Jonathan de Boyne Pollard wrote in a message to Mike Bilow:

MB>
  >  JdeBP> [...] but what becomes of Process 1?
  >
  > It starts the first half of PMSHELL and dies.
MB>

 JdBP>   Hmmmm.  It looks like the designers of the OS/2 kernel
 JdBP> never read Douglas Comer's books on XINU.  (-:

To be fair, the original OS/2 architecture had to run on a 286.  That
placed severe limitations on what could be done, especially since the
machine had to be able to switch between real and protected mode, using
Possibly the Worst Kludge Ever Invented.

XINU isn't really the ultimate model here.  The architecture we are talking
of was designed by J. Gordon Letwin, and he was working on this in the
stone age of operating systems.  The core problem is that the device
drivers had to be bimodal, meaning that selectors had to be rigged up so
that they would address the same memory in either real or protected mode,
and that is not easy.  Device driver initializing therefore was a royal
pain.

When OS/2 was upgraded to the 386 most of this became irrelevant, but
compatibility had to be preserved.  As a result, we have device drivers
initialized in a context-free state by a process that dies.
 
-- Mike


---