🚀 go-pugleaf

   Boy, did I open a can of worms.  To summarize, I originally said:

in article <2536@legs.UUCP>, dash@legs.UUCP (Darrell Shively) says:
>
[...]
> 	By the way; since your machine is a 386 based system, do NOT use
> IBM OS/2 1.1 (either EE or SE).   The IBM versions of OS/2 1.1 for the AT 
> does not support the 386.  After all, IBM does not make a 386 equipped AT
> machine.  Starting with version 1.2 however, both IBM and Microsoft has 
> been/will be using the enhanced kernel that supports all 80x86 CPU's.
[...]

   To which Tim replied:

> 	This sounds like baloney to me. I have been running IBM 1.1 SE
> 	on a Compaq 386s for quite a while now with _NO_ problems. We
> 	have OS/2 from a couple of different vendors and use it on
> 	various 386 machines (both straight 386 and 386sx) and have
> 	had _NO_ problems, especially no 'Brand X won't run on a 386'
> 	problems. In fact, the only problems we have seen at all are
> 	mouse drivers incompatibilities.
> 
> Tim Henrion
> tim@banyan.com --or-- ...!bu.edu!banyan!tim

   And Kaleb:

> I'll second that, I run IBM 1.1 SE on a generic Taiwan 386SX; I use a Logitech
> HiRez bus mouse, using the Microsoft Bus mouse xxx/199 selection in the setup.
> 
> kaleb@mars.jpl.nasa.gov            Jet Propeller Labs
> Kaleb Keithley

   But Paul  was somewhat supportive, in that:

> No, you've just been lucky! Older 386 compatibles (like mine) may not
> report the machine type correctly. I had to hack the install script on IBM's
> 1.1 install disk to work with my machine.
> 
> 					Paul Verket
>					verket@venice.sedd.trw.com

   However, Tim and Kaleb are quite correct, and I am guilty of gross over-
simplification.  While what I said was not really *inaccurate*, is was
definitely *incomplete*.  Here are the details of why running IBM OS/2 1.1
SE and EE can be (but are often not) problematic on a 386 machine.

   1) IBM OS/2 1.1 for the AT is designed for a '286.  The PS/2 version
      (3.5" disks) utilizes the 386 architecture.  But on an AT clone,
      this means:

      A) It will use the "shift into reverse while on the freeway" technique
	 for mode switching (e.g. triple fault).

      B) The IBM 1.1 kernel uses the undocumented 286 LOADALL instruction,
	 which is not present in the 386.

   2) As Paul points out, there can be a problem with machine type identi-
      fication, which can cause a problem during installation.  But this
      should not continue to be a problem once one gets OS/2 on their
      machine.

   Problem 1A isn't a killer.  It will only cause a performance degradation
while running in the DOS compatibility box.  I'm not sure how great a deg-
radation, but it is probably significant (~20% ?) since OS/2 is constantly
switching in and out of real mode while in the 3x box in order to do back-
ground processing of protected mode tasks.

   Problem 1B sounds like a killer, and is on some older machine with older
BIOS's.  But the solution is to use a good BIOS that implements software
emulation of the LOADALL instruction for the 386.  Nowadays, I should
imagine that just about any 386 one can buy (ranging from Tim's Compaqs to
Kaleb's Taiwan clone) should be just fine.  So even with an older machine,
upgrading the BIOS will probably make it palatable to IBM OS/2 1.1.

   Problem 2 can also be solved with a BIOS upgrade.  Or you can modify
the install script as Paul did.


   CONCLUSION:

   So really, IBM OS/2 1.1 will PROBABLY run on any reasonable 386 AT
clone.	But this is not guaranteed.  And if the clone's manufacturer
has gone to the trouble of porting OS/2 to their machine, it will
without a doubt run faster in the compatibility box.  They will also
support it on your hardware, which I doubt IBM will do.

   As an aside, to reiterate a previous point, IBM OS/2 1.2 does not have
any of these idiosyncrasies.  The IBM and Microsoft OS/2 1.2 kernels are
functionally identical (at least according to Microsoft) and both support
the 386 enhancements.

   Sorry for the misunderstanding; I should have thought about what I was
saying a bit more.  As a friend of mine says, "I was running off at the
keyboard".


  "Any opinions expressed are my own, and in no way represent my employer."
-- 
Darrell M. Shively II (dash)        uucp:      ...!uunet!ccicpg!legs!dash
AST Research Inc. Dept 660          MS OnLine: MS05455
P.O. Box 19658                      Voice:     (714) 727-8637
Irvine, CA  92713-9658              Fax:       (714) 727-9358