Martin Franco wrote:
On Sat, May 24, 2008 at 12:02:24AM -0700, [EMAIL PROTECTED] wrote:
I noticed you never mentioned the core reason that makes Hurd different.
You do know about microkernels? I guess the answer to what you think is better
depends on your take of the micro vs. mono - kernel debate.
I have been reading _Design and Implementation of 4.4BSD_ and _Operating
System Concepts_, but so far I don't remember even reading mention of
microkernels. The thing I find most striking from reading material
posted on hurd's website is that most of what has been traditionally
handled in the kernel can be moved into userspace, even made
unprivileged. Each user can individually replace drivers, etc with
thier own versions, without affecting other users, and the worst result
of letting users into what has been kernel space isn't crashing the
system, but a hung process.
Yes and no. Microkernels have significant advantages for certain
things. One of the *big* things they have advantages in is security
since everything is very compartmentalized.
While that means things tend to operate in user space and can do less
damage, it also makes them tougher to write. Writing distributed
systems is always tougher, no argument. However, in this case, it
actually makes security *possible*.
If you want more iformation, the wikipedia pages are good starting
points for hunting down more information:
http://en.wikipedia.org/wiki/L4_microkernel_family
http://en.wikipedia.org/wiki/Coyotos
The other interesting thing is that the L4 folks keep up-to-date with
the latest Linux releases and create L4Linux:
http://os.inf.tu-dresden.de/L4/LinuxOnL4/
Given that they seem to keep up pretty well anymore, they seem to have
it under control. In addition, you can actually see what the
microkernel performance penalty is.
-a
--
KPLUG-LPSG@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-lpsg
