Re: [v6ops] Operational Consensus on deployment

[Lorenzo Colitti <[e-mail]>]

https://www.nanog.org/sites/default/files/wednesday_general_byrne_breakingfree_11.pdf

Slide 4: "Over 50% of IPv6-user traffic is end-to-end IPv6 (no translation
needed)"

The 5% number is just a guess. If you only look at the handset, then t's
just traffic from IPv4-only apps, and at least in the Android case, it's
quite hard to write an IPv4-only app - you have to really go out of your
way to do it in Java, or use native C/C++ code via JNI. The Netflix app
used to be that way, and now I think the major one is Pandora. It sort of
depends on the market though - from what I've seen on T-Mobile's deployment
it didn't take that long for the major apps to get IPv6 support (with the
exception of Skype and maybe Google Hangouts).

Tethering is still IPv4-only on T-Mobile (and on many devices), and that's
probably more than 5%.

Cameron, do you have a better estimate of how much traffic uses 464xlat?


On Fri, Aug 8, 2014 at 9:11 PM, Eric Vyncke (evyncke) <[e-mail]>
wrote:

>  Lorenzo
>
>  For my education, do you have a pointer for a data point on:
>
>   From: Lorenzo Colitti <[e-mail]>
>
>    A 4G handset with 464xlat will have ~50% of traffic native IPv6, ~45%
> NAT64, and ~5% 464xlat. 464 conversion is lossy and brittle, but if it's
> only used for 5% of traffic, then the operator might just say, "Who cares?
> I don't; and if somebody else does, they're free to use IPv6."
>
>