> On Oct 16, 2014, at 9:18 AM, Lorenzo Colitti <lore...@google.com> wrote: > > So every time a new prefix comes in, hosts should restart DHCPv6? That > seems pretty dubious (and expensive). I don't think any DHCP > implementation works that way. > > How do they work, then? And why would you describe this as expensive? > How often do new prefixes appear on your networks?
[responding on homenet, since I don't think my response is about what should or shouldn't be done to RFC 3315] Many hosts don't even implement DHCPv6. It's totally optional (RFC 6434). Even when implemented, it may be disabled. So relying on host DHCPv6 behavior (of any sort) for proper homenet functioning, including proper address selection, would be a bad idea, IMO. If DHCPv6 is *required* in any way in order to implement address selection policy in a host, I think we're in trouble [no matter if something is or isn't done to RFC 3315 to make information request behavior more predictable]. I think support for receiving more specific routes in RA messages (RFC 4191) would be easier to get hosts to implement than DHCPv6. In any case, I think there needs to be a solution that *always* works in homenet hosts. Barbara _______________________________________________ homenet mailing list homenet@ietf.org https://www.ietf.org/mailman/listinfo/homenet
