RE: geo short vs long term? [Re: Geo pros and cons]

["Tony Li" <Tony.Li@procket.com>]

|    So how would intra-site routing work? By clearing the 45 
|    routing bits 
|    when the packet arrives at the edge?


That would be one way.

Note that you can also salt the routing bits (however many we choose)
with any other constant and that will work too.  This can be used to
make it easier to migrate.  Suppose that you have a site that is
initially singly homed and already has a fixed address.  By simply
swapping the routing bits at the border router, you can now renumber
the entire site trivially.  You can also multihome, again by only
tweaking the border routers.  All local configuration can remain
the same.

    
|    > |    I guess it would be possible to do this without changing
|    > |    the network layer but I wouldn't bet on it.
|    
|    > I don't believe that it's a requirement to change the syntax
|    > of the network layer header.  The semantics, of course, must
|    > change.  IMHO, this is a good thing.
|    
|    Why is this good?


Because today's semantics are "locator === identifier"

  
|    In earlier discussions, I think we (that is, at least some of us) 
|    reached the conclusion that a 16+16 mechanism, where we 
|    have something 
|    that looks enough like a regular IPv6 unicast address to 
|    be forwarded 
|    by existing routers, would be similar in nature, but 
|    easier to deploy 
|    than 8+8 and presumably also 5.625+6.375.


Yes, but given the above argument about swapping all of the routing
bits, I'm not sure I agree.  It seems to me that a border router that
swaps just routing bits or swaps all 128 bits isn't that big a
difference.

Tony