Hi Dean!

[..]
 DR> I agree with Mark. Significant compromises were made to support
 DR> Dos/Windoze. That support might be justifiable for a while, but
 DR> soon OS/2 will have to break with the past and move forward. For
 DR> instance, instead of having a 4Meg linear virtual address space
 DR> per process, OS/2 only has a 512 Meg space to support 16 bit OS/2 and
 DR> Dos apps. I shudder to think how much more stable, fast, and small
 DR> OS/2 would be now without the requirements of the Dos/Windoze
 DR> emulation. Even if you don't load them, the architectural
 DR> limitations/compromises are still there.
As you are absolutely correct, I agree with you. Indeed there is already a
really "clean" OS/2 in the works - OS/2 for PowerPC. But for my
best guess we will _not_ see it as an OS/2 for intel processors comparable
to our current OS/2. The only thing we can do at the moment, is to set
IOPL=NO, PROTECTONLY=YES and to remove all DOS/Win device driver relicts
from the system. Well, but seriously, the 512-MB per process limit is not a
real _limiting_ limit these days. I don't know anyone who bounced it :)
Normally if you need such huge amounts of system resources you'll use other
computer architektures like Vax,Sun, etc.

There are only two and a half ways to crash OS/2.
- (1) a process with IOPL I/O privilege can halt the system, other
processes cannot execute the concerning instructions. Only DOS/Win/OS2
16-Bit processes have this IOPL-Bit set. So a 32-Bit OS/2 application just
can't crash OS/2 by itsself...
- (1,5) A bug in OS/2's kernel or device driver layer. 

Most problems result from hardware giving the device driver a hard time.
Like doubled IRQs, I/O-port conflicts, etc.

        __      2:246/2320.18{at}Fidonet 14:75/1010.18{at}NFDN
        |_)     7:7000/7998{at}ZebraNet  81:449/8810.18{at}OS2Net
cu/2.99 |atrick Haller{at}ti-nov1.fbe.fh-weingarten.de

--- FleetStreet 0.95.2wb