What's all that for?

Wed Feb 3 11:13:45 CET 1999

> Well, the message is important (from my point of view), but I hope you
> will receive it just once :-) 
Got only one of this one! :)

> A few months ago, I was searching for options to building a BIOS for a
> custom PC: No graphics card, no hard disk, with sound card, network card
> (well, say: an MP3 player for the HiFi tower). Of course, I wanted to
> avoid anything I would not need and I wanted to have anything I need to
> boot up the thing quickly and easily.
A little off topic maybe, but have you seen the cool car MP3-player at
http://www.empeg.com ? Looks cool!
If you manage to make a working mp3-box, let me know, would be cool to build one
myself!

> As you see, I found the openbios movement, but what I found was, say,
> nothing but a vision. Neither a concept nor an idea. Just a vision. A
> vision of building something similar to what we all already have in our
> PCs. Of course, the idea of reinventing the BIOS-wheel is great and I
> thought: Hey, finally _we_ have the choice of how the PC boots up, we can
> work around any hard disk or booting problem by ourselves. 
You mention the 32bit boot loaders, which seem to be a good thing. I think the
vision is to make a bios that we can add needed/cool features to later.

> But as far as I can see the intentions grown up to now, we just sit in
> here and construct a new BIOS after the Phoenix, AMI, Award etc. with the
> same API (for best compatibility) which means, with the same main booting
> features. The idea of booting directly from 32 bit mode was fascinating
> for me, but what's it for if we have to switch back to 16 bit, implement
> all API calls again just to boot any OS. I thought, the main purpose of a
> new BIOS would not _only_ be the fact that it is OSS. We would have the
> opportunity to make something very new, especially concerning the boot
> process. The idea of putting the remote boot BIOS directly into the system
> BIOS is great - but what's it for if we don't invent something really new?
> 
> Of course, this is a large project, and my ideas make it even larger :-). 
> But, would f.e. Linux be that great if it would have just been a "better
> DOS", say, an "OSS DOS"? OK, it would be for free, but what for? 
And we even have FreeDOS! I think we just need a start, and maybe there could be
configurable if you want a clean 32bit bios, or you want a bios16-compatible
binary. I think that is the big issue here, as I have said before: 

                         Make it configurable!
Then the visions and new ideas can be added and used when wanted.

> OK, back to topic: My idea (better: brain storm) of the boot process is as
> follows: 
> 
> 1) Processor switches to 32 bit. 32bit only, no compatibility API in the
> MAIN BIOS. Well, I think after we implemented all in 32bit, there will be
> not very much space left for such things.
> 
> 1a) If, somehow, 16bit compatibility is needed, we could easily
> implement (if one wants to call it that way) a loader for the old BIOS.
> Well, only if the BIOS behaves "well" and doesn't try to reload parts
> directly from the EEPROM chip (well, there is shadow RAM etc. and I am
> not familiar with it anyway). Alternatively, one (of us) could implement
> a new 16bit API to enable such compatibility mode. Of course, this topic
> is not yet finished in any way.
I mentioned this possibility in a mail a while ago. The problem is where should
we store this old original bios where it can't be accidentally deleted? I like
this idea, and use (2) for loading linux and other OSS oses, and as you say,
maybe we could get M$ to help us make win 2k/9x/NT-loaders later.

> 2) (1) implies, that we need NEW OS LOADERS FOR EVERY OS running with this
> BIOS. Of course, 32bit OSS OSes (FreeBSD, Linux) are easy to handle, Win9x
> with its DOS loader will certainly not run if booted from 32 bit mode
> (i.e. will neeed option (1a) or others). Windows NT could profit from our
> ideas, too, but we would need strong support from MS to locate and setup
> the NT kernel and all its modules and configurations etc. in memory and to
> boot it correctly. Other OSes would profit, too, because they all have the
> same problems: 504MB, 8GB, many others, and of course, the other problems
> with the old, old API16.
> 
> 3a) support for new hardware and, compared to API16, its changed role in
> boot-up in the API. F.e. Network Cards as (a) Input (remote boot) and (b)
> Output (total remote control), of course serial ports, USB, firewire ...
> modularized (f.e. burn-time selectable).
This is things that are new. Would be cool with a totally remote controllable
machine with no display adapter/keyboard at all.

> 3b) Security API, support for random number devices and security devices
> to allow security immediately after/at boot up. Eventually strong
> encryption/authentication, including passphrase etc.
Yes, this is definitely something we would want.

> 4) fast boot up, quick OS restore supported by BIOS (well, similar to
> ACPI).
Fast boot up is something I would really like. Just turn on the PC, and it's
there! :)

> Well, the main topics from my point of view for now. A very flat point of
> view, but a beginning, I think.
I think you mention a lot of things to discuss here. Come on folks, let's here
what you think of this!

One thing I would like is a possibility to load bios images from floppy. That
would be a nice way to test the image before flashing it and leave the box
unusable... Would be really nice during development.

Regards,
Karl Erik

--------------------------------------------
Karl Erik Asbjornsen
karl_erik.asbjornsen at capgemini.no
http://www.samfundet.no/karlea
A: 73540248 P: 73837153 M: 90738663