I think you misunderstood what I was saying there. I didn't say that the behavior of the end node must change - only that either all nodes attaching to the network get NETLMM service or none of them do. There can't be a mix - otherwise, on shared media, there will need to be multicast RAs, which will lead to the problem Marcelo was bringing up.
Vidya > -----Original Message----- > From: Jari Arkko [mailto:[EMAIL PROTECTED] > Sent: Tuesday, August 08, 2006 3:17 PM > To: Narayanan, Vidya > Cc: INT Area > Subject: Re: [Int-area] IPv6 addressing model, per-MN subnet > prefix, and broadcast domain > > > >all nodes attaching to links within the NETLMM domain will > have to follow NETLMM defined behavior - I don't see how > there can be a mix of nodes on the network, some that obtain > NETLMM services and some that don't. > > > > > We need to be very careful about "NETLMM defined behaviour". > In particular, what part of this behaviour is in the network > and what is in the host. The value of a NETLMM solution > depends in part on the host not needing to have specialized > support for NETLMM. My ideal model of the solution is where > the hosts are happily doing what they would be doing anyway > (DAD, DNA, DHCP, etc) and the network makes sure that their > globally routable addresses do not need to change unless you > leave the entire network. This also makes it possible to > deploy NETLMM on top of an existing, non-LMM link technology > without having to update the mobile nodes in sync with the > deployment. If we start to require changes in router > discovery or other parts of the MN behaviour this causes > complications. > > (Having said that, I am not aware of the current documents in > NETLMM proposing any deviations from the ideal model.) > > --Jari > > _______________________________________________ Int-area mailing list [email protected] https://www1.ietf.org/mailman/listinfo/int-area
