Hi Mads!

(Mon Nov 16 1998) Mads Orbesen Troest wrote to Andrew Belov...

 MOT> Actually, I don't really see why it should be impossible to implement
 MOT> "real" multiple input queues, as PM programs have
always, since day 1,
 MOT> allocated/deallocated a "message queue" through a
standard API. It should
 MOT> be very possible - but demanding a major rewrite of some central PM DLLs
 MOT> - to fix this;

...and, again, some software will refuse to work. We've learned it with
Warp Fixpak 17. Some people were still using PMWord by then and a lot of
noise was made.

 MOT>  I fear the problem is that no-one knowing anything about
 MOT> the old design of OS/2 remains at IBM and they are simply afraid to
 MOT> touch something that's not /really/ broke.

However, they reworked the kernel in OS/2 4.5. So, they still know the OS/2
internals.

I think it's IBM's policy not to implement something its corporate clients
won't need. IBM's infamous "patches" (PM_ASYNC_FOCUS_CHANGE,
CMPXCHG8.SYS, and a CONFIG.SYS switch in OS/2 4.5 to specify the memory per
process limitation) exactly reflect their policy.

 MOT> Unfortunately, I find it increasingly hard to live with a system
 MOT> with the lame SIQ problems OS/2 present from time to time... :-\

The problem is also that OS/2 usually will survive after an interface lock.
If it would trap, or compromise system integrity in any other way, the SIQ
problem might be fixed.

                                                       Bye.

---