On 2004-02-24, Francis Dupont wrote: > [Greg Daley wrote:] > > > This is not going to happen in Nick's optimistic DAD draft. > > => I disagree: even in Nick's draft the second system has both > a service disruption and has to switch to another address. > So as to be the second system is just a matter of race Murphy's law > implies it will be the important one.
... and the result to the second system of an address collision under unmodified DAD is ...? We're talking about an extremely rare event here, an address collision of _well-distributed addresses_, eg: 3041-like or SEND-CGA-like addresses. Opti-DAD explicitly excludes manually configured addresses (although it may be used with EUI-64s). Just how many nodes _per network_ were you thinking of having? And when it happens, the first node may get a few stray packets and the second node will have its address configuration delayed by a few milliseconds -- but not for as long as it would have taken to perform DAD anyway. I don't see the disaster in this. cheers, -----Nick -------------------------------------------------------------------- IETF IPv6 working group mailing list [EMAIL PROTECTED] Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
