* Bill Arbaugh <[EMAIL PROTECTED]> [020520 19:12]:
> Agreed. But, a specialized mini-os is huge effort. I'd prefer a small
> general purpose OS. Just look at the wishlist and you'll somebody
> wants a little bit of everything.
There already is an effort for this: OpenBIOS. It provides a small
virtual machine, called an FCode evaluator, and a bytecode tokenizer and
detokenizer. Big advantage is there is already hardware that is
supported. Big disadvantage is that the virtual machine and the driver
API are quite some time before prime time and can not be used yet.
Still I think this is the way to go, because these bytecode drivers are
a lot smaller than writing an size optimized driver in assembler or C.
The current implementation of the virtual machine, paflof, is very clean
and small, it fits into less than 16k, uncompressed. Using Open
Firmware's object oriented design, it should be possible to write new
drivers with a minimal overhead.
There was not much going in the last month, as we're currently only two
overloaded people writing code for OpenBIOS and desperately need more
hardcore hackers to get the thing going, but it's not an unsolvable
task.
Best regards,
Stefan Reinauer
--
The x86 isn't all that complex - it just doesn't make a lot of
sense. -- Mike Johnson, Leader of 80x86 Design at AMD
Microprocessor Report (1994)
