On Thu, 4 May 2006 22:18:45 +0200 (CEST), Hubert Feyrer <hubert@feyrer.de>
wrote:

> On Thu, 4 May 2006, Steven M. Bellovin wrote:
> > No, I don't like it, either, which is why I conclude that this isn't a
> > fruitful path to pursue.  Your goal is very reasonable, but I don't think
> > you can do it the way you've suggested.
> 
> I had an idea on the whole "setup my wlan laptop as router" scenario:
> 
> Initially, you have no address at all.
> Then, you ask for one ip address (via DHCP, autoconfiguration, whatever)
> That's the standard 'laptop' case so far.
> 
> When you want your 'laptop' to become a router all of a sudden, you 
> basically need more address space. I wonder why not just ask for it, as 
> you do in the first case (DHCP, ...), and then advertize it on the 
> "internal" network/interface, and just route it.
> 
> I'm not sure DHCP or anything on this planet is prepared for the second 
> inquiry, but it would prevent the need for ad-hoc addresses and 
> NAT.
> 
Right.  And the easiest way to ask for such an address is DHCP with a
different MAC address in the packet (note that DHCP wants the MAC address
in the body of the packet; it doesn't use the Ethernet header for that).
That will give you multiple IP addresses, each with a different MAC
address.  Where do you get those other MAC addresses?  From your clients
-- that's the easy place.  And the IP addresses you'll get will be on the
same net as your public interface, so at best you're doing proxy ARP.  At
worst -- well, you're already so close to a bridged net that you may as
well do that in the first place -- why write new code and new protocols?

		--Steven M. Bellovin, http://www.cs.columbia.edu/~smb