Google's IPv6 Statistics | Hacker News Reader

Google's IPv6 Statistics

google.com

53 pointsbensummers33 comments

That is depressing, isn't it?

On the other hand, I'm still wondering why Google hasn't switched away from the Flash-based graphing tool seen Analytics and Finance.

Has anyone had any issues with it? I think it is very robust and integrates so seamlessly one doesn't notice it's Flash unless you have Flash disabled. This is one of the 'appropriate and well implemented' Flash products. It's not broken in any way, so why fix it?

I know I've disabled flash myself; my browsing experience is more often hurt by flash than it is helped, so it is a net gain.

I think on the graphing tool its as simple as the need to support IE6, 7, and 8. There are decent Canvas wrappers for relatively static content on IE (excanvas etc) and an increasingly interesting SVG wrapper at http://code.google.com/p/svgweb/ but for more dynamic graphics with sliding and zooming I've never found one that's good enough for production.

There are times when I wonder if we will ever actually switch over to IPv6. The rate of adoption is super slow and yet at some point surely we'll hit some sort of critical mass right?

Hurricane Electric has a nice little widget that shows how many IPv4 addresses are left for allocation.

The site is http://ipv6.he.net/

I was reading somewhere that there are still a pretty decent number of allocated but unused addresses available out there that will still need to be used but that can't go on forever.

It's hard to say what will actually happen when we do start to run short, though. I doubt it will have any noticeable impact on the average consumer, though.

I find these examples of scarcity and cost really interesting and wonder if economists do too... IP4 is like a miniature version of the energy crisis — at some point the cost of not switching _to someone_ become greater than those of switching. I'd be interested to know how it works in this situation? Who will be, in actuality, faced with the headache of not enough addresses, and will they be sufficiently powerful to motivate others who don't directly experience the pain to transition?

I wonder why there's a spike every Monday?

Unless the graph changed, the spikes are showing up on Sundays on my end. My guess is simply from weekend use - people going home and using their ipv6 connected computers, then using their ipv4 work computers during the week.

Are there actually any residential ISPs that do native IPv6? I'd switch in a heart beat if there were...

Where do you live?

My ISP does[1]. It's not trivial to switch over though - for example, most home ADSL devices won't support it[2].

[1] http://ipv6.internode.on.net/

[2] http://ipv6.internode.on.net/access/adsl/

I should have mentioned. I'm in the UK. The East Midlands to be precise.

Comcast seems to be working hard on it: http://comcast6.net/

>any residential ISPs that do [native] IPv6

It seems there must be (though I guess not necessarily end-to-end), as jjcm notes there is a peak every Sunday and a local minimum nearly every Friday suggesting that the usage is primarily domestic. If this analysis is correct then this surprises me.

Here's the problem, I only ever hear about the advantages of IPv6 in terms of its much larger address space, and thus as a solution to the limited address space of IPv4. If the only reason to expend all the effort necessary to move away from IPv4 merely to avoid the problem of running out of addresses then of course it's going to be a slow process, and may not be started in earnest until there's really a crisis.

This article is worth reading: http://arstechnica.com/business/news/2010/09/there-is-no-pla...

It explains things like why having a IPv6 capable computer doesn't solve the problem, why we can't just use the class A address space etc etc.

Over 99.7% of hosts hitting Google.com are IPv4 only. That's insane! Some estimates say that IPv4 addresses will run out in under a year. And still under 0.3% of all hosts have IPv6 capability. Long live ISP-side NAT.

However, I think the reason for this is obvious. Had IPv6 not been designed as forcefully backwards-incompatible...

> Over 99.7% of hosts hitting Google.com are IPv4 only.

Is it the hosts, or is it the mid-span links ? I'm sitting here in a fast-food place with a Macbook is certainly IPv6 capable, but I suspect the cable-modem and/or the AP they use here isn't. How many last mile providers are IPv6 capable today ?

Your computer being IPv6 compatible is incidental to the actual transition to IPv6. IPv4 to IPv6 NAT will be used as the circuit that's provided has IPv6 only connectivity in order to support older clients. And then, in order to talk to any hosts that don't have IPv6 support (a large number of them, as in, insanely large number), you'll need an IPv6 to IPv4 bridge (provided by the ISP, transparent to the end user). If done properly, hosts that support IPv6 will start getting traffic as the onsite CPEs and gateways start getting IPv6 addresses rolled out to them (likely to be starting on date X new customers get IPv6 IPs and existing customers maintain IPv4 for the transition period).

Why is that so insane? Does it surprise you that more than 99.7% of cars out there are not hybrids? There is no pressing financial reason to switch yet. As a node on the Internet you can either request an IPv4 address for, say, $0.25/month and make that go as long as possible, or you could drop whatever other work you are doing and replace/upgrade, test and deploy your entire hardware architecture just so that at some point when IPv4 support starts getting dropped (not happening in one year), you can be ready.

Once again, everybody relax. IPv6 will happen. As soon as it is cheaper to upgrade your entire infrastructure than getting another batch of IPv4 addresses, everyone will jump on board. In the mean time, just breathe.

And lastly, two 16 byte addresses cannot possibly fit into a 20 byte header, so there is no way that IPv6 would have been compatible with IPv4.

It could have been done with software updates only, to routers and hosts.

The problem is, you can't just update IOS on your routers, you also have to add IPv6 configuration to the routers.

If two 32 bit addresses are talking to each other, use IPv4 packets. If at least one address is 128 bits, use IPv6 packets. It could have been that simple, with software updates (and hardware, in the case of many routers) over the past what, 15 years, we could have been running ipv6 effortlessly.

"If two 32 bit addresses are talking to each other, use IPv4 packets. If at least one address is 128 bits, use IPv6 packets. "

That, actually is a pretty good description of a dual-stack host. :-)

So what about all the organisations that were assigned class A address blocks? Should they not do something and give up these classes, come 2011? We certainly don't have 2^32 devices out there.

Because I hate downvoting without explanation: This question arises every time when exhaustion of IPv4 is discussed.

Reallocating class A blocks would only delay the inevitable by few months (less than a year for sure).

Are 50% of all IPv4 addresses not class A? Can you explain a bit more in-depth why it would not make a significant difference?

Talking about "class A" is a red herring. There are used addresses and unused addresses. It's not clear how many allocated-but-unused addresses exist because the audits necessary to find them would be expensive, but people have made estimates and those estimates are not that large. At any rate, the transfer market should flush out some of these "dark" addresses once they are needed.

50% of IPv4 addresses were in class A (0-127). But only 40 of the blocks were assigned to organizations in legacy assignments.

Organizations with unused addresses will start selling them soon. https://www.arin.net/resources/transfer_listing/

I wonder how they collect this data. Maybe by having a list of IPv6 capable ISPs

The typical way to do this is to include a 1x1 pixel image from an IPv6-only server, and use JS to detect whether it loads or not.

Can't we just make a backward compatible IPV6 bis, move on and forget this fiasco ?

No. Even if some new protocol was finalized today, it couldn't be widely deployed until 2015, which is after the end of the world. Also, there's no way to create a protocol which has larger addresses and is completely backwards-compatible; IPv6+NAT64+DNS64 already exists and is as backwards-compatible as you're going to get.

To the rate at ipv6 is installing we are still nowhere today, we have been there for more than 5 years, and we could still be there 5 years from now. I have read several times that it ipv6 had been designed non backward compatible, so I suppose there was a backward compatible way to design it. It seems that it would be possible to use 0 0 0 0 prefix as a special case for ipv6 and truncate it for ipv4.

I suppose there is a reason that prevent that because this is way too much of an evidence, so someone certainly has though about it : what's prevent us to do that ?

Reply on news.ycombinator.com