ADAM Sulmicki <adam@cfar.umd.edu> said:
> when you have a problem you want to be able to quickly have a netbootable
> kernel not spend time making initrd images.
The time to build a initrd is a fraction of the time you need to compile a
kernel.
> One could argue that you should have ready emergency disks ahead of time,
> but in my experience by time you need a boot disk something changes and
> they are not what you expect them be. for example I have there this box
> with 460+ days of uptime... the sources for the kernel I used to compile
> it are long gone, heck I even don't remember what ethernet card it has
> anymore (dmesg output is long gone, an I'm too lazy to open the box up,
> but for sure it is not a pci card.. )
If you don't plan ahead, you pay the price. Big news.
> Another issue is development of new machines. when you port to a new
> machine, it come out handy be to able to simply netboot it.
OK.
> perhaps it would be a good idea to instead putting everything into
> initrd, to follow the patch netbsd took and simply make two stage
> boot loader. I think it might be more flexible than initrd is.
It would be _less_ flexible, and much more complex. initrd can contain
anything at all, a copy of netscape if you want to. A first stage
bootloader that passes options to the kernel means that only options planed
ahead are available. Plus there are now two independent ways of setting
this stuff; the options way will be less tested, and quickly fall into
disrepair. Back at square one, with a bloated kernel and unneeded booting
complexity.
-- Horst von Brand vonbrand@sleipnir.valparaiso.cl Casilla 9G, Viņa del Mar, Chile +56 32 672616- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sat Feb 19 2000 - 20:53:14 EST