dhcprequest function

[dhcp.50.CHRIS94561 at spamgourmet.com]

On Mon, 29 Mar 2004, Ted Lemon  - mellon at nominum.com wrote:

> On Mar 29, 2004, at 6:36 PM, dhcp.50.CHRIS94561 at spamgourmet.com wrote:
> > Am I missing something? Is there some clarification to be made? Is
> > there a clear statement somewhere including "MUST" or "required" that
> > clearly illustrates the requirement to broadcast NAKs?
>
> What's going on here is that there are flaws in the RFC that make it
> somewhat self-contradictory.   Most implementations just do what seems
> reasonable, and generally interoperate reasonably well, but there have
> been serious interoperability problems because of imprecise
> specifications.   In this case, though, I think it's pretty clear what
> to do.   If you unicast a DHCPNAK directly to a client in the
> REQUESTING state, you are not following the protocol.

I dunno about pretty clear. Were I implementing this I'd be inclined to
ignore the multicast requirement and use the broadcast flag as my
barrometer but that's just me.

Chris