernel. I agree that the original statement is a red herring: this has nothing to do with open-sourceness. -- KrisJohnson

I know people hate it when such a good discussion gets spoiled by the facts, but here it is: Linux now has "graphics" support in the kernel (the framebuffer device). The reason why the framebuffer stuff in Linux was moved (on x86) from userland to kernelland had nothing to do with speed (after all, your CPU doesn't magically become faster when running in priviliged mode) but because of stability and security reasons. By the way, with this move, x86 Linux now has the same model as commercial Unices and non-x86 Linux. -- StephanHouben

I hate to spoil your assertions with facts, but in NT the interfaces between different parts of the kernel are more efficient that the kernel-userland interfaces, so moving the GUI stuff into the NT kernel did improve performance. -- KrisJohnson

True for the horribly inefficient kernel-userland interfaces in NT, but not true in general for moving code into the kernel in other operating systems.

Kernel mode *is* faster than user mode although and it's nothing to do with processor speed.

Windows 2000 and XP are very, very stable given reasonable hardware and correctly written device drivers. The argument that Windows 98 is unstable, therefore Windows 2000 is unstable is rubbish. The argument that a particular install of Windows XYZ is unstable, therefore all installs of Windows XYZ are unstable is rubbish too.

It wouldn't be challenging to set up Solaris, BSD, Linux etc and make them fall over all the time. That wouldn't mean that they are unstable.

The whole stability issue can only be measured with science, not your subjective views.

I will say that I have found Windows 2000, when properly configured, to have uptimes measured in terms of months. The major way to cause problems under Win2k is to have crappy drivers. And many of the consumer hardware companies who are building drivers for Win2k have some issues. Like not working when you have two processors, for example.

I will also say that the biggest benefit to an open source Windows OS is that various finicky internal behaviour could be properly explained. According to some people I know, there are some areas where GDI doesn't work the way MS says it should. Emulation could be done with much greater ease.

View edit of July 29, 2003 or FindPage with title or text search