>>>>> "Christian" == Christian Vogt <christian.vogt@xxxxxxxxxxxx> writes:
Christian> Right. There is one limitation, though: With stateless
Christian> NAT'ing alone, failover of active communication
Christian> sessions between providers is not possible.
I agree with this statement.
Christian> This is
Christian> because statelessness requires one-to-one address
Christian> mappings, hence a separate internal prefix for every
Christian> provider-assigned external prefix. Many-to-one address
Christian> mapping, such as by mapping a single internal prefix
Christian> onto multiple external prefixes, would require stateful
Christian> demultiplexing.
I don't think this follows. Statelessness only requires that when a
packet crosses from inside to outside, I be able to select the correct
external prefix without state. There are a number of ways to do this,
including hashing the six-tuple (five tuple plus flow ID) to choose an
exit. The return direction does not require state.
None of this allows you to fail over a connection. However,
maintaining state does not help either. If you have multiple external
prefixes most transports will not permit you to change the external
address on an ongoing connection.
We have a lot of tools if you want multihoming better than that. Some
of them, like BGP multihoming, LISP and HIP, work quite nicely with
NAT66. Others, like SHIM6, would need some work to work in a NAT66
environment.
_______________________________________________
Ietf mailing list
Ietf@xxxxxxxx
https://www.ietf.org/mailman/listinfo/ietf