RE: [mobile-ip] Re: hmip and mobile routers

[Michael Thomas <mat@cisco.com>]

Hesham Soliman  (ERA) writes:
 > I don't know if this mail will make it, but I'll try.
 > 
 > I'm not sure who sent the original mail because I 
 > only received Thierry's response.

   C'est moi.

 > > > Section 6 describes a mechanism of how mobile
 > > > routers might be supported by HMIP. However, I
 > > > think there is a fundamental category error being
 > > > made here. Namely, it is expected that behind the
 > > > mobile router are, in fact, mobile nodes.
 > > 
 >      => There is no error in that. If they are not mobile 
 >      nodes then HMIPv6 does not solve that problem.
 >      Clearly there are two cases to consider, 
 > 
 >      a) the nodes behind a router have (H)MIPv6 implementation
 >      b) They don't

   Yes, but they could just as easily be mobile nodes
   behind mobile routers where the mobile router functions
   as a home agent for those mobile nodes, ad infinitum.
   I don't think that HMIP brings much special to that
   situation: any place that you could put MIP, you
   can also put HMIP, but that should go without saying.
   In other words, this doesn't look any different
   that multiple layers of hierarchy.

   Case (b) is what I would consider to be the "mobile
   router" problem.

 >      HMIPv6 addresses a and not b. Thierry's draft 
 >      addresses b and not a. 

   Thierry's draft addresses one aspect of (b), but
   probably not the whole story. It may have gotten
   lost in the email meltdown, but I wrote a long
   post about some of the detrimental aspects of
   the home address destination option, one of which
   is (b). I'll probably repost it.

 > > As currently defined,
 > > > mobile IP is only a requirement if the device
 > > > (router, host) is itself mobile. I do not see why
 > > > we should expand that requirement to anything
 > > > which is *behind* something that is mobile. In
 > > > fact there could be a huge downsides to doing so,
 > > > not to mention a whole lot of uncharted territory.
 > > 
 >      => What is the down side ? This is an unfinished
 >      statement. 

   Did you read the planet-earth-behind-a-mobile-
   router question? I thought it was part of this
   post, but maybe it got clipped by the time you
   saw it.

 > > > Also: what happens if there is more than one
 > > > router in between MR and the stationary host
 > > > in MR's subnet? 
 > > 
 >      => The MAP option is propagated as usual.
 > 
 > > What happens if there's more
 > > > than one MR? 
 > > 
 >      => What happens if there is more than one AR ?

   Well, that's easy. I forward packets to the first one,
   and I immediately lose track of what happens from there on out.
   In the layered ([H]MIP) case, it would seem that each mobile
   might need to deal with binding updates to the
   HA for each time a mobile router moves too. 

   If this isn't what you mean, it's probably
   because the draft is unclear, and in either
   case it would be good to clean in up to be
   explicit with the interactions.

 >      => The section says that the MAP (MR) sends a MAP
 >      option to the "mobile network" with its on link
 >      CoA (LCoA) as the MAP address. 
 >      Does it need more explanation ? We can add 
 >      more.

    What is the mobile network? What happens if
    the mobile network in turn contains a mobile
    router? Is it a flood fill? What do the mobile
    nodes need to do when a mobile router of a 
    mobile router moves?

 > > > non-mobile hosts shouldn't be forced to become
 > > > mobile aware by virtue of a mobile router some
 > > > arbitrary number hops away.
 > > 
 >      => Does the draft say that ? Please let us know 
 >      where it does so we can remove that assumption.

   The draft doesn't make a differentiation
   between (a) and (b). I'm not convinced that
   (b) is very well understood, however, but
   that's probably not HMIP's fault (other than
   claiming to support mobile routers and accidentally
   stepping on a large land mine).

               Mike