Syslinux - Oct 2012 - Thoughts About Direction?

Some thoughts and questions for Syslinux:

1. It?d be great to see the multi-disk (or multiple FS) support completed. ?

2. It?d be interesting if Syslinux could support a CPIO FS, which has been
discussed before.  This would make sense for a new Syslinux "kernel
variant."

3. With Matt Fleming?s ELF support, we could allow the FS drivers to compile
into "stand-alone" ELF modules?

4. Syslinux, having multiple FS support, could have a CPIO archive embedded for
all variants, similar to the Linux kernel.  We could shift from choosing what to
link in, to choosing what to CPIO in.  (I'm thinking of optional filesystem
drivers not required for the variant's "boot" filesystem.)  Is
this sensible, or does the more efficient linking seem better?  Or both?  (Maybe
both capabilities are best; link in the CPIO _and_ "boot" FS driver,
but allow a user to CPIO in other stuff.)

  - As well as continuing to be able to use a CPIO archive passed by another
boot-loader, for the kernel variant, as established in #2.

Do any of these ideas have merit?  Do they fit with the vision of the future of
Syslinux or not really?

- Shao Miller

On 10/26/2012 03:03 PM, Shao Miller wrote:
> Some thoughts and questions for Syslinux:
> 
> 1. It?d be great to see the multi-disk (or multiple FS) support completed.
?
> 
> 2. It?d be interesting if Syslinux could support a CPIO FS, which has been
discussed before.  This would make sense for a new Syslinux "kernel
variant."
> 
> 3. With Matt Fleming?s ELF support, we could allow the FS drivers to
compile into "stand-alone" ELF modules?
> 
> 4. Syslinux, having multiple FS support, could have a CPIO archive embedded
for all variants, similar to the Linux kernel.  We could shift from choosing
what to link in, to choosing what to CPIO in.  (I'm thinking of optional
filesystem drivers not required for the variant's "boot"
filesystem.)  Is this sensible, or does the more efficient linking seem better? 
Or both?  (Maybe both capabilities are best; link in the CPIO _and_
"boot" FS driver, but allow a user to CPIO in other stuff.)
> 
>   - As well as continuing to be able to use a CPIO archive passed by
another boot-loader, for the kernel variant, as established in #2.
> 
> Do any of these ideas have merit?  Do they fit with the vision of the
future of Syslinux or not really?
> 
I think they all do (although 3 may or may not be worth it.)  The issue
is mostly a matter of resources.  Getting to EFI support is by far
priority 1.

Getting 1 done would definitely be a win.

	-hpa