[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: IPv6<->IPv4 address translation is not NAT?

On 2007-12-10 06:53, Geoff Huston wrote:

Brian E Carpenter wrote:


I think an important point of commonality is that when translating
between a smaller and a larger address space, you're forced to
deal with port mapping as well as with address translation.


not "forced", surely. "Might elct to" is close to the truth.


I think it's forced iff you are seeking a purely stateless
solution, which I realise from your explanation below is
not what you're assuming.


And the algorithmic embedding of A in B applies asymmetrically.


Correct. (Shanti is an example of that.)




That is the same problem whether the smaller/larger pair
is single-IPv4/RFC1918 or IPv4/IPv6.

Maybe we should talk about address sharing rather than address
translation?


I don't agree.


Well, isn't the goal to allow entities in a smaller address
space to access entities in a larger address space? I think
that necessarily requires sharing (or multiplexing if you
prefer).



What is going on appears to be the "type" of translation.


Yes, I agree with the distinction you make below - it just
isn't quite what I was writing about.

     Brian
One approach preserves stateless capability in the network elements and all that is a consequence avoids the use of placing translation sate i the network elements, and "embeds" the translation into the packet so that a remote entity (or the same entity at a later time) can reverse the translation and recover the orginal intended context.
Another approach attempts to place state into network elements and then all the packets within a context are treated in the same way. NAT bindings are a good example of this state.
Analogy of the differences between approaches: Imagine that you are flying home from Vancouver and you have no short term memory (it happens :-)). What do you do when you get off the flight in Canberra and you have absolutely no memory that you left your car at the Airport in Row 13? But you knew when you parked your car a week ago that you would forget. Do you write down where you parked it and keep this mesage with you through your trip so that its at hand upon your return (option a). Or pin the message to the airport information board and hope that it will still be there upon your return (option b).
The algorithmic embedding is one of these "packet-state" approaches where any network element could undo the operation given that the packet contains sufficient information to reverse the operation.
That's a long way of saying that I don't agree that embedding is sharing! Embedding is another approach to translation :-) 


  Geoff