Steve Dirickson b894 WestWin <[log in to unmask]> wrote:
 
One last time -- keeping this thread alive one more time.
 
>I agree that the effort would likely be significant--*if* there were any
>meaningful new items involved.
 
Well, I disagree.  Our experience has been that when you mix old
peripherals with new processors, you uncover timing issues that you
didn't know existed.  A good portion of our processor development time
is spent in certifying existing peripherals.  In this case, there is a
meaningful new item involved -- a new processor with different
characteristics, especially I/O speed.
>
 
<snip>
 
>If some future generation of boxes simply doesn't accept the current
>HPIB adapter, and provides no replacement, then it becomes
>self-disqualifying, so there is again no need for a time limit.
 
Somehow, I kinda think we will still be having a similar conversation.
As in, "Why doesn't this new system support HPIB?"  To tell the truth,
I really don't see much distinction here -- HP does not support HPIB
on our latest box.
 
>
><<I am not sure that this is "purely a marketing issue".  I believe that
>this decision was more driven from a technical standpoint.>>
>
>Except that the hardware and software both appear to work fine; where's
>the technical limitation? That keeps coming up as the key issue. There
>may very well be some kind of show-stopper somewhere; I just haven't
>found it (and neither have my HPIB printer, HPIB 7980A, HPIB C1511A,
>HPIBDIAG, REELDIAG, HPIBDDS, or CIPERLPD).
 
There are two technical issues:
 
   1)  HPIB was not tested on these systems.  Our testing includes
       verification of not only "normal" use but also "stressed"
       usage.  Do these HPIB peripherals recover from powerfail?
       Have we tested all to corner cases?  etc.  As I've said,
       new processors sometimes open up timing windows that can
       mask serious (e.g. undetected data corruption) problems.
       We cannot guarantee your data integrity if you use untested
       peripherals.
 
   2)  Performance:  HPIB was developed in the 1960s and your 9x9KS
       processor was developed in the 1990s.  They are a poor match.
 
Remember, your peripherals are the lifeblood of your system.  The bottom
line is the HPIB peripherals are untested and unsupported.  I'm glad you
are happy with your HPIB peripherals -- they were built to last, but you
are taking a risk in using them.
 
 
>But we're talking about existing, supported HPIB devices connected to an
>existing, supported HPIB controller plugged in to an existing, supported
>CIO adapter. What requires certification?
 
   Typically, the follow issues are considered when certifying
   peripherals on new platforms:  "normal" operation, "stressed"
   operation, powerfail recovery (both system and peripheral),
   max configuration, and performance.  Certification is a
   non-trivial exercise.
 
>
>You can-as long as you aren't trying to solve a network problem.
 
I replied to the "recommended printers" question before reading this
paragraph.  As I pointed out, there are a variety of solutions.  For
example, the HP5000s are available in Network, Serial, Parallel, and
SCSI flavors.
>
>Thanks for the continued dialog.
>
>Steve Dirickson         WestWin Consulting
>(360) 598-6111  [log in to unmask]
 
Thank you, too.
 
Jon Cohen
HP Commercial Systems Division