Talk:Operating system/Kernel models

I think there is more kernel models than just microkernel vs monolitic kernels. There are others options like exokernels. But unfortunately I don't know enough of it to complete this page. --Mildred 22:40, 13 March 2008 (UTC)

On &micro;kernels
L4 has proved that the speed problems of a &micro;kernel is not inherited to the &micro;kernel concept, but due to problems with specific implementations of the concept -- I think this should be noted.

On Hybrid kernels
The image displaying how a hybrid kernel works looks to me like a &micro;kernel with a single-server implementation which has been pulled back into kernelspace -- I suggest that this is either clarified or that hybrid kernel is purged from the article.

Also the introduction to hybrid kernels make no sense, since the design of a &micro;kernel is mutually excluse to how a monolithic kernel is designed &mdash; having IPC into and out of a kernel, does not constitute &micro;kernel design. &micro;kernels are (most often) defined as having a minimality of features not in having certain features. This in itself makes a hybrid kernel (as defined as a combination of &micro; and monolithic kernel design) bogus. See for a definition of &micro;kernels.

On monolithic kernels
I have never seen anyone claim that monolithic kernels are simpler in design than &micro;kernels (there is no arguement against so-called hybrid kernels, which must be at least as complex as monolithic kernels and &micro;kernels according to the definition in this article). The operating system at large may be simpler in design, but that is not what the article claims &mdash; I seriousely doubt that monolithic kernels can be simpler than &micro;kernels in design. Of course, this depends on what you consider part of the kernel and what not (for example, are servers for a &micro;kernel part of the kernel or external to the kernel?)