Details of stateless autoconfiguration
	address space is split in two 64bit halves
	upper 64bit are used to specify a particular network segment
	lower 64bit are used for individual nodes in one segment
	lower 64bit are generated from 48bit mac address with 'fffe' in the middle
	potential problem: privacy

DNS and IPv6
	forward resolval (hostname -> address)
		ipv4 uses 'IN A' record
		ipv6 uses 'IN AAAA' record
		a particular hostname can have A and AAAA records
	reverse resolval
		uses .ip6.arpa.	suffix
		uses hexadecimal instead of decimal notation:
			4.4.0.0.0.0.0.0.0.8.7.0.1.0.0.2.ip6.arpa.
	portable applications under *BSD/Linux do round-robin between all records, with a preference of ipv6 for the first try.

BSD Sockets API and IPv6
	struct in_addr has become in6_addr
	new API's like getaddrinfo() instead of gethostbyname() support _both_ ipv4 and ipv6
	apart from that, everything is the same.

configuration under linux
	router/gateway
		runs radvd or zebra for sending router advertisements
	client
		just has to load 'ipv6' module and configure an interface up
		recevies prefix-advertisement(s) and auto-configures address accordingly

IPv6 specific security issues
	packet filter has to explicitly allow neighbour discovery, since it's inside ipv6/icmpv6
	special attention to option headers
		most sites won't want routing or hop-by-hop options
	neighbour cache DoS:
		compare with existing neighbour cache issues in large (/16) networks
		in ipv6, the standard is /64 for every segment (!)

	one advantage: port scanning of whole networks way more difficult :)