Pekka, > 2) I'm not sure if this is the right approach. Something better suited > could be found, I believe, in adding functionality to first-hop routers' > ND cache behaviour; a router could answer directly which could be > interpreted like "don't use that address, I just recently saw it used here > by another node.. but if you're really sure you want it, perform DAD as > specified in RFC2461/2".
This is the approach I've been looking at. But there is a problem. Nodes that implement the required ND extension can explicitely inform the router about their IP addresses. Router can learn about others (i.e., regular DAD hosts) via listening to multicast NSs for DAD. No problems so far... But when a router starts with no state (a new one, or rebooting after crash), it might never learn about regular nodes that have already done DAD. So, unless this problem is solved, the applicability of this solution is limited to networks where all nodes support the required extensions.. Any ideas? > Optimistic DAD is a useful optimization because DAD is far more > likely to succeed than fail, by a factor of at least 10,000,000,000 > to one[SOTO]. This makes it worth a little disruption in the failure > case to provide faster handovers in the successful case, as long as > the disruption is recoverable. > > ==> this is totally, and completely wrong. [SOTO] only provide analysis > in *some* cases, in particular autoconfigured vs privacy addresses. For > manually assigned addresses, I believe the ratio is closer to 1:10 or > 1:100 (unmeasurable, of course). I must be missing something.. How come this probability is so high? alper -------------------------------------------------------------------- IETF IPng Working Group Mailing List IPng Home Page: http://playground.sun.com/ipng FTP archive: ftp://playground.sun.com/pub/ipng Direct all administrative requests to [EMAIL PROTECTED] --------------------------------------------------------------------
