Re: on the point of mobility & multihoming

[Eliot Lear <lear@cisco.com>]

I like the idea of having a "mobility considerations" section in 
peoples' drafts.  While I agree with John that if we focus on trying to 
kill two birds with one stone, sometimes the other bird just happens to 
get killed by being in the right place at the right time.  To stretch 
this analogy past the bounds of taste for a moment, if we sufficiently 
wound the second bird and it only requires a small amount of thinking to 
knock him off entirely, we're better off.

That having been said, Geoff's (and Dave's) clear message is that we 
could create a nightmare of overhead and complexity if we are not 
careful.  Consider a case where we are using some sort of shim layer 
solution that sits atop MIPv6.  The cases we have to deal with look like 
this:

--  what happens when a provider goes down?
Will the correspondent node attempt to send packets to the home agent?

-- what if the home agent is behind the same provider?  will the home 
agent be using a multihoming mechanism, and how will it impact the 
correspondent node?

-- suppose the mobile node reseats itself back to its home network in 
the midst of all of this?

Messy messy.  In the meantime, is what we would really have from an 
architectural standpoint one form of transport isolation running atop 
another (I hesitate to use the word "tunnels" because it's not quite 
correct, but we're in the same ball park).

BTW...

Excellent presentation, Geoff!

My apologies to bird lovers everywhere...

Eliot