A version with biosdisk, pc and ext2
before: 24800
after: 24687
But another matter is expandability - preboot hooks are important. 
Without them implementation of sendkey or map is impossible. And adding 
preboot hooks to kernel would increase its size by 150-200 bytes.
Also I don't see how this change would prevent integrating e.g. _linux 
in the core. But modules for loading kernels are often reltively big 
since they have to treat multiple versions of boot protocol.
It also seems that some parts of my patch were missing. I send a 
completed version.
Thanks
Vladimir 'phcoder' Serbinenko
Robert Millan wrote:
> On Mon, Sep 08, 2008 at 10:11:33PM +0200, Javier Martín wrote:
>> El lun, 08-09-2008 a las 22:48 +0300, Vesa Jääskeläinen escribió:
>>> phcoder wrote:
>>>> Hello. As I said in another email there is no need for it. I send a
>>>> patch for it.
>>> Doesn't this break our rescue mode ?
>> How would it? As with other commands, there would be two versions of it,
>> one for rescue and one for normal. Besides, what is the point in having
>> the "boot" command in kernel -and thus always available- if you can't
>> load a kernel to boot?
> 
> Vladimir, you're proposing an important change because it modifies the goals
> of the rescue mode, but I see there's neither significant support for or
> against your proposal.
> 
> I think it would help if you provide numbers on how much size do kernel and
> boot.mod have before and after the proposed change (on i386-pc).
>