Re: [v6ops] IANA assigned prefix for IPv6 benchmarking

[Tore Anderson <[e-mail]>]

* Erik Kline

> Regardless of ULA or some other space, I think a useful property for
> the GUA(-like) prefix used by Apple's NAT64 test network would be
> randomness.
> 
> If, for instance, *every time a NAT64 test network was started* a new
> random ULA was generated or, say, 2001:2:0:$RAND16::/64 were used,
> then developers would be less likely to accidentally make assumptions
> about the consistency of addresses.

Same thing goes for NAT64/DNS64 translation prefix. This could be
tempting for developers to hard-code as well - my gut feeling tells me
that most NAT64 deployments in 3GPP networks out are using the WKP from
RFC6052 (64:ff9b::/96) for their translation prefix.

As I also mentioned in the WG session I think that Apple should simply
obtain a GUA prefix through the registry system for this purpose (never
to be advertised into the DFZ). That way they wouldn't have to squat on
prefixes reserved for other purposes (like 2001:2::/48), and they would
be able to document what the prefix is actually used for and link to
further information in the prefix' whois entry, so people who
accidentally find themselves connected to such a network has a chance
to figure out what's going on.

The cost would be negligible - a /48 PI costs only €50/year here.

Tore