Mail Archives: djgpp/1997/09/15/15:01:18
There are three reasons for desiring a 32-bit DOS: (1) Windows 95 and Windows
NT do not run all existing DOS programs, such as my KnowledgeMan runtime
system, my Alsys Ada compiler, my Oracle 5.0 for DOS developers edition,
programs which use extensive amounts of extended memory, programs which
write to the robot controller ports, and programs which light up dots
directly on the screen;
(2) realtime programs run a lot slower, such as programs that need control
of the mouse, keyboard, dots on the screen, parallel ports, serial ports,
etc., although the combination of the system developers kit and the game
developers kit under NT give some control back, they do not approach
the speed of DOS; and
(3) DOS does not have perfect multi-programming-ness, and needs help
help from DOS extenders to get certain features: good device drivers
for the keyboard and mouse and ports, reentrant system calls (software
interrupts), multi-threaded system calls (disk I/O, timer, interprocess
communications, memory access, un*x utilities, and a realtime executive to
round-robin around the tasks in memory preemptively handling interrupts).
Since djgpp solves the first 2 and a portion of the third, the desire
might be better solved by taking all the multitude of labor being
proposed for this new 32 bit operating system and add the missing
multi-threaded features into djgpp. If this labor has something
emotionally against adding multi-thread capability into djgpp, then
add DOS capabilities into something like Linux. If there is not
enough labor to add the extra features into djgpp, then it appears
there is not enough labor to create a new operating system, which
would be quite a bit more work (if you maintained the backwards
capability, without which it would not be reasonable to call it
DOS32).
Mike Brenner
- Raw text -