> I believe the objection is against "non-deterministic translation", rather than > stateful versus stateless. By non-deterministic, I mean that the subscriber's > equipment (e.g., CPE) cannot determine the mapping it will have on the > Internet. A+P mechanisms are deterministic (including 4rd, Dual-IVI, and > draft-ymbk-aplus-p). > > A stateful CGN, as commonly deployed, is not deterministic. I don't understand why that is significant enough factor for IETF to (not) recommend some double translation variants. I mean does existing applications work better if double translation is done in deterministic manner? One reasoning against double translation has been that it breaks some class of applications. Is it now so that some forms of double translation do not break applications while some others do? Teemu
<<attachment: smime.p7s>>
_______________________________________________ Ietf mailing list Ietf@xxxxxxxx https://www.ietf.org/mailman/listinfo/ietf