> Well, BPF is one way to go.  However, I do not really want to put the work
> into it.  (I have a job, and I am lazy! ;-) )  Why should I do all the work
> of writing a filter,

you don't have to write a filter.  If all you want to do is boot the
machine, use bpf without a filter (but don't put the interface into
promiscuous mode), and you will get _all_ packets received.

> that parses the UDP/BOOTP packets, and also only accepts
> packets for the local interfaces, or ether broadcast addresses.  Why should I
> write all the packet stuffing code for stuffing a UDP/ether packet to be sent
> on a raw interface, when I could have the IP stack to it?

Why bloat the kernel IP stack with functionality that is only required
at boot time ?  This specific functionality belongs properly in
the booting code, in this case, a usermode bootp client.

-- 
/* ronald@cpm.com.my +60 3 241 5232  |  ronald@demon.net +44 181 371 1000 */