IPv6 Challenges and Opportunities 315
1sockchuck writes "Opinions differ on when the Internet will run out of IPv4 addresses, prompting a wholesale transition to IPv6. In recent videos, John Curran of ARIN provides an overview of issues involved in the IPv6 transition, while Martin Levy of Hurricane Electric discusses his company's view that early-mover status on IPv6 readiness can be a competitive advantage for service providers. Levy's company has published an IPv4 DeathWatch app for the iPhone to raise awareness of the transition."
Re:Current deadline, in case anyone's interested (Score:2, Interesting)
Does that take into account universities and large companies giving back all the class A ip addresses they have that were initially given out back in the day?
(I'm genuinely asking, I don't know)
All I care about (Score:1, Interesting)
Re:IpV6 reality check (Score:3, Interesting)
Since this rant, google has actually gone IPv6 for IPv6 ready ISPs.
http://www.google.com/intl/en/ipv6/ [google.com]
By no means is the internet IPv6 friendly, and a lot of the points Dan makes are good ones, but he fails to offer any solutions either.
The thing that gets me... (Score:4, Interesting)
...is that even new devices don't support IPv6, even when they're in entirely controlled address spaces.
For example, why the hell don't, for example, cell phones internet capabilities have IPv6? I mean the IPv6 routing would seem exactly designed for cell phones, devices external to the network don't need to reach them, and it's a frickin closed system with device upgrades fairly quickly. If we can't even use IPv6 in closed systems like that, it has failed.
The reason, of course, is because IPv6 is, in fact, an EPIC FAIL in actually working, because no one apparently bothered to figure out any sort of actual transition for it.
It's like, if instead of self-driving cars, they invented self-driving micro-monorails and expected us to buy them. But, don't worry, they have a handy monorail carrying rack we can install on top of our car that not that hard to set up so we can carry our monorail to the monorail tracks fifty miles away.
D. J. Bernstein is an ass, but he's right about this.
IPv6 should have been built by changing the damn format of the packets, but using the exact same IPv4 addresses with a specific prefix, routed exactly the same place. Any router that talked to devices that didn't understand IPv6 could just 'dumb it down' to IPv4, and, they should eventually do the same in reverse!
We could actually include a bit in the packet that upconverted IPv6 packets get, so we could keep statistics on how many packets were IPv6 their entire distance, and how many got converted down and back up at some point. So we could see what networks are actually switching out their equipment, and see what misconfigured gear thinks it's talking to IPv4 devices when it's talking to IPv6, so it needlessly converting. (IEEE 802.2 specifics a way to autonegotiate IPv4 or IPv6 using the EtherType, but it might not always work, and it's only for Ethernet anyway.)
At some point, as routers and OSes got replaced, large amounts of traffic on the internet would end up being IPv6 their entire distance, and at that point we can start assigning the IPv6 addresses that don't have a equivalent IPv4 one.
And, incidentally, we should keep the IPv4 network operational forever. 95% of the people can give their IPv4 addresses back, and as people stop connecting IPv4 devices, routers and whatnot will lose the ability to speak to them but there will still be some devices that cannot be upgraded, some embedded device that speaks only IPv4 or whatever. The company should be able to keep an IPv4 address, and require people to install one of the routers that can still upconvert in front of the device, and it gets routed over the internet and back just like anything else, because, for almost all the trip, it's IPv6. There would be no reason to ever turn off the subset of IPv6 that is IPv4.
Instead we invented a new fucking network that doesn't interact with IPv4 at all. Yes, yes, you can get IPv6 versions of IPv4 addresses, but routers and OSes do not automatically translate them. And it's actually against the rules for someone to try to contact a IPv4 server 'over' IPv6. They have to use their IPv4 address, like there should be a difference.
Re:marketing speak = teh suck (Score:4, Interesting)
Re:marketing speak = teh suck (Score:4, Interesting)
He doesn't need to. He may want to. He has that option today.
You can assign IPv6 addresses manually to your heart's content as long as you have a block assigned to you, but for client machines there is rarely a reason to do this (just like how you normally don't go about handing out static IPs to every workstation, you set up a DHCP server (or many depending on the size of your organisation) and hand out dynamic addresses to most machines).
/Mikael
Previous address expansions (Score:4, Interesting)
IPv6 should have been built by changing the damn format of the packets, but using the exact same IPv4 addresses with a specific prefix, routed exactly the same place.
Yes, that's what was done the last two times the address space was upgraded.
When ARPANET IMP addresses went from one byte to two bytes, to allow the number of nodes to increase beyond 256, the old addresses retained their 8-bit value, with a new prefix.
When the ARPANET was extended to the Internet, the two byte IMP address was the low two octets of the IP address, and the first two octets were 10 and 0, so IMP addresses converted to IP addresses as [10.0.xxx.xxx]. And that's where "network 10" came from. When the ARPANET went down, it freed up that address space for other uses.
But we have DNS now.
Re:marketing speak = teh suck (Score:4, Interesting)
No it doesn't, at best there are 4,294,967,296 available IPv4 addresses, in reality there aren't nearly as many since the entire network isn't one huge subnet. With IPv6 there are 3.4*10^38 addresses. There is no real competition in terms of "we give you your own class C" vs "We give you one address" when it comes to IPv4 because most ISPs can't actually hand out addresses like they're candy. With IPv6 an ISP would have no problem whatsoever handing out a /64 to each customer since they'll have a shitload of /64s to hand out.
/Mikael
Re:Current deadline, in case anyone's interested (Score:3, Interesting)
Problem with that prediction is it's bollocks.
The potaroo exhaustion counter that these dates come from hasn't changed significantly in the last year that I've been following it. It dips somethimes to 650 days or so, then climbs over 1000 days sometimes.. but the average stays around the 700 mark.
If the prediction had been remotely accurate when it said 700 last year it should be at around 350 this year, and it just isn't.
ipv6experiment.com (Score:1, Interesting)
Could somebody please tell me whatever happened to the ipv6experiment.com ?
Re:Readiness test checklist (Score:4, Interesting)
I can do everything in your list just fine -- if you pick a web site that actually supports v6. There is no AAAA record for slashdot.org. Were Slashdot to configure their servers for IPv6 and add appropriate AAAA records, then it would be reachable. Of course, most sites don't bother with v6 support, because few users have v6 addresses, which is because most ISPs don't support v6.
When the ISPs move, everything else will follow.
Re:marketing speak = teh suck (Score:4, Interesting)
I haven't counted, but I think I have *at least* 6 ISPs to choose from --- Not counting wireless, of course. North of Copenhagen, nothing special. And none of them seems to be able to deliver an unblocked port 25 (just inbound would be cool, I can relayed outgoing no problem). Sad, right?
Re:marketing speak = teh suck (Score:4, Interesting)
He doesn't need to. He may want to. He has that option today.
You can assign IPv6 addresses manually to your heart's content as long as you have a block assigned to you, but for client machines there is rarely a reason to do this (just like how you normally don't go about handing out static IPs to every workstation, you set up a DHCP server (or many depending on the size of your organisation) and hand out dynamic addresses to most machines).
/Mikael
I never do. I set up DHCP with static addresses for the known computers, and dynamic for the guests. So much easier to ssh between machines with proper ip addresses and names.
Please fix IPv6 (Score:1, Interesting)
Dear inventors of IPv6 please fix your protocol. I'm not asking for wholesale changes such as use of reasonable address lengths or a politically sane and realisitic transition scenario.. All I'm asking for is that some minor details be fixed OR ELSE!!!!
Not following law of demeter on ISO layer separation is a sin punishable by rotting in hell for eternity.
Using interface names to disambiguate IPv6 addresses..really? All %eth0 nonsense needs to be outlawed immediatly and forever. I don't know or care how many bis RFCs need to be created to make that happen but it must be done.
All hosts must be able to talk to themselves. If my servers IP address is AAA:BBBB:CCCC::BADD that address must work from the server itself without interface name decorations applied.
These above two demands are non-negotiable. The rest are nits I really don't expect anyone to take seriously:
The socket options to prevent use of IPv4 mapped addresses don't need to exist if dualstack is the accepted global transition strategy which it looks like will definately be the case. Make the IPv4 mapped space reserved like you know -- your precious massive Class E IPv4 block you just pissed away while the Internet is hurting for Internet addresses.
People should be assigned network prefixes that can remember using :: compression to its highest effect. The flawed notion that address will ever denote structure in any meaningful way needs to be dispensed with.
Re:marketing speak = teh suck (Score:3, Interesting)
I never do. I set up DHCP with static addresses for the known computers, and dynamic for the guests. So much easier to ssh between machines with proper ip addresses and names.
And... what's stopping you from doing that with ipv6?