Re: [v6ops] draft-ietf-v6ops-ula-usage-recommendations - work or abandon?

[Joe Touch <[e-mail]>]

On 11/8/2015 6:04 PM, Lorenzo Colitti wrote:
> The main reason you wouldn't design it like that today is that NAT has
> broken the original architecture of IPv4.

NATs actually provide a very sensible model, just not the one most
designers think.

>From the private side, a NAT acts like a router.

>From the public side, a NAT is (not just "acts like") the endpoint.

So it's not surprising that running an application that isn't on the NAT
box causes problems. The NAT is the problem - it wants to act like an
endpoint (sourcing/sinking packets from/to a public IP address) but
doesn't want (or have enough information) to run the app correctly.

The IP achitecture is fine, and NATs can be a valid part of that
architecture, but only to the extent of what they are. They ARE hosts on
the public side. E.g., we would expect that it's hard to run 6 web
servers on port 80 on a single host with a single IP address; why is it
so surpising that a NAT can't support that?

Joe