On Tue, 10 Dec 96 00:40:37 -0700 
 Christian Kuhtz <kuhtz@ix.netcom.com> wrote:

 > *grrr* ;-) A better approach is to port NetBSD to be hosted on a Mach mk and  
 > port the mk to a new hardware.  Otherwise we will most likely never see  
 > efficient multiprocessing support.

I don't agree :-)

Don't get me wrong... Mach is fine and dandy... But there are performance
considerations to think about in u-kernel environments; all that message
passing carries a price in overhead.

NetBSD _will_ have efficient multiprocessing support; a group of folks are
actively working on the design now... eventually, NetBSD will have a
synchronous multithreaded kernel.

I don't see how much a NetBSD server could benefit from Mach's MP
capabilities without adding the locking primitives to the NetBSD
code anyhow.  Since you'd have to do this anyway, why not do it to
the NetBSD kernel itself? :-)

 > Just because NetBSD is a standalone kernel, doesn't mean you can't rip its  
 > guts out and mutilate it to be an OS server on top of a Mach mk.

...and, likewise... just because we _can_ make a NetBSD server for Mach
doesn't mean we _should_ :-)  (I'd certainly never build a mission-critical
server out of such a system...)

Jason R. Thorpe                                       thorpej@nas.nasa.gov
NASA Ames Research Center                               Home: 408.866.1912
NAS: M/S 258-6                                          Work: 415.604.0935
Moffett Field, CA 94035                                Pager: 415.428.6939