[ghshephard]

I'd be very interested in knowing how these stats were calculated - 12% of the United States on IPv6 seems a bit high. Maybe what Google is saying is that "It's Available, but we're not saying people are using it." - For example, Comcast has had IPv6 widely deployed for at least a year, so most of their customer might be identified as "Available" - even if their browsers aren't doing a AAAA lookup for www.google.com.

[tw04]

Not in the least. For all their atrocities, Comcast was one of the early IPv6 adopters and has rolled it out to their entire residential customer base (people with older modems/routers that don't support IPv6 obviously won't pull an IPv6 IP address). That's EASILY 11% of US internet traffic.

[furyg3]

Just curious, do the internal interfaces on Comcast's IPv6 routers give out V4 or V6 addresses?

[ghshephard]

I think that's the point I was trying to make.

[DangerousPie]

The page says "The graph shows the percentage of users that access Google over IPv6.", so I would assume this would mean people are actually using it.

[Spooky23]

This reflects the shift of traffic to mobile.

Everyone using Verizon Wireless with an IPv6 capable device is using IPv6. It's behind some sort of carrier grade proxy, but is IPv6.

[ay]

Also, if you are using T-mobile, you can use IPv6-only, and they have a few million subscribers.

I got a T-mobile SIM when I was in the US last month, and switched to IPv6-only APN. Worked pretty well for the casual browsing - and the non-IPv6 websites were still reachable using NAT64 on T-mobile's side. Though the possibility to do IPv6-only might depend on your handset.

[chime]

Indeed. You can test it on Verizon by doing a Google search for: ip

It should give you your mobile's IPv6 address.

[darren_]

> I'd be very interested in knowing how these stats were calculated

Reasonably sure it's based off something similar to this: https://www.ietf.org/proceedings/73/slides/v6ops-4.pdf

And yes, it is basically 12% of US users are capable of connecting to google over ipv6, not that they necessarily did/have. A browser might, for a URL that resolves to both v4 and v6 addresses, always try the v4 address first; this'd give you users that _can_ connect over v6 but don't. FWIW I have a v6-capable connection and find on chrome I pretty much always connect to google and facebook over ipv6, but I have no idea how IE/FF/Safari behave.

[ay]

The presentation is from... long time ago :-)

Now they serve to everyone over IPv4 and IPv6 and these stats are the actual IPv4 requests vs. IPv6 requests that they get, not some projected number.

On address selection:

Safari uses Apple API which has a proprietary mechanism for determining whether IPv4 or IPv6 gives a better user experience. You can approximate this as an RTT race between IPv4 and IPv6 connection (thus, somewhat of a flavour of what we describe in RFC6555).

Firefox/Chrome give a small headstart (~150-300ms) to IPv6.

IE strongly prefers IPv6, except if a particular MS-hosted IPv6 site is not reachable, then it strongly prefers IPv4.

[0x006A]

They should all use the "Happy Eyeballs" algorithm (https://tools.ietf.org/html/rfc6555)

[coldpie]

I think mobile ISPs are leading the charge here. I know my phone on T-Mobile is using IPv6.

[takeda]

Everyone who uses IPv6 also uses IPv4, but IPv6 is used whenever site supports it, and when they have to fallback they most likely connect through CGNAT).

For example everyone on Verizon who uses 4G uses IPv6 by default and CGNAT for sites that are IPv4.

When IPv6 is implemented correctly, the user doesn't even know when he/she is using IPv6.