> > Hmm. Is there anyway one can help with these things (writing drivers etc) without
> > acutally having JJOS/decaf up-and-running? Is the Linux-host working?
>
> Well, "real" drivers will be impossible to do under Linux.
Yes, some of them. The graphics stuff can be done with SVGAlib, as mentioned
(although it will require root access).
>Perhaps we can
> resurrect Robert's GRUB work... Would it be OK if you could recompile the
> jjos+decaf "kernel" and reboot your Linux box? It would be tedious, but it
> would be workable...
As in pluging something into the Linux kernel? Modules can be added at boot
time, if that's what you're refering to.
But; don't go to any trouble because of me - I'm not experienced in driver
development and such, so I can't really promise I'll deliver.
> How about doing second-tier work? Like, filesystem implementations?
Yes, that's a good possiblity. ext2fs would be a good starting point. I
might even learn something in the process :)
(didn't someone write a FAT implementation a long time ago? I think I've
seen it on the Wiki ages ago when I was new to the project)
>I think
> maybe we could use "raw" disk I/O (I seem to recall something about this being
> possible in Linux -- bypassing the kernel caches...).
It is, but I'm not sure how yet. I'll try to find out... hopefully one
won't need a 2.2 kernel (I'm on 2.0).
>What would be necessary
> to pull this off is arch-specific bytecode (we can already do arch-specific
> native code).
Hmm. You mean some sort of alternative (with less overhead) to JNI? Beats
me. But I guess it needs to be done. How much overheard *is* involved in
JNI? Doing JNI method calls for every disk access might become slow pretty
quickly, even with caching. Or am I misstaken?
--
/ Peter Schuller
PGP userID: 0x5584BD98 or 'Peter Schuller <[EMAIL PROTECTED]>'
Key retrival: Send an E-Mail to [EMAIL PROTECTED]
E-Mail: [EMAIL PROTECTED] Web: http://hem.passagen.se/petersch
PGP signature
