-----Original Message-----
From: IBM Mainframe Discussion List [mailto:[EMAIL PROTECTED] On
Behalf Of Nigel Hadfield
Sent: Saturday, November 04, 2006 11:47 AM
To: [email protected]
Subject: Re: The PSI Letter V4
<snip>
ISTR there was a perception at that time that IBM could not legally
refuse
to license its software on whatever machine the customer wanted to use.
("You could license it on a washing machine if you wanted to."). I guess
this was related to the anti-trust nonsense. I don't suppose IBM always
liked to admit quite how powerful Amdahl and Itel/NAS machines were, so
PCM
customers probably got a fair deal. But if that was the case, when and
why
did it change?
<snip>
Forgive me for rambling a bit (its Monday, and my interruptions are
being interrupted), but some things I will mention might trigger some
thoughts for/from some others specific to this subject.
As I recall, AMDAHL had something called TIDA (Tech Info Disclosure
Agreement) with or from IBM. And there was an "EU" issue where IBM had
to disclose the interfaces to its machines at some office in Europe
somewhere (I used to work for Amdahl, and I honestly don't know the
particulars of this).
In 1997 when IBM disclosed to certain parties that what we now know as
z/ARCHITECTURE was coming, apparently AMDAHL and HDS realized that they
were going to have a problem because the period of time that IBM had to
do the disclosures in Europe was coming to an end.
Well, Amdahl imploded. I don't know what happened to HDS or any of the
other PCMs. But Amdahl had known, because of patent filings, the
probable direction IBM was going to take in many cases (at least up
through 1993). One example was ESA (which we referred to as "Eat S**T
Amdahl"), where Amdahl guessed fairly correctly how IBM would implement
it. But because Amdahl's management lost the vision of what Amdahl was
and did...
So, here you are (H/W MFG co), knowing IEF (SIE) and RMF plus a few
other things that are not in the PoO or the related manuals (which names
I've forgotten) that were issued "publicly" to disclose how this assist
or that worked. And you see the disclosure windows closing. What do you
do? As I said, Amdahl imploded (between 1989 and 1997 with all the
layoffs, I don't know how many of the Macrocode developers were
left...).
As for the new machines (emulation based) that are out, I would assume
that if they had some way of setting the CPU ID on the Motherboard, or
via some special CE/FE only utility, then IBM might be interesting in
talking.
One of the other things being discussed has to do with who you gonna
call when you have a problem? Well, IBM knew that the Amdahl systems
were quite correct in the area of architecture (at least when I was
there). If we found some inconsistency, we had to recreate it on an IBM
box. If we could, and it seemed to not match the PoO (or equivalent
manual), we got to write a letter, detailing a test (without disclosing
what we were doing -- I really enjoyed writing "hand loops" that had to
be entered through the console) and then ask what result we should have
expected if what we got was incorrect. I can't tell you how many ECs
(micro-code/firmware updates) from IBM resulted from Amdahl testing. As
a result we got the idea that we were IBM's West Coast Beta Test Center.
So, should any of these new companies doing emulation have access to the
old Amdahl Architectural Testing Programs (e.g., DIRT, 8E7, Alpha), they
would be able to prove rather quickly that they were compliant with
S/390 Architecture. Now as for z/ARCHITECTURE...
This would mean that the finger-pointing would get rather limited. If
you can define the problem and IBM can recreate it on their platforms,
then the problem is not "your hardware is not compliant."
[So what problems we did have had to do with proving that an interrupt
(of a particular type) had been properly presented, or an I/O request
had been properly handled, or that the service processor had provided
the correct response to your "B2xx" instruction (or similar).]
If the emulation machines get to this level then I think things would be
at the level they were when there were several PCMs making "IBM" type
mainframes.
Later,
Steve Thompson
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
