[no subject]

Apart from that, all binaries were recompiled just because I insisted on
having all source code at hand, and in that process things were
optimized
where possible for a minimum memory and disk footprint.
Other requirements were met as well, such as having it work on a
primitive
shell like /bin/ash and to not require perl.

The only binaries that we had to use as-is, and work around the cobalt
distribution hooks in there, were the lcd-related binaries. I'd still
like
to receive source code to it, but Cobalt just doesn't seem to hear me
asking. I've worked my not-so-quick-but-sure-dirty way around that.

So, to finally answer your question, thanks to the netboot feature of
the
flash inside the RaQ-1 and RaQ-2, it wasn't that difficult to get it
work.
It was, however, a lot of work. Most of this work is unrelated to the
actual
underlying target hardware, and had more to do with the specifications
that
we used: minimal memory footprint, no /bin/bash dependancy, no perl.

Maintenance of this system is done either from the prompt over the
console,
or with yet another self-built software package using ssh to administer
a 
whole cluster of these boxes as a single item, using some sort of a cgi
interface that has nothing to do with cobalt's gui.

-- 
Niels Poppe - org.net bv <niels@xxxxxxxxxx>