March 2008 Archives

Apple's Airport Express has long been a popular, cheap wireless access point. They're popular with several of my colleagues for business trips -- they're a nice way to get wireless in your hotel room when you travel.

This morning, Apple rev'ed the Airport Express. They added draft 802.11n support. They also added 6to4 tunneling support. This feature has been around in Apple's higher-end Airport Extreme for over a year, but the cheaper Express has lacked it. With today's announcement, Apple supports IPv6 tunneling on all of their wireless access points (the Airport Express, Airport Extreme and Time Capsule).

To the best of my knowledge, Apple is the only consumer electronics vendor who can say this. Most other vendors don't support IPv6 at all; a few support it on one or two of their products. Way to go Apple!

CAIDA has published their 2008 graph for IPv6 AS topology:

Not surprisingly, the graph shows that Europe has better IPv6 interconnectivity than the Asia/Pacific region or North America.

| | Comments (9) | TrackBacks (0)

Google is reachable via IPv6 :

So far, it's only their search engine, not any of their apps. And none of their nameservers are reachable via IPv6 yet. But it's a start.

I keep hearing that IPv6 will only be "successful" when the major web sites (google, yahoo, facebook, ebay, paypal, etc) start using it. This is a big step in the right direction.

I should point out that latency over IPv6 is terrible (from AS3999): 20.6 ms over IPv4 -vs- 251.3 ms over IPv6. It looks like the Internet2 → Global Crossing → Google routing is still very sub-optimal:

$ traceroute6 -n
traceroute6 to (2001:4860:0:1001::68) from 2610:8:6800:1::408, 30 hops max, 12 byte packets
 1  2610:8:6800:1::1  39.817 ms  10.278 ms  10.136 ms
 2  2001:5e8::fffe:0:1:4:102  10.058 ms  10.055 ms  10.049 ms
 3  2001:5e8::fffe:0:1:4:101  10.048 ms  10.045 ms  9.395 ms
 4  2001:5e8:0:1::1:fd  10.704 ms *  6.579 ms
 5  2001:5e8::fffd:0:2:2:2  14.951 ms  18.307 ms  12.993 ms
 6  2001:468:ff:109::1  30.25 ms  26.884 ms  36.155 ms
 7  2001:468:ff:103::1  48.942 ms  51.061 ms  48.879 ms
 8  2001:468:ff:304::2  81.231 ms  85.703 ms  87.645 ms
 9  3ffe:80a::c  98.942 ms  90.235 ms  98.992 ms
10  2001:450:2001:1000::670:1708:179  335.792 ms  168.063 ms  174.705 ms
11  2001:7f8:1::a501:5169:1  251.204 ms  248.744 ms  250.161 ms
12  * * *
13  2001:4860:0:1001::68  246.633 ms  249.771 ms  248.692 ms
(As an aside, why is there a 6bone address (3ffe:80a::c) in there? The 6bone address space was deprecated years ago.)

If I had to guess, I'd say that the IPv6 servers are in Europe, since 2001:7f8:1::/48 is assigned to the AMS-IX Exchange Point.

This morning I read an essay claiming that:

It's time to start talking about what the Internet will be like in a future where we abandon all our efforts toward the IPv6 transition. Because the transition isn't happening. It's not going to happen. We're going to be living on IPv4/NAT for the rest of our lives.

I think the author actually ends up arguing in favor of IPv6, even though he doesn't realize it.

Go give it a read, then come back here for my take.

He correctly point out that the IPv4 routing table is continuing to grow in size, and that there are issues with multi-layer NAT. To me, that's an argument against the long-term viability of our fragmented IPv4 system and of NAT. He claims that "the costs and benefits of just limping along forever with an IPv4/NAT-only architecture are predictable and well-understood," but then argues against this claim by saying that we don't understand the scalability limits of NAT, by pointing out that IP-in-IP tunneling has performance issues, and by saying that maintaining IPv4 will lead to uncertain cost increases. That doesn't sound like "predictable and well-understood" costs to me.

He correctly point out that an IPv4 market that's heavily regulated by the UN and WTO will be... shall I say, "painful" to work with. Again, that to me is a reason to adopt IPv6 sooner rather than later.

He correctly point out that the cost of routing IPv4 is increasing, and that it could increase sharply in the near future:

The monthly bill from your ISP is about to have an uncertain new fraction of it devoted to the cost of maintaining that globally routed IPv4 address you're using (and possibly sharing with your neighbors). How much? Nobody knows.... It could be pennies a year. It could be the better part of a hundred dollars a month. Nobody knows. Nobody freaking knows.
Isn't this uncertainty a reason to adopt IPv6 sooner rather than later?

I disagree with the claim that the increased cost of maintaining IPv4+NAT won't drive people to IPv6. There are already some ISPs who are actively promoting IPv6 precisely because of the economic uncertainty of IPv4. And these ISPs are having some success.

IPv6 allocations have been steadily increasing in every RIR since 2000. See this graph from the RIPE NCC 2006 annual report.

There certainly are issues with IPv6. A fair amount of software doesn't support it (but a fair amount already does), DHCPv6 support is still buggy on several platforms, and many ISPs still don't offer IPv6. But these issues are being addressed.

IPv6 is here. Its deployment is growing. It's just happening very slowly and very much under the radar.

ARIN and CAIDA are conducting a survey of IPv6 deployment in the ARIN region. If your organization has deployed IPv6 or plans to do so soon, please fill it out. The survey closes on March 24.
6to4 is a popular IPv4-to-IPv6 transition mechanism. It works by tunneling IPv6 over IPv4. It's useful for connecting an IPv6 "island" to the IPv6 internet if you don't have IPv6 connectivity. For example, the Apple Airport Extreme supports 6to4: You can use IPv6 in your home even if your ISP only gives you IPv4.

Due to the distributed nature of 6to4, no one has quite figured out how to make reverse DNS work. Until now.

Yesterday, the IETF approved RFC 5158, which specifies how reverse DNS should be handled for 6to4 islands. As RFCs go, it's pretty readable. In short, you go to from a client inside your 6to4 island and register your DNS servers.

I run 6to4 at home, so I'm interested in this. I just need to scrounge up two machines to use as servers.
In January, 2008, Google held an IPv6 Conference. The videos were just uploaded to YouTube yesterday:

Panel Discussion: What will the IPv6 Internet look like (54:44)

IPv6 on Windows (22:19)

Planning for the IPv6 Integration (20:19)

IPv6 and the DNS (18:11)

IPv6, Nokia and Google (23:53)

Some of the videos are pretty good; others are fairly repetitive. I found the Windows and DNS ones good.
Just a quick note: I just noticed that has IPv6-enabled their web, DNS and SMTP servers. It's great to see that a very public project has stepped up to the plate.