[sdiy] bootable linux cd

[Magnus Danielson]

From: Rainer Buchty <buchty at cs.tum.edu>
Subject: Re: [sdiy] bootable linux cd
Date: Sun, 27 Jul 2003 22:38:11 +0200 (CEST)

> > DOS might, in fact, be a good place to start experimenting because it
> > doesn't get in the way very much.
> 
> On the other hand, it doesn't offer very much which you can't re-do
> yourself in a few minutes, but OTOH lacks a lot of what you really need
> (and which needs to be patched into using sometimes rather strange ways).
> Besides, it's entirely IBM-PC centric.

Let alone that hardware is much more complex these days than it used to be.

> So why not creating a very own light-weighted "MusicOS" just providing the
> necessary framework like
> 
> - disk I/O
> - serial & MIDI I/O
> - USB
> - Ethernet / TCP/IP stack
> - video output
> - system timers
> 
> and on a higher level
> 
> - task switcher
> - memory management

... you also want to make sure you make good use of VM since that saves a
bundel. You want a good disk I/O to handle realtime streams. You want to have
rate monotonic scheduling as well as some spare asynchronic for some control
processes. 

> A friend of mine did something similar (lacking USB and TCP/IP)  for a
> Motorola Dragonball based system he designed being the heart of his
> home-brew multimedia system. It's just below 50kB, where USB would add
> another ~50kB (he just did a port for the X-Box FreeBIOS), TCP/IP around
> 4kB.

Trouble is when you want to be generic, it all adds up. If you do something
specific it is _much_ easier, since you (hopefully) know what you can trim off
since it will never appear.

If you want to hack another OS, please go ahead. Personally I think it is much
easier to do something good with what exists. Much of the OS-theories I hear
about doesn't match well with real life anyway and while the Linux kernel isn't
the most RT-system we know (there are hacks) it does have all the features that
I think is needed if you go for 2.5/2.6. It is fairly easy to trim off the fat
you don't need and there is alot of people doing it, so it is not big science,
you might even buy books about it.

Sorry, but writing a new OS, even if dedicated, is not what I think is the best
solution if you go beyond a certain level. I think you spend more time and
efforts on the wrong things that way and you have to stand fairly much on your
own. If you want to do a music app, you want to concentrate on doing that well
rather than concentrating on doing an OS well.

Cheers,
Magnus