Hallam-Baker, Phillip wrote: > There are really two separate questions here: > > 1) Should application developers write protocols that make assumptions > as to the Source/Destination IP headers remaining constant end to end? no, that's not the question at all. the question is whether application developers should be able to assume that the address of a peer as viewed from one location, can be used to reach that peer from a different location? and the answer is clearly yes. but NATs break that. however, if NATs adhere to certain rules (e.g. they support hairpinning, and their bindings are not dependent on external peers' addresses, and they have a STUN-like facility built in, and internal apps can explicitly request bindings for the purpose of listening to incoming traffic) and certain disciplines are imposed on their deployment, then applications can obtain addresses which have these characteristics. which might or might not be adequate overall, but it's a lot easier and more reliable than what can be done with existing NATs in IPv4. > 2) Should NAT66 be prohibited in order to allow such assumptions to be made? no, that's not a well-formed question either - because it's not at all clear that IETF prohibiting NAT66 would have any significant beneficial effect. the question is, what can IETF do to (a) make NATs as sane as possible (b) discourage NATs as much as possible and (c) encourage such NATs as are used to be sane ones? Keith _______________________________________________ Ietf@xxxxxxxx https://www.ietf.org/mailman/listinfo/ietf