1. Implement a tool that sniffs all incoming traffic and responds with 
   ICMPv6 NA and then Echo Reply to all NS and Echo Request packets,
   making any IPv6 address on the local network successfully ping6-able.
   Use Scapy or libdnet.

   Caveat: Netfluke's sniffing loop seems to be buggy on MacOSX and OpenBSD,
   crashing or exiting with 'Resource busy'. This may be due to the loop
   itself, or to libdnet's python bindings. 

2. Configure a router (or write a tool) to emit RA for any IPv6 
   nodes on the local network to route their traffic to a Hurricane Electric 
   tunnel. Ping Google (or somesuch) from a host behind such router.

   Caveat: no NAT yet, but none needed: HE gives a /64 routing prefix
   of (globally) routable IPv6 addresses. 

3. Test the capacity of a network card's programmable multicast address
   filter (p. 294-295, Ch. 6.4) by setting filters and trying host 
   response time (e.g., to pings). Extra interrupts to CPU for handling
   multicast packets should affect either latency or bandwidth (or both)
   of the host.

4. Experiment with the order and number of chained IPv6 extension headers
   and options: E.g., hop-by-hop header if any should only follow the IPv6  
   header; what happens it more than one copy of it occurs there or elsewhere?
   Same for destination headers and fragment headers. Will these be rejected
   or accepted by hosts? routers?