[net.micro] Small micro OS for Education

knudsen@ihnss.UUCP (04/06/84)

If you want cheap hardware and a REAL OS for educational tinkering,
get a bunch of Radio Shack Color COmputers with one disk drive and
the OS-9 system.  Inspired by UN*X, OS9 supports multi-tasking,
multi-users, I/O redirection, etc.  It is well documented and fairly
easy to tinker with, being entirely modular in construction.
An assembler is included, and a FULL C comp9iler is available for $100.
SOme possible disadvantages:  source not supplied (but disassemblers are
readily available); no swapping or paging (Version II shuffles & relocates
using an MMU, but needs bigger 6809 system than Coco).
However, OS9 uses interrupts, timers, shcedulers with variable task priority,
etc.  Lots to play with in one semester, eh?  mike k

emjej@uokvax.UUCP (04/12/84)

#R:ihnss:-200600:uokvax:3400039:000:573
uokvax!emjej    Apr 12 05:54:00 1984

/***** uokvax:net.micro / ihnss!knudsen /  9:45 pm  Apr  6, 1984 */
SOme possible disadvantages: ... no swapping or paging (Version II
shuffles & relocates using an MMU, but needs bigger 6809 system than Coco).
/* ---------- */

I heartily agree with Mike Knudsen's sentiments about a machine for
teaching OSs, but wish to point out that not swapping or paging is
a disadvantage only in the context of teaching--the decision not to
swap is a MAJOR win for OS-9 over Unix and its clones. (If you can
wait, OS-9/68000 Level III for the 68020 will page...)

						James Jones

dont@tekig1.UUCP (Don Taylor) (04/14/84)

>SOme possible disadvantages:  source not supplied (but disassemblers are
>readily available);

You have got to be kidding.  Teaching the fundamentals of operating systems is
going to be hard enough, let alone turning them loose with a disassembler and
a block of object code.
And now for something completely different.  "Operating System Design, The XINU
Approach" by Douglas Comer, of Bell Labs, Published by Prentice Hall, 1984.
Rather than showing lots of different ways that something might be implemented,
he picks one way, and does a, layer by layer, design of a whole system.  He
starts with link lists and works all the way up through a file system.  He
states out front that this is not UNIX compatible internally.  On the other hand
he claims a big chunk of this fits in 4k bytes, and that the sources are
"available at nominal charge".
Has anybody got any information on this tape?
Don Taylor
tektronix!tekig1!dont

sdyer@bbncca.ARPA (Steve Dyer) (04/14/84)

Much as I like OS9, I find the comment that "the decision not to swap is a
MAJOR win of OS9 over UNIX and its clones" a bit strong.  What this means
is that if you're not careful, especially in a dynamic multi-process
environment, memory gets fragmented, and processes which ran successfully
at one point in time won't have sufficient contiguous memory to start up.
Sure, this is ordinarily an intelligent decision for floppy-only based
systems wherefrom OS9 springs, but *I* would like to be able to make that
decision myself, thank you very much.  Especially with the advent of low
cost hard disks, the additional system slowness of a swapping system
might possibly be worth the increase in flexibility and predictability.
I could imagine micro-based versions of UNIX which would allow the user
to configure it to swap, or not, depending on the system-wide requirements.
As it is, OS9 avoids the problem altogether--that's not exactly a win,
that's a cop-out.
-- 
/Steve Dyer
{decvax,linus,ima}!bbncca!sdyer
sdyer@bbncca.ARPA

emjej@uokvax.UUCP (04/18/84)

#R:ihnss:-200600:uokvax:3400042:000:1364
uokvax!emjej    Apr 18 09:36:00 1984

/***** uokvax:net.micro / bbncca!sdyer / 10:12 pm  Apr 14, 1984 */
>What this means is that if you're not careful, especially in a dynamic
>multi-process environment, memory gets fragmented, and processes which
>ran successfully at one point in time won't have sufficient contiguous
>memory to start up.

True, for Level One. (Too bad a CoCo won't run Level Two...)

>Especially with the advent of low
>cost hard disks, the additional system slowness of a swapping system
>might possibly be worth the increase in flexibility and predictability.

This is not a rhetorical question--how fast are those inexpensive hard
disks? I quote from Robert Gammill's "Position Paper on OS's for Small
Systems: the Unix Operating System," SIGPC Notes, v.4 no. 3, fall 1981:

	Many of the smaller Winchester disks now entering the market
	use stepper motors for head movement, providing average
	access times of 200 msec or more but keeping costs low. Others,
	using more expensive voice coil actuators, provide average
	access nearer 20 msec but at higher cost. Such differences will
	be very noticeable in Unix performance, since its performance
	under load is so disk access and main memory dependent....We
	are likely soon to begin hearing complaints from those who
	have had the bad judgement to try to run Unix on hardware
	unable to properly support it!

/* ---------- */