Slashdot Mirror
IPv6 Traffic Volumes Are Low, But Nobody Knows How Low
netbuzz writes "As the June 8 World IPv6 Day experiment draws near, there is universal agreement that little IPv6 traffic is traversing the Internet at the moment. The event is designed in part to increase that volume. However, it will be difficult for Internet policymakers, engineers and the user community at large to tell how the upgrade to IPv6 is progressing because no one has accurate or comprehensive statistics about how much Internet traffic is IPv6 versus IPv4."
And in case you don't know much about IPv6 and why it matters, dave.io has kindly provided "a primer on the IPv6 transition: why it's cool, how to get started with it and what's changed."
Since the ISP:s don't want to offer IPv6 to their customers the traffic is a lot lower than it could have been.
Right now it's necessary to do tunneling to an access point for IPv6 and that's not convenient for the majority of the internet users.
If builders built buildings the way programmers wrote programs, then the first woodpecker would destroy civilization.
A timely article - I just got full native IPv6 running for my home internet connection last week (dual stack, of course).
Works well - the DSL modem connects like usual and the ISP assigns you a dynamic IPv6 /64 for the PPP session (ie. the modem's public IPv6 address), a static /60 for your LAN (your router then dishes out IPs within this subnet to the machines on the network via prefix delegation), and of course your good old standard single IPv4 address.
My Linux, Win 7, Mac OSX machines, iPad and iPhone all had no issue correctly picking up their IPv6 address and using it. The only things on the home network that are still IPv4 only are my old D-link NAS and the Wii. Attempting to access something, IPv6 is tried first, and it that fails it'll fall back to IPv4. Most Google sites are IPv6 enabled it seems, though other than that, the vast majority of stuff I access is still IPv4 only at this stage.
It really is weird having every machine in the house with a unique, globally addressable IP again after all these years behind a single public address using NAT. No more port forwarding.
Consumer Routers don't oar barely offer IPv6 support. My router supposedly does IPv6, except it doesn't. There are no upgrades to the firmware to support it. Comcast (my ISP) supposedly offers IPv6 support. I suspect the consumer router companies are selling IPv4 routers now when we run out of IPV4 addresses, in hopes of selling the "upgrade" to IPv6 in a year or two, as that can be the only reason why IPv6 support isn't offered.
Sad
Agent K: A *person* is smart. People are dumb, stupid, panicky animals, and you know it.
There isn't enough porn. What ever happened to the free IPv6 Porn project? :)
http://banjo.employees.org/~dwing/aaaa-stats.html
Plus, ipv4 is easy to manage; your average network engineer has IPs memorized for when things break, or at least a somewhat logical addressing scheme so it's super-easy to guess the IP of a specific component when DNS breaks or is inaccessible, to be able to log into the device and fix it. the dot-quads make things really easy, four integers with a max of three digits (people memorize numbers and spelling most easily when broken down into chunks of three or less) per integer.
You can make it as hard as you want to. It does not have to be difficult. I have a substantial network at home and my scheme is:
"My /48" : "the VLAN" : "host"
My /48 is pretty easy to remember after I type it in 50 billion times. Its just one number. I have no problem memorizing multiple CCs, SS#, phone #s, so memorizing my /48 prefix isn't very challenging. I will be very pissed when/if I ever get "native" ipv6 and lose my tunnel and my ISP gives me a new /48 via DHCP every week.
Anyway, the VLAN is encoded very simply, blah:0100:blah is the /64 for vlan 100. I could do something ridiculous and convert 100 decimal into 64 hex and then encode that as blah:0064:blah but that is a complete waste of time and brain cycles.
The host is also beyond simple. Take a wild guess what my static host address is for a router? How bout blah::1? If, as usual, I have multiple routers in a vlan they number up from ::1. Luckily I have less than 24 routers... can you guess why? My DNS server lives at blah::53 and web server at blah::80. Take a wild guess what address my ntp server lives at?
I only use static addresses for stuff that matters... pure clients just get whatever radvd gives out, much as I don't care what ipv4 address my dhcp server gives pure client machines.
Also, frankly, lets be honest here, the days of having to justify buying a dedicated $15000 sparcstation with 4 megs of ram to barely handle running BINDv4 over my thinnet coaxial ethernet are kinda long since over... I have no shortage of secondary/backup DNS servers, and I can't remember the last time I completely lost DNS ...
"Science flies us to the moon. Religion flies us into buildings." - Victor Stenger
The problem is that ISPs and router makers have been dragging their feet over IPv6 for years - there was just no ROI in the short term for them. Rolling your own solution is doable, but doing it properly without ISP or router support is still quite tricky.
Now of course, as IPv4 running out becomes a concrete problem, it's cheaper and simpler to focus on deploying carrier grade NAT - i.e. multiple end-users sharing a single globally routable IPv4 address.
I do have IPv6 on my home network; I've got a dlink 825 flashed with openWRT as my primary router (linked to cheapie DSL modem with PPPoE) specifically so I could run the AICCU client for sixxs.net for my IPv6 tunnel on it. RADVD handles advertising the tunnel prefix to the home LAN, so all my PCs, VMs, laptops etc have IPv6 addresses using one /64 out of my allocated /48. I had to do it this way as I have a dynamic IPv4 address, and the handful of expensive routers that do support proper 6in4 tunnels generally only work if you have a fixed real IPv4 address.
I have a similar setup at work, but there it's just a linux box with the main fixed IP router forwarding the 6in4 packets to it.
The main use for this for me is to be able to connect direct over IPv6 to any of my machines at home (mostly my NAS or VMs), using SSH or RDP etc - I've just put the static IPv6 addresses into my external DNS for my own domain. Very handy if I want to test how one of our hosted services looks from outside the work network, or to queue up a download so it's ready when I get home. I even use it at home to connect to work; since the IPv6 takes a different (shorter) route, it's quite a bit lower latency than connecting to the same machine via IPv4 and VPN (my firewall allows such connections from and to work, but not the general outside world)
So it has its uses for a techie like me; but for the average home user? It's way way beyond their ability to setup. Even setting up a single machine with a dynamic IPv6 tunnel is too complex, and certainly using 6to4 or toredo or the like relies too much on having a nearby translation gateway, and they're still pretty thin on the ground leading to a pretty rubbish IPv6 connection.
I honestly think we're going to see a lot more carrier-grade NAT from ISPs - it's already happening for mobile devices - than we see major IPv6 rollouts in the near future. Of course, that will break even more than it already is P2P apps like skype, bittorrent, IM file transfer etc etc, and of course running your own IPv6 tunnel will be that much harder behind a double NAT firewall.
Remember kids, it's all fun and games until someone commits wholesale galactic genocide.
...The embarrassing thing is that Facebook, a site for doing social things that isn't about tech is available over IPv6, but Slashdot, which is all about tech still is not available over IPv6.
Oolite: Elite-like game. For Mac, Linux and Windows
How can the average homeowner tell if their cable modem/router is IPv6 capable? Or, is this a non-issue?
WRT to cablemodems:
You can only run, eh, "8 megs" or so over a single downstream channel... If your local cableco is selling services running faster than that, they must be doing channel bonding to do it, which requires DOCSIS 3 link layer protocol, and DOCSIS 3 certification / licensing / whatever has mandatory ipv6 support. Also no one in China has manufactured a non DOCSIS 3 hardware compatible cablemodem for I would guess a couple years now. Does not exclude the possibility of your local cableco having a warehouse full of brand new, "old" DOCSIS 2 modems.
Most people "get the cablemodem for free from their provider". Its possible you live in an area were you own and pay for the modem, much like the DSL guys do. Assuming you purchased it, look for "DOCSIS 3 support" on the shipping box, or just google for your model cablemodem and "docsis3" etc.
"Science flies us to the moon. Religion flies us into buildings." - Victor Stenger
WTF is this crap? Don't just make something up and post it as a fact. DOCSIS 1 .x - 2 supports up to a 42mbit (minus overhead traffic) downstreams on non Euro-DOCSIS systems. This is because ATSC uses a 6Mhz channel for the downstream. If were talking EuroDOCSIS it's PAL and has 8Mhz channels so you could get up to 55Mbit (minus overhead traffic). Now this is total channel capacity so if you have multiple high usage users you'd need to implemement load-balancing.
DOCSIS 3 takes over from any speed above 42Mbit for Docsis and about 55mbit for Euro-DOCSIS. IPv6 is natively supported in D3.
How can the average homeowner tell if their cable modem/router is IPv6 capable? Or, is this a non-issue?
But, if I unplug my modem and take it over to my computer so I can type the model in, google doesn't work.
un-ALTERED reproduction and dissimination of this IMPORTANT information is ENCOURAGED
It doesn't have to be government mandate.
Seems like a very simple example of something that can be encouraged through the tax system.
1) Define a company as an ISP using a set of criteria that work for this purpose (something like, supplies network bandwidth and IP address/es to paying customers)
2) Increase their corporation tax by some amount proportional to the number if IP4 addresses they assign
3) Decrease their corporation tax by some amount proportional to the number of IP6 addresses they assign
They'll figure out a way to sell it to the consumer.
As stated above, we don't have to force them to make every IP a v6. Just to make them assign enough IP6 addresses that the chicken/egg problem is overcome.