Iljitsch van Beijnum wrote:
...
But in any event, compared to the backflips through flaming hoops we
have to do in IPv4, the asking a remote server what our source address
looks like from the outside to make address based referrals work
doesn't seem too onerous. Or do you disagree?
Who do you ask??? Your note assumes there is only one 'outside', so any
server could answer the question. There is absolutely no restriction on
where and how topology warts are deployed, so asking a server in network A
what your address will appear to be to network B is fundamentally absurd. I
have heard similar comments from the document authors recognizing this
problem, but hand-waving something about asking a service before populating
DNS, while completely ignoring the fact that there is no way to predict in
advance who will want to know or where they will be attached. Essentially a
server is not reachable until it guesses that network B exists, someone
wants to contact it from there, and where the service is to ask about the
address that the server appears to be.
There is no valid reason for 66nat. The only justifications being given are
'people will do it anyway', and 'we have to move quickly because vendors are
trying to build it'. This is called railroading in any other context, and
absolutely no long term thought is going into the impact and inability to
remove this once it is unleashed.
Tony
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/ietf