Ah, once again Denys issued some pretty wise thoughts. I'd like to add my 2
cents (the Euro is divided in cents, too) here.

> When people mention the reliability of the HP 3000,
> some confusion
> may ensue.  Does the person mean the reliability of the
> hardware or the
> reliability of the software?  Actually it is both.  If you
> had an excellent
> OS on substandard hardware, it would not be a reliable
> system.  Conversely,
> if you had magnificent hardware running a so-so OS, you would
> also not have
> a reliable system.

Yep. Couldn't agree more. When customers talk about their computers, it's
just that - computers. They don't separate hardware and software. Why should
they ? One cannot exist without the other. But these days HPe3000 and
HP9000s are the same hardware plaftorm. MPE is reliable, but so is HP-UX.
OK, HP-UX does not have an XM-like feature, but who (expect our community of
technicians) cares ?

> Just try to use USB on Windows
> 95 or Windows
> NT 4.0.

My colleague tried exactly that : connect a USB peripheral to his W95
laptop. Took him about 3 days, the laptop almost flew thru the windows (no
pun intended) countless times, foul language was used, etc. After 3 days, he
gave up, reinstalled the PC with W98, lost countless megabytes of data
(including lots of very valuable stuff), and got it to work.

Now we all have W2K laptops, and I have to say the OS is light-years ahead
of W95 and even W98 in terms of stability.

> It is only since MPE gained access to the PCI architecture,
> itself over 8
> years old, that MPE gained access to more current mainstream
> devices running
> on LVD SCSI, way late. It is a shame PCI took so long to
> appear on the 3000.
> I understand it was quite an effort and it was very well
> executed, but it
> was so very late.

Yes, but what would we have done with PCI backplanes based on, say, the
K-Class ? PCI is able to run at 133 MB/s, but the K-Class, AFAIK, only had
NIO 32 MB/s backplanes. Not sure about GSC, and pretty sure the HP3000
K-Class never had HSC. So PCI could probably have worked, but would have
totally hogged the backplanes, anoud the net result would have been an
actual performance DECREASE. This is my best guess for the reason why it was
never implemented.

Now come the A- and N-Class boxes. These babies have aggregate bandwidths
running in the gigahertz ranges, with 440 MB/s-capable backplanes. With its
much higher throughput, PCI is perfectly suited to the task of providing
better performance. It now makes sense to have it.

Besides, Denys perfectly underscores that implementing PCI in MPE was no
small feat. From my past travels within source, I can say that it must have
been a big tast to do that.

Christian Lheureux

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *