65

According to my IPv6 tunnelbroker Hurricane Electric the world will end in approximately 712 days when we finally run out of IPv4 addresses.

Would it be possible to enable IPv6 access to the Stack Exchange family of websites?

3
  • 8
    The counter just dropped from 1 day to zero. And indeed, the last pools have been assigned to regional offices, and some "significant announcement" is due tomorrow...
    – Arjan
    Feb 2, 2011 at 20:34
  • I know this is (or is it?) [status-deferred]. Still a bounty, if only to get some attention for World IPv6 Day, Wednesday June 8th.
    – Arjan
    Jun 4, 2011 at 11:27
  • "According to my IPv6 tunnelbroker Hurricane Electric the world will end in approximately 712 days when we finally run out of IPv4 addresses." reading this +4000 days later :D Mar 17, 2021 at 20:16

2 Answers 2

37
+100

I've done some checking and www.stackoverflow.com is currently hosted on an IPv4 address that belongs to peer1, so this must be PA space.

Looking on ARIN's whois, Peer1 (PER1) have lots of IPv6 space:

PEER1-IPV6-01 (NET6-2001-1978-1)    2001:1978:: - 2001:1978:FFFF:FFFF:FFFF:FFFF:FFFF:FFFF
PEER1IX-IPV6-01 (NET6-2001-504-19-1)    2001:504:19:: - 2001:504:19:FFFF:FFFF:FFFF:FFFF:FFFF

So you can get an allocation from the 2001:1978:: space.

I've had a look at Peer1's IPv6 routing and it seems to be pretty much first-class, so there's no ISP excuses for not having IPv6.

I know that stackoverflow have private BGP (from the SF blog) so they should be able to just get an IPv6 allocation, and announce their routes through BGP.

The web servers (being Windows 2008) support IPv6 straight out of the box, so the constraints I can see are:

  1. Do they store / use IPv4 addresses anywhere, e.g. in spam-blocking? If so, they have a development job there to incorporate IPv6 addressing. Not that huge, but still some work.
  2. What about their load-balancers, routers, etc? - HAProxy and Quagga both support IPv6 in recent versions, and if they have switched from Quagga to Cisco hardware routers, then all recent versions of IOS support it too.

Conclusion: StackOverflow could operate on native IPv6 without too much of a hassle, and it would be nice to do so and will get steadily more and more important.

How about World IPv6 day (8 June 2011) as a target for IPv6 native throughout?

16
  • 2
    Spam and low quality questions blocking could actually improve, as often IPv6 addresses give one both some subscriber-specific prefix (kind of like the IPv4 address, which could be shared with others) and the computer specific MAC address. Blocking both makes it harder to bypass the block.
    – Arjan
    Feb 7, 2011 at 18:03
  • Richard, if you happen to know a lot about IPv6, and the IPSEC that all implementations must support, then: do you know if that makes using HTTPS obsolete (well, for the encryption part)? I'm wondering if supporting IPv6 would also solve issues with cookie theft, like that of Firesheep.
    – Arjan
    Feb 7, 2011 at 18:13
  • 3
    @Arjan No, IPv6 doesn't make HTTPS obsolete - it's quite hard to require a connection to us IPSec from the client level (it's a network level feature). TLS sits up at the application layer so it's easy for the application to require TLS: IPSec is way down in the land of datagrams. Feb 7, 2011 at 18:44
  • @Arjan - yes, IPv6 spam / LQQ blocking might work better - though bear in mind that most implementations don't actually use the MAC address for blocking; you're probably better off blocking a /64 - but it will definitely need rewriting to deal with a different set of issues. Feb 7, 2011 at 18:46
  • (Okay, thanks, @Richard. I copied your comment to the Firesheep question.)
    – Arjan
    Feb 7, 2011 at 18:48
  • 5
    Have just noticed that Facebook's IPv6 address is 2620:0:1CFE:FACE:B00C::3 which is rather neat. [nslookup -type=AAAA www.v6.facebook.com for those on Windows, dig www.v6.facebook.com AAAA for UNIX-types]. Anyone 7EE7 enough to do the same in hex for StackOverflow? Feb 8, 2011 at 12:26
  • 1
    +1 for the World IPv6 day suggestion!!! Feb 9, 2011 at 0:26
  • (To extend on my earlier comment a bit: I just learned that exposing MAC addresses in IPv6 addresses can be avoided by a setting on a computer (it's not something a router configures). See How to avoid exposing my MAC address when using IPv6? on Super User.)
    – Arjan
    Feb 9, 2011 at 19:01
  • @Arjan that settng is the default in Windows 7, which is likely to be a really large fraction of early IPv6 usage. Feb 14, 2011 at 21:16
  • I suddenly think IPv6 blocking is harder than IPv4: how would one know how large the address space is that an ISP has given a subscriber?
    – Arjan
    Apr 12, 2011 at 8:20
  • 1
    @Arjan You don't know but you can guess relatively sensibly: a /64 and if you are blocking a lot (how many a "lot" is will have to be determined empirically) of /64s in a /56 then block the whole /56, the same for a /48. Nothing higher, as can be determined by examining the IPv6 allocation policies of the RIRs. Apr 12, 2011 at 11:16
  • @Richard, my consumer grade subscription has a /48 block. It seems that matches an Arin proposal: All customers get one /48 unless they can show that they need more than 65k subnets (though it also states If you have lots of consumer customers you may want to assign /56s to private residence sites). But still then, indeed: SE would get few requests from my /48 (and only to some SE sites), so SE might guess right.
    – Arjan
    Apr 12, 2011 at 11:41
  • @Arjan yes - I expect that a lot of ISPs will ignore that and use /64s for private residences. The interesting one is going to be mobile phones. Will they get allocated a /64 (for tethering) or a /128? If they get /128s then they're going to be practically unblockable by IP address. Apr 12, 2011 at 11:44
  • 1
    Good data! But Peer1 just finally blogged about getting their own web site an ipv6 address for ipv6 day, and many blog posts suggest they don't offer ipv6 to customers yet, at least not to small ones. E.g. Peer1 not doing AAAA records yet. I wouldn't think SE would want to operate over a tunnel, though in the short term that might be ok. But hopefully this all won't take too much longer.....
    – nealmcb
    Jun 9, 2011 at 16:39
  • 1
    @RichardGadsden how about 57AC:07E8:F703 or 57AC:07E8:F103 or even 57AC:71:07E8:F703 -- K & R is tricky!
    – Azendale
    Nov 7, 2013 at 19:41
8

Are you suggesting to start by spoiling the IPv6 space with sites that already have an IP address assigned?

We'll run out of IPv6 addresses in the year 53985251 AD, and people living by then are going to be seriously pissed off when they find out about this.

4
  • 4
    I think that year will be sooner when everyone's toaster is online.
    – Troggy
    Sep 13, 2009 at 15:42
  • 14
    The toasters are already online. You're only getting burnt toast because there's a buffer underrun leaving the charring as a result. Bad packet delivery in action.
    – random
    Sep 13, 2009 at 15:50
  • 5
    I read that as "butter underrun". Jul 7, 2014 at 10:07
  • I'm suggesting to start spoiling the IPv6 space with sites that can't distinguish one subscriber to an ISP from the 1,000 other subscribers to the same ISP on the same NATted IPv4 address. Jan 5, 2020 at 20:28

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .