Micro Kernel Question

Alaric Williams alaric@alaric-williams.com
Sun, 4 Jul 1999 15:08:51 +0100 (BST)


On Sat, 3 Jul 1999, Brian Rice wrote:

> >In my long debate with Fare about this some time ago, I decided that
> >calling a system kernelless was about splitting hairs on definitions -
> >there would always be some fundamental metamodule somewhere, which was
> >basically a kernel unless you went for some predefine strict definition of
> >kernel which depended on the person :-)

> I still disagree.  Compilers for every OS that I know of translate some
> high-level language into running code that invokes kernel services. 

How about Windows systems - the "kernel" there is just a DLL that happens
to be called "kernel" because it contains the basic runtime system.
AFAIK, there's nothing special about the Windows kernel DLL other than the
fact that it gets magically loaded by whatever boot loading system Windows
uses these days rather than being brought in on demand...

> The
> issue is to change this notion into one where the added code (perhaps
> considered aspect code) allows the program to use memory, i/o, and processor
> resources in a co-operatively stable way. 

You're talking about the definition of a kernel as being where the
priveleged code happens, then. That's not the case in Windows as far as I
can tell, since drivers and the like are just DLLs (it's been a while
since I studied Windows programming, so don't quote me on this, but this
is certainly how things were at some point... or just substitute AmigaOS
for Windows in my argument :-). But I wouldn't call Windows kernelless
since there's a bit of it called "kernel" :-)

>     Brian

ABW

---==========================[ http://www.alaric-williams.com/ ]============---

              They flutter behind you, your possible pasts;
          Some bright-eyed and crazy, some frightened and lost

                                         - Pink Floyd, The Final Cut

---=========[ alaric@alaric-williams.com ]==================================---