[IDEA] make the BIOS the OS.

Eric W. Biederman ebiederman at lnxi.com
Wed Apr 21 02:40:01 CEST 2004

"Chris" <linuxbios at rebel.com.au> writes:

> No not really I was just passing comment
> But I do have a question, based a little on laziness cause I could not find
> adequate mention of my ideas in respect to LinuxBios.
> I have thought about a embedded operating system for some time, for two main
> reasons the first is obvious and well documented
> That is the desire to have a computer come with an OS that runs as soon as
> it is started. perfect for all sorts of things
> The second does not seem to have been covered or at least I have not been
> able to unearth much information about it.
> So I wanted to see if it was in the charter of this group.
> What I am referring to is a method to enable device manufacturers to have a
> consistant interface to the hardware.
> If an OS writer such as M$ does not support a device it is very hard to get
> that device to work if at all.
> Manufacturers have to write device drivers for many different os's and
> versions, consequently they only write for the most popular and forget less
> popular and older systems.

No, OS developers do.  Manufacturers only need to provide documentation.

> What I envisage is a common device interface. Allowing any manufacturer
> create one piece of software for their hardware and except for constraints
> such as recompiling for different cpu's etc their hardware will work on any
> system. (within reason).

You are describing an OS.  An OS is the hardware abstraction layer.  All
attempts to the contrary are non sense.

> So all you need is one API written for each OS that gives the the
> manufacturer access to that particular device REGARDLESS of whether the OS
> was written to handle that device.
> For example NT 4.0 has no idea what a USB device is. So you can't use one.
> period! this was the main reason I went to XP and its griped me ever since.

This is a problem with closed source not with the driver development model.
If you have the source and modification rights to it.  When something
is important enough you can do whatever you like.

> I am sure that in the linux world it is annoying to have to write new
> drivers for every little piece of hardware that comes out.

If they are close enough a generic driver can be written.  This is why
there are standards for hardware manufacturers.
> Wouldn't it be nice if the device manufacturer could simply recompile their
> driver to suit linux or sun or mac(bsd)

Only if it was not utter crap.  Occasionally that happens.  Mostly hardware
manufacturers are not good at writing code.
> the only real place for such an interface is at the BIOS level, lest OS
> writers pervert it for their own good.
> and if that BIOS was provided licence free to Mainboard manufacturers
> wouldn't that be something.

It is almost impossible to update a BIOS in the field, (human factors).
And it is very dangerous.  Who wants to update take the risk of killing
their machine just so their new whatzit will work?
> Sorry If I am going over old material, As I said I have not found articles
> relating to this.

I don't know if it has ever been written about.  But the idea comes up often.
You would probably find some sympathy at Intel among their EFI group.  But
then EFI is peculiar DOS/WIN clone anyway...


More information about the coreboot mailing list