In your previous mail you wrote: Is it not true, then, that the better multiplexing ratio disappears (while it is often presented as the main advantage of ISP's running the NATs instead of CPE's)? => IMHO this better multiplexing ratio is an illusion: it relies on a heterogeneous population but this doesn't match what one should get in the real world: a small number of end-users sharing the same address and likely living in the same block... I believe the same thing about the similar argument on the provisioning: stateful is supposed to make overbooking easier, i.e., you can put a hard limit L to the number of ports per user with L * <#user> >> <#ports>. With other words, between BSDs which pre-allocated the swap space when processes request it, and Linux which kills arbitrary a process when the swap space is full, you should guess what I prefer...
Regards francis.dup...@fdupont.fr PS: when applicable stateless solutions are better just because by definition they are scalable. Exactly the opposite for a CGN. _______________________________________________ Int-area mailing list Int-area@ietf.org https://www.ietf.org/mailman/listinfo/int-area