Pekka Savola writes:

   o  Among home agents with equal preference, their IP addresses in the
      Home Agent Addresses field SHOULD be listed in an order randomized
      with respect to other home agents with equal preference, each time
      a Home Agent Address Discovery Reply message is returned by this
      home agent.

   o  If more than one global IP address is associated with a home
      agent, these addresses SHOULD be listed in a randomized order.

==> these 3 bullets seem rather redundant optimizations to me.  A MN cannot
depend on them being done (SHOULD's), so it must implement same preference
ordering functions itself anyway!  This seems only useful when the number of
used items exceed MTU, and in that case, even simpler clipping could be
used.

------------------------

Jari Arkko responds to Pekka Savola:

I have looked at this. Unfortunately I don't think we can easily remove
the randomization request. First of all, if it happens to be the case
that the message doesn't fit the MTU, do we want always the same addresses
in the response? I think not, particularly if the N first addresses
just happen to belong to a home agent that is down at the moment.

Secondly, since there is no structure within the set of addresses in
the reply message, it would be hard for the mobile node to do this
randomization in the right way i.e. by weighing the home agents with 
highest priority first.

Finally, a randomization served by the home agent makes the mobile node
a tiny bit less complicated ;-)

The first issue above seems to be the trickiest one to deal with. 
I did think about doing the randomization only for the case where
MTU is exceeded, but I feared that the resulting text might get
more complex than it is now.

In short, I'm not sure we can do anything else than what we are
doing currently in draft 20. If you have any ideas, please post them --
otherwise I keep the current text...

------------------------


------------------------


------------------------


------------------------


------------------------


------------------------


------------------------