[rescue] Mac Appliance

[Dave McGuire]

On August 3, Mike Dombrowski wrote:
> >designing chipsets that drive 2 CPUs well; and MSFT is slowly figuring 
> out
> >how to write SMP code...
> 
> Home PC's don't need SMP.

  Wow, thanks for deciding that for the rest of the world.

  *I* will decide what my computing needs are, thank you.

> NT has been doing SMP for years. For that 

  ...and NT has been too unstable to use for anything important for
years, too.  Go figure.

> matter let's beat down NetBSD and OpenBSD because AFAIK, they don't 
> have SMP support yet.

  NetBSD has SMP on Alpha, VAX, and Intel now.  SPARC is just around
the corner.

  I feel the need to explain something about NetBSD with regard to
things like SMP.  Linux and FreeBSD, despite < 1% arguments to the
contrary, really usefully only run on one platform.  It's relatively
easy to do something like that on one platform.  NetBSD runs on FORTY
different platforms, supporting like A DOZEN different [incompatible]
processor architectures.  Now that 90% of the SMP stuff is done in an
architecture-independent way, the other 10% need only be done for the
other processors.  As a case in point, NetBSD/vax went SMP a matter of
weeks after NetBSD/i386 did.

#define RANT
#ifdef RANT
  And don't even get me started about OpenBSD.  It exists because a guy
with an attitude problem got kicked out of the NetBSD core group
because he couldn't get along with anyone.  Now granted they've done
some great work, but still...OpenBSD = NetBSD + Theo's Patches.

  In a world where "turn off everything in the default
/etc/inetd.conf" can be advertised as "proactively secure", them's the
facts, boys.
#endif

> Hrm, so graphics accelerators like Geforce don't count? Nor sound 
> accelerators like SB Live? Intel CPUs are so stinking powerful that 
> they can simply brute force everything. God, that NeXT, they would 

  "Fast-clocked" != "powerful".  Clock speed isn't a measure of
computing performance.

  I have a nine-year-old machine in my computer room whose primary clock
is 33MHz per processor.  Yes, thirty-three.  For floating-point
applications, you have to clock an Intel processor at like 500MHz to
get similar performance.

> never put no hardware in a device and make the CPU do all the work 
> would they? Never, they'd never do anything like that. Wait, they did, 
> NeXT Laser Printer.

  Though I don't agree with the idea, we must understand why they did
it.  NeXTs have a postscript display system...Why interpret the
postscript twice?  Not trying to be agumentative, but I it's an
important point.


     -Dave

-- 
Dave McGuire
Laurel, MD