Re: New issue 16: No packets from other end?

I am not sure whether there is a consensus for this issue or not. But the
arguments are convincing (at least for me) that we need some IKEv2 message
to find the working pair of source and destination address. To make sure
i understand how this message is used, let me consider the following three
cases.

1) Mobility without multi-homing: In this case a single pair of address exists between
    the two nodes. When one of the node moves, MOBIKE detects at some point
    in time that there is a change at layer 3 e.g. using DNA/MIP6. MOBIKE
    learns about the new address using whatever mechanism the local OS provides,
    and sends an  address update to the other end. In this case, there is no need
    for any extra probing to look for a working pair.

2) Multi-homing without mobility : In this case there are multiple pair of addresses
    between the two nodes (Si, Dj). IKEv2 starts off with some address pair
    (S1, D1). At the end of successful IKE SA establishment, IKE knows the working
    pair of address which will be used for the further communication. MOBIKE kicks
    in at some point and each node updates the known local addresses with the other node.
    Whenever a new address is obtained e.g. a link to the ISP becomes UP, the other
     end is updated with this new address (it is really a matter of policy whether it is
     updated or not). Whenever an address ceases to work (hints from outside of MOBIKE),
     and that address is part of the current working pair, each end discovers this independently
     and starts to find a working pair. This is when you need a message at the IKEv2 level to
     find which address pair (Si, Dj) works. (I still don't understand the details of how both ends
     agree on a single pair of address. But that detail can be discussed later i guess) 

3) Multi-homing with mobility:  In this case there are multiple pair of addresses between
    the two nodes (Si, Dj) and one of the end is mobile. This case is not much different
    from (2). Mobility makes an address disappear or a new address appear. Whenever
   the node moves, an existing address becomes invalid. And if it is part of the existing working
   pair of address, MOBIKE can try and choose a new address by probing the other end
   (using the new IKEv2 message) or just update the other end with the new address just
   obtained (and only when that fails, start finding a different working pair).

Does this make sense ?

-mohan