caveat: i have not checked the entire thread yet, so it could be a duplicate of someone's words. > First, giving each end user a /64 means that the user can > have up to 2^^64 devices at their site/home/office. That 2^64 interface ID is picked to reduce the possibility (or "probability" if you prefer) of address collisions when we use RA. it does not need to be tied with IEEE 802 MAC address - in fact, if i were to use bridged ethernet and wireless segment, i prefer to use MD5(hostname) just for the ease of migration between wired and wireless. (i guess i would like to mention a bit about session layer protocol, but that is outside of the point) > Second, Ethernet bridging is a well understood technology > and it works just fine even with very large numbers of devices. that's true, but when link MTU is different, it gets very nasty. > Third, DHCP is a well understood technology that is deployed > at millions of sites world-wide and generally works quite well. ISC dhcp server takes like 5 minutes to reboot when it serves /16 segment, due to the re-read of /var/run/dhcpd.leases. dhcp server is a single point of failure, unless you have a replicated lease database among dhcp servers. if there's implementations or practices, i would like to know that. > Fourth, lots of folks (me included) happen to find it convenient > to use NAT between my site/house/office and my upstream provider. eeeeks. i wouldn't re-start this religious war, but i just can mention that (1) NAT is a single point of failure, and (2) NAT is not future-proven so you have to upgrade it forever. itojun _______________________________________________ Ietf@xxxxxxxx https://www1.ietf.org/mailman/listinfo/ietf