mcr@sandelman.ottawa.on.ca said:
>   A workaround, I just realized, is to put a ":" at the end of the
> gateway name.

That's not just a workaround. The idea is to pass
gateway:netmask at once.

> I also think that the netboot shouldn't require that you specify
> a gateway.

OK - but since bootparamd doesn't answer at all if an entry
is not present, it takes timeouts and retries until the
booter can realize that.
The kernel bootparam code does it that way, but the delay
in the boot process is suboptimal.

btw, this reminds me of a problem I wanted to fix a while ago
but forgot: The gateway passed by our "bootparamd" in the
"whoami" call is not worth much (it's the loopback address)
unless a valid one is specified with -r.
This doesn't affect NetBSD clients, only original SUNs.
NetBSD clients don't believe this gateway address anyway
if you don't force them to do so. (The whole protocol idea
is not well thought, and the gateway stuff has always
been broken.)

best regards
Matthias