Booting the installation over network

[Nikita Schmidt <cetus@snowball.ucd.ie>]

I've just added a rule into the boot-floppies' Makefile that generates a
network-bootable installation system image (tftpboot.img).  I've also
built such image using generic kernel and root.bin from David's latest
disk set (ftp://genie.ucd.ie/pub/alpha/test/tftpboot.img).  It worked fine
for me, so it might be useful to move it onto David's page for people to
try.  It can be booted with just `boot ewa0' - the `root=/dev/ram'
parameter is passed automatically (but can of course be overridden with
-flags).

Technically, this thing needs an unpacked 2.2.x kernel source to build, so
I've not yet included tftpboot.img as a default target.  Moreover, the
kernel source must be (at least) partially compiled - bootpheader uses two
small libraries from there.  I don't know whether it is enough to just say
we need a compiled kernel source, or is it better to do the whole thing
automatically (unpack, configure, compile the libraries we need).

Nikita