On Mon, 17 Jan 2011 08:39:27 -0600 Jack Bates <jba...@brightok.net> wrote:
> On 1/16/2011 2:52 PM, Mark Smith wrote: > >> It's a basic and necessary feature of ipv6-supporting loadbalancer devices. > >> > > > I don't think that is necessarily the case either. A group of hosts > > with the same downstream address (the global one announced in DNS), and > > a router-type device (i.e. isn't assigned the global address announced > > in DNS) spreading traffic across them such that the same session landed > > on the same downstream host would provide "load balancing" without NAT - > > effectively a more advanced version of anycast where all hosts are > > active, and there is more intelligence in the selection of to which > > anycast host the traffic is sent by the upstream router, spreading the > > load. > > Current methods commonly used are NAT and DSR. v6 implementations of > traditional devices have been using NAT only, and still working on DSR > support. A quick search shows IPVS (linux) for IPv6 with both NAT and DR > support. > I think that is unfortunate and a missed opportunity. However, I suspect these people haven't been on the other side of those sorts of devices very often, trying to troubleshoot problems through them. They might then appreciate how useful end-to-end transparency between the actual communicating end-nodes is after trying to do that a dozen or so times. > By strictest definition, NAT66 has been implemented and deployed already. > > It is not so much NAT66 which is the danger, but where it is deployed. > Having NAT66 in CPE for residential users would be extremely bad. > Special use cases, or environments where the breaking of communications > (certain corporate firewalls), should be considered completely > acceptable. IPv6 gives us the ability to not REQUIRE NAT66, but it does > not remove the market demand for NAT66, though it does lessen it. > > > Jack > _______________________________________________ Int-area mailing list Int-area@ietf.org https://www.ietf.org/mailman/listinfo/int-area
