On 12/12/2011 09:03, Randy Bush wrote:
he is saying that we have long experience of what causes routing table bloat and that we are about to repeat a mistake we made once already.
that's exactly what I'm saying, except for the context in which I say it: we need to repeat the mistake of PI for ipv6 because there are no good alternatives for small-site v6 multihoming, and while historical context has shown us that routing bloat is a problem, it's a problem which was and remains manageable for at least v4 I still maintain my position on 2011-02 that removing the multihoming requirement is not a good idea because we no longer have the {address counting | multihoming} limiters which act as natural barriers in the case of ipv4 PI assignments. Nick