On Tuesday, Nov 5, 2002, at 11:37 America/Montreal, Pekka Nikander wrote:While RR (Return Routability) works, to a degree, with *address* ownership, it certainly doesn't work with *ID* ownership (that is, if IDs are separated from locators).
RJ Atkinson wrote:
Given the known common presence of man-in-the-middle attacks, I don't see that RR actually buys anything in the way of trust or assurance that one is talking with the party one thinks one is talking with.
Right. It just prevents someone from "stealing" addresses with MIPv6 BUs from an arbitrary location in the Internet, and limits the viable attack locations to those on the path. I *think* (but haven't analyzed in detail) that it would work in the same way with end-host multi-homing based on secondary addresses. And we must not forget the danger of and the prevention of flooding, either. Thus, for example, SCTP should use some sort of RR when falling to secondary addresses. I don't know if it currently does. --Pekka