Mail Archives: djgpp/1999/12/02/00:50:59
> I don't think there's something principally new to be done about
> DOS, at least as I see it. My system currently runs on top of DOS and
> is not bound to any specific vendor or version. It takes like 200 K of
> memory and doesn't impose further system requirements.
> The most significant difference probably is that I'm still flexible.
> I'm not sure by now that I want to develop a whole DOS package. I can
> customize a multitasking solution to a legacy DOS-based system or
> custom DOS-based embedded system rather than pull you to use something
> given and unchangeable.
Interesting. You may see how the kernel in Lineo's EMM386 behaves -
especially for making the mouse interrupt live under each task
when it has been set before launching multi-tasking mode, concurrency
along with the built-in DPMI server (works!). Their API documentation
may be an inspiration too (if still online...).
Converging to a thing that works on more than 1 machine may be non-trivial...
How to share resources can be a challenge, too: DR-DOS shares a
RAM-disk (and other disks), MS-Smartdrive cache, but (hopefully) the
mouse cursors are independent on each task, as well as graphics
modes, screen memory etc.
I suspect there is a lot of subtle choices to do (the IRQs, the
network resources). The DR-DOS multitasking kernel is also helped by
structures prepared in the real-mode machine (IBMDOS / IBMBIO / COMMAND)
for a switching to a multitasking mode. Look for instances of
the word `multitasking' in the sources of DR-DOS real-mode kernel.
It can help seeing how to implement things in a vendor-independent way.
Anyway, good luck and there will surely be plenty of alpha testers
on these NG...
--
Gautier
_____\\________________\_______\
http://members.xoom.com/gdemont/
- Raw text -