2.1. Host selection of an address [RFC3484] describes an architecture by which a network administrator can define which source address prefixes should be used on datagrams sent to various destination prefixes. This proposal assumes that if remote non-default prefixes are propagated within a network, this technology governs the choice of address. As such, traffic headed to destinations for which there is routing other than the default route will never be sent to an upstream that will discard them.
Hmm. Is it really that simple? If so, why do we have draft-bagnulo-6man-rfc3484-update and draft-arifumi-6man-addr-select-sol? In fact, is it really possible to avoid some sort of reachability probe?
2.2. Host selection of a router
...
if DHCP [RFC3315] is in use, it may be possible to rely on the Router Advertisements bring broadcast periodically. This case requires further thought.
Wouldn't this be an argument for recommending that RAs SHOULD be sent if a site has multiple prefixes in use, regardless of DHCP usage? Brian -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------