From: Jon M. Taylor <taylorj@ecs.csus.edu>
  To  : ggi-develop@eskimo.com
  Date: Thu, 20 Aug 1998 20:40:16 -0700 (PDT)

Re: Ring 0/2 KGIcon?

On Thu, 20 Aug 1998, Sengan Baring-Gould wrote:

> http://www.lwn.net/980820/a/ac-x.html
> 
> What's the point of splitting into 2 rings? Security? 

	Among other things.  The idea is that you can do some
microkernel-type stuff with privileges, without having to actually write a
true microkernel, if you run drivers in a privilege level between the
kernel and userspace.  If the drivers oops, their ability to affect that
rest of the kernel is minimized by the ring-based privilege locks.

	Alan is analogizing the KGI/LibGGI driver library split to the
OS/2 ring 0/ring 2 split.  Traditional Unix kernels have drivers which
serve two functions: to multiplex the hardware and to link specific
hardware functionality to a standardized API.  KGI multiplexes the
hardware, LibGGI implements the API(s) on top of that hardware. 

> How would this map
> into non-x86 architectures?

	Not well, IIRC.  Most non-x86 processors only have two priviledge
levels.  That's really all you need - IMHO the KGI/LibGGI split is more
sensible than what OS/2 does.

Jon

---
'Cloning and the reprogramming of DNA is the first serious step in 
becoming one with God.'
	- Scientist G. Richard Seed