RE: WGLC ent-analysis-03: DSTM

>    While DHCPv4 may or may not require minor tweaks (e.g., relating to
>    link-layer adaptation),
> 
> => using DHCPv4 over a tunnel is *not* a minor tweak.

Let see. Assume that you have a native IPv6 "site", over which you can
reliably send multicast packets. Assume that we have defined IPv6
multicast addresses that reach all the nodes within that scope. Then,
you can pretty much port DHCP and ARP over that scope, using the node's
IPv6 address instead of the Ethernet address. You obtain a basic "IPv4
over IPv6" solution. It is hard to be simpler than that.

Obviously, there are a few tricky issues. We need a payload type for
IPv4 (done) and another for ARP (not done). We need a way to properly
delineate the scope of the IPv6 "site", probably using an IPv6 prefix
that shall be configured on every node. We need to get the multicast
addresses registered, and probably registered in a way that attaches
them to the selected IPv6 prefix.

The design becomes a little bit harder if we cannot use multicast,
either because it is not reliable or because it creates too much
overhead. But there are solutions -- certainly simpler than doing IPv4
over ATM.

-- Christian Huitema