"Preston L. Bannister" <[EMAIL PROTECTED]> writes:

> Is this mainly because of the issues with size, or something else?

Size is certainly what led me to seriously consider alternatives.
 
> I'd expect something like etherboot to boot faster (unless of course the
> Linux version in BIOS was the one you wanted to run).  On the other hand I'd
> expect you to have more work to adapt to new hardware.

etherboot loads a final kernel faster.  Actually writing a driver is a small
amount more work, but it is not work that always must be undertaken.
Additionally the conditions when a bootloader runs are noticeably
different from when a kernel runs.  The best example I have is that
the Linux kernel does not get hard drive spinup correct.  So I would
have had to hack that driver anyway, thus their would have been no
savings of work.  Given that the bootloader is motherboard independent
the developer pool is potentially as large as the Linux Kernel
developer pool.  At this point using the Linux kernel appears to be a
false economy. 
 
> If the BIOS ROM (or the equivalent) was big enough to hold the Linux version
> you wanted to run, does this change your point of view?

The systems I build a fundamentally general purpose systems, so I putting
the final Linux kernel in the ROM is not something I would do.  

And size is an issue that simply will not go away for me.  So improving
the Linux kernel to act as a better bootloader does not look to
be an especially productive course of action.

Eric

Reply via email to