Re: [v6ops] AWS ipv6-only features

[[e-mail]]

Mark,

> > True, and I can't condone it, but as long as they don't leak it, the only
> > operator that can be damaged is AWS itself, so it's an own goal. In fact,
> > even if they do leak it, any competent ISP will drop it.
> >
> > The damage is not to operators, it is to application developers. Using fd00:ec2::/16 pretty much guarantees that there will be collisions within EC2 itself. If collisions can happen, that means that applications will need to learn to work with NAT66 or at least with NPTv6. That's pretty much the worst thing they could have done for IPv6 I think.
> 
> I agree that damage is bad. Unfortunately that cat is already out of the bag.
> IPv6 applications already need to work through NAT64.
> 
> And likely enterprises running on ULAs with NPTv6 gateways and "firewall in the cloud" style services which typically use NAT66/NPTv6 too.
> 
> 
> Here's an example I think of to demonstrate the point.
> 
> I've had the same mobile phone number since 1995, and anybody who knows it can still call me on it.
> 
> That's across multiple carriers due to number portability (and I'm quite aware of the scaling issue of doing that, however it seems to be working well enough).
> 
> Imagine not even knowing your own phone number. That's what NAT is doing. It makes things callers-only, even when being a receiver would be far better.

Absolutely. And my point wasn't to argue the qualitative properties of NAT.
I was making the point that NATs in various forms also are prevalent in IPv6 and that an application cannot assume end to end transparent addressing.
(Addresses are ephemeral and of course applications shouldn't use them as identifiers anyhow.)

O.