OSDev.org

No? What about an OS that uses a micro-kernel design for 64-bit drivers and a monolitihic design for 32-bit drivers. Is it monolithic or a micro-kernel? As an example, in a mainly 32-bit configuration it could run all drivers in kernel, while for a 64-bit configuration it could run all drivers in user-space. Would you assign different labels to it based on configuration?

a monolithic kernel remains monolithic even if you move drivers out of it and into userspace. the reason is that in microkernels there is an elaborate IPC between the different "user space" services whereas in a monolithic kernel there are just user space programs.

the windowing system is almost irrelevant. the compositor does the real work and the rendering API like opengl and directx renders actually renders the images to be composited.
in windows you can write your drivers to be in userspace. I don't know if there are restriction as to which ones, maybe someone can tell me.
i think in linux the printer drivers are in userspace too (like in windows)

Bogus argument. Linux has a much broader IPC collection than most microkernels, including a read() and write() pair, send() and recv(), shared memory and named pipes, therefore it is a microkernel?

The micro vs monolithic kernels is quite uninteresting because there are so many shades of gray between the two. However, the trend is obvious that both Linux and Windows are migrating more and more drivers to user space because it provides a number of benefits while the performance is still acceptable.

Still I'm intrigued by QNX why they decided to put their network drivers in the kernel. I'm a bit confused by this, can someone who knows more about QNX give more information about this and what part of the network driver is in the kernel.

The first bit to me just looks like a chance to bash Windows XP, id like to see how it pairs to 7 or 8 now.





Nobody cares if you care. You either do or don't, otherwise don't post here at all.

There is also the exokernel design, which provides memory protection and a hardware multiplexer and not much more.

I found one more great article about Windows Nt Vs Kernels. Have a look
http://technozed.com/windows-kernel-vs-linux-kernel/

Hi,


I have read the article and I decided to make numerous remarks about it.

"Operated"? Also every piece of software is "maintained" by developers, so this statement is redundant.

The Unix kernel didn't include drivers for all of the modern hardware, as they obviously didn't exist back then. Linux instead includes drivers for this hardware. So how can they be similar?

The user interface (I presume this is meant, since the article is most likely written by a noob) is not a part of the kernel.

Again, confusing Unix and Linux.

Facepalm. The NT kernel architecture was present a lot earlier.

This is utterly wrong. FreeBSD can run in many different platforms, and there is a port of Windows 10 to (at least) Raspberry PI.

SYSENTER and SYSEXIT are processor instructions, and not system calls.

Not sure what does the author mean here, but I presume he either means they switch to ring 0 in the same way (which is not true), or he means they have the same system calls (which is not true either).

"Page tablets" - I don't have to say more.

This is operating system development theory, not applicable to Windows kernel against Linux kernel.

Wrong, Linux swaps too. And Windows has access to the memory too.

I didn't know you could open the kernel executable in a window on the desktop.

I didn't know there are operating systems without a kernel.

Nope. Linux tolerates insufficient memory in a better way. It even lets the user to set the "swappiness" parameter to fine-tune it according to usage cases.

I short, I might have missed something, but the wrongness of that article is much above tolerability level. Most likely it is terminology issues here.


Regards,
glauxosdever

An appalling article.