Slashdot Mirror
IPv6 Traffic Remains Minuscule
judgecorp writes "Even though we are running out of IPv4 addresses, IPv6 traffic is still not taking off. In fact it is less than one percent and falling, according to a report from Arbor Networks."
← Back to Stories (view on slashdot.org)
what is filling in the gap between ipv4 and ipv6? ipv5?
How many home routers support IPv6?
I've tried to set up my home network to prefer my IPv6 tunnel from Sixxs over IPv4 but there are oh so many hosts on the net that only support IPv4. Slashdot.org is a great example of such a host...
Maybe if more websites and other services actually supported IPv6 we'd see it "take off". Currently it's a bit like complaining about no one taking the train when there are only two stations in the whole country.
Greylisting is to SMTP as NAT is to IPv4
IPv6 adoption wasn't just going to happen overnight.
What about the rest of the world? Brazil, Russia, India, China, France? Other countries? Aren't they using it more than the U.S.? I think the U.S. will be the slowest to migrate. We have the least need. Or so people perceive it.
That's because most people's home internet - eg the fantastically expensive Verizon FiOS network - don't even do IPv6 routing yet.
StoneCypher is Full of BS
For the players of the internet infrastructure market ipv6 is uninteresting. They've divided the v4 address space between them and have crafted strategies to dominate the market. This is nothing exceptional or strange, but rather applied business strategy. I'd bet that they're willing and even motivated to fight v6 adoption for the purpose of maintaining their position at current terms of technological environment. As they see it, v4 is delimited to something ownable, controllable. V6 is for them something that would reset the playing field and make a significant portion of their investments obsolete.
i don't know how it looks in other parts of the world but here in austria it is close to impossible to even find a provider that will offer you a routable ipv6 address. i checked the biggest providers available in my area and the only thing i could get would have been a tunnel.
My ISP (Virgin Media) have said that they've "got enough" IPv4 addresses and that they'll start to look at IPv6 "sometime in 2012", so it's not like people are falling over each other to get IPv6 support up and running.
Let's say I'm an ISP and I have a bunch of IPv4 addresses. I can invest and convert my customers to IPv6 or only add new IPv6 customers. Or I can make like the IPv4 addresses are a rare commodity and charge more for them.
Hmmmm... Gouge or invest, what will it be, what would Ma Bell do, where's my federal subsidy ?
Nullius in verba
IMHO, if the IPv6 spec drops the Interface ID requirement, then IPv6 use may change. I don't think that anyone is particularly jumping for joy to have their machine uniquely identified on the net.
Nothing but the blind men and an elephant in the internet age.
http://en.wikipedia.org/wiki/Blind_men_and_an_elephant
From what little I have access to, I see it increasing. From what little they have access to, they see it increasing in absolute but decreasing in relative. I'm sure someone else out there can get an equally meaningless datapoint. Who cares.
I've switched at least some of my infrastructure over to v6. It just works. How boring. In other news, the sun rose in the east today.
"Science flies us to the moon. Religion flies us into buildings." - Victor Stenger
Guys, I am sure there's 100 reasons why this is a dumb idea, but let me just ask.
Why couldn't every ISP that offers consumer-grade connectivity (that doesn't allow serving) do NAT at the ISP level, and give you a few IPs out of their internal Class A? Why couldn't portable devices on a wireless network (looking at smartphones and other 3G stuff) have a NATted IP from Verizon/Sprint/AT&T? Yes, I know it's not as cool to have a NAT address, but you could pay a couple bucks extra for a static, public IP as many broadband customers do now. As far as I understand, for any day-to-day web surfing, chat, media, etc., that would work just fine, unless you wanted to create a public server (at which point you'd buy a static/public IP).
Seems to me this would allow everybody's IPv4 routers and stuff to continue to work just fine, and reclaim tens of millions of IPs from ISPs that would no longer need them (as they could drop to a small number of public IPs to share across the NAT users).
I'm sure I'm missing something, as this scenario seems entirely too convenient to have been ignored. But it should make for good discussion.
...right on time for when the Mayan calendar ends :)
In that case, the transition from NTSC to ATSC might be a better analogy. It needed an act of Congress to make it happen.
Something I've been wondering...can't ISPs do IPv6 externally (what outside world sees) and IPv4 internally (their subscribers) - a bit like NAT but bigger?
I doubt many ISPs have more than 2^32 subscribers on a single subnet yet so it seems to me this would solve the problem for a very long time to come. All we need is some routers which do IPv4 to IPv6 conversion at the very top level.
No sig today...
but people with cable or satellite tv where ok well maybe with cable you need a box but in some areas before that analog cable was cut down and you needed a box to get most channels any ways.
Now IPV6 is says that you cable box needs to be swapped out as well the back end systems at the head end as well.
It's like pulling all the SD boxes or all the MPEG 2 HD boxes and going MPEG 4 only. It's a lot of hardware to swap out.
When it becomes difficult for the average user or corporation to get an IPv4 address
Hosting companies such as Go Daddy charge per IP address. And given that a lot of deployed web browsers still require a distinct IPv4 address for each distinct site, SSL site operators have to pay up.
I know I don't. Could they have come up with a more hard to remember addressing scheme?
Why not force the mobile phone segment over to IPv6? It's the fastest growing tech segment on the planet in need of address space! The home, corporate, isp IP need, though growing, pales in comparison to what the mobile market is demanding now and in the near future.
Has anyone at IANA even suggested this? Too much burden on the telco's to get their shit together? For all the 'claimed' tech. infrastructure being deployed around the world, doesn't seem like IPv6 is getting as much lip service as an implemented priority, compared to the constant bitching of impending IPv4 doom.
I'm a network admin and I honestly don't know enough about it to be proficient or even comfortable. I, along with many in my position, are so swamped and overwhelmed in day-to-day operations that there is no chance of learning enough about it to be able to undertake the kinds of overhauls and ripple effects it would bring. I'd love to get some training and utilize it if there were some gains to be had without needing to replace massive amounts of gear or reorganizing/restructuring things... I just don't see it happening.
http://teasphere.wordpress.com - A little spot of tea
... all the pr0n, warez, tunez, and moveez sites were to allow free access for non-tunnel IPv6 users.
now we need to go OSS in diesel cars
The thing is that there is a difference between not having any spare IPv4 networks to hand out from the top and Internet not working. Internet is kept together by way of network address translation. Correct me if I am wrong bearded network gurus, but to my understanding it is the 65536 ports that fill in for lacking addresses, correct? I mean, that's how and why NAT works, right?
Put another way, a home network usually is given a single address by its connecting entity - the ISP usually, but that doesn't restrict it to a single user. Same thing, different scale is happening on Internet. We are essentially NAT-ting everything we can. Maybe it is because of that that IPv6 won't kick in for another X years or so - I mean, why, what's the problem? NAT keeps Intertubez connected and blinking.
I can tunnel over to IPv6 (waste of time and bandwidth) or I can continue not to care about IPv6 and use the Internet like 95% of internet users who do not care for anything under the hood except for being able to access their email, banks, and facebook account.
All the cool kids use IPV6.
It could work as you describe, but not without some massive investment by the ISPs. That investment would be better made on IPv6 which is a definitive solution as opposed to the band-aid that NAT is.
Score: i, Imaginary
Wow, the sheer incompetence is mind blowing. Yes, we could go almost indefinitely without IPv6, but it would be a situation of NAT upon NAT upon turtles, and well, lets be honest from there it's turtles all the way down.
The problem is that there are a lot of services which don't work with NAT, and if we limit ourselves to just the ones that do, there's all sorts of cool things which nobody will bother to invent because they're impossible.
As has often been suggested around here, just because something is good enough, does not mean that it's acceptable. If the telecoms weren't so damned greedy, we could have IPv6, they're already gouging us on service as it is, requiring them to actually provide proper IPv6 shouldn't require rate changes.
It's not that you are missing anything, and it's not even the area you are talking about that is the main problem... it is corporations. (isn't it always :) Through greed, inefficient use, and myriad other issues *that* is where the bulk of the waste is. ISPs could definitely run like any global or even national company does with a private address space and NATing. I'd say 75% of users would never care or know the difference, the other 25% would gladly pay a few dollars to have an actual address. I'd much rather pay for something like that than artificial bandwidth restrictions and bullshit like that.
Smartphones/wifi devices are a pain in the ass network-wise in general, but there is almost *zero* reason NAT would not be perfect here. People aren't hosting content on them and they are basically Internet viewers as it is. Give yourself a bit more credit!
http://teasphere.wordpress.com - A little spot of tea
In truth, IPv6 for an internal network doesn't make any sense at all, it's not worth the switch for most people. For the internet, it may make some sense if the cost of a fixed IP address is too much, and you provide or use a service that can't use NAT, and the people who are trying to reach you are from a new audience who are not IPv4 bound, and other means like dynamic DNS are not practical. The key question, isn't the number of IPv4 addresses available, but the number that absolutely must be fixed for people to go about their business ... and that number is probably closer to a few million, than to 4 billion.
IMHO, the key problem here is that the powers that be are not letting IP addresses be allocated by the market, but rather by assignment. The market would automatically adjust supply, and demand, and once the cost reached a certain threshold (if ever) ... that would determine when people think it's worth it to switch.
I remember a few years ago, I talked about how IPv6 was overrated on slashdot and in the tech community, and promptly got blown off and down voted. They may have had a fundamental understanding about the technology, but didn't jack fuck about the marketplace.
Can't speak to the expense, but wouldn't it be relatively easy to turn off NAT on the home routers, and give people 10 IPs with every connection, and charge an extra dollar for every IP needed after 10?
The reason it's getting complaints on /. is while most of the general public will be absolutely fine, the techie nature here means quite a few people are likely to be running servers on their domestic connection.
You really want one layer of NAT for that at most - the layer at your gateway. If your ISP puts you on carrier-grade NAT, you're stuffed.
Doubtless ISPs will offer a real, honest-to-FSM IPv4 address, but they won't offer it to domestic subscribers. It'll be business users only, and it'll cost extra. I'm not even going to get into the mess that'll come about if you're in the middle of a contract when the ISP puts you on a NAT'ed connection. Even if you can resolve the inevitable dispute (presumably by getting out of your contract early), doing so is unlikely to be quick or easy.
Not when I look it up. It returns no AAAA records. And I have IPv6 access.
Companies tend to use IPv6 DNS whitelisting, meaning if they don't think you really have IPv6 connectivity, then they don't return their IPv6 addresses in queries and so you end up using IPv4. Google does this for sure.
This makes it tough to measure how many people have/are using IPv6. If companies just switch their DNS whitelisting off (as they are expected to do on IPv6 day), then we'll see how much IPv6 traffic there really is.
And the article states:
'It’s ironic considering Netflix is one of the few major companies with an IPv6-accessible website.'
What does it matter if netflix has an IPv6 accessible website? Most of their traffic is through their VoD service, not their website.
http://lkml.org/lkml/2005/8/20/95
That is the key reason we will never see IPv6: the entities that have to do something to make it happen have no incentive to do it, and a significant disincentive. IPv4 can be controlled by a few large organizations -- large telcos, governments, large technology corporations. IPv4 addresses are scarce and it is impossible for any new entity to come along and start challenging Verizon or Bell. Things like RFC 1918 addresses, NAT and tunneling make is possible for users to get stuff done in the face of IPv4 limits, so there is little subscriber-driven requirement to upgrade. End subscribers -- even very large ones -- essentially depend on the connectivity providers to lead the way in this sort of upgrade transition, and the large telcos have nothing to gain by giving up their de-facto oligopoly power in the market. Why should any guy with a couple of microwave dishes be able to go into business up against AT&T? That would be bad for business. As long as he does all that with RFC 1918 addresses, that's fine. But if IPv6 came to town, a guy like that would be selling fully routable connectivity, and that's no good at all.
...SonicWall supports IPv6, for example, but don't document how you actually enable that support in their firewall or router products. I sent a tech support request asking for more information. Many days on, it's listed as "being researched". It's their bloody product, how the hell can they NOT know how to enable a function?
With home routers, the problem is worsened. Many run Linux, which means they do nominally support IPv6. IPv6 has been available for Linux since 2.0.20 and a standard part of the kernel since about 2.1.8. Any home router running an older Linux kernel than that has issues going well beyond IPv6. The first problem is that this capability isn't provided within the standard administration functions. The second problem is that they don't document logging into the router directly (rather than using the provided gui), making it almost impossible for anyone but an expert to actually do what is needed to load in the kernel module and update the router tables.
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
This only works when your provider has a deal with google to do so, as explained on Google's IPv6 page. For my IPv6 connection at home this works, on my IPv6 capable VPS it doesn't work... So it doesn't just work for everybody, although it will probably work for the rest of the world on IPv6 Day.
Yet Another ArborNetworks Advertorial. Brought to you by Slashdot.
I can't wait to get back on the v6 net.
I'm waiting to get my sixxs account re-activated. I moved house, and then I registered again, so I now have a -RIPE handle and a -SIXXS handle.
Get your own free personal location tracker
That is a workable idea as a stopgap; in fact, it is the solution ISPs came to about a decade ago when the IP addresses were originally set to run out. But now even with that solution, we are still about to run out of IP addresses. The fact is, even NAT has numerical limits.
"First they came for the slanderers and i said nothing."
If Facebook, Google, Ebay and Paypal gave notice that they would be IPV6 only in 12 months, ISPs would jump. Imagine the rage if Auntie Joy couldn't see baby photos or search for recipes.
Deadlines can be a wonderful motivator.
Why couldn't every ISP that offers consumer-grade connectivity (that doesn't allow serving) do NAT at the ISP level
Because breaking the internet into two classes of users fundamentally breaks the internet. Having an internet connection without a publicly addressable IP address is like having a home telephone without a phone number. Can you imagine the phone tag hell you'd get into if half the phone users out there couldn't be dialed directly?
No, NAT is a horrible hack. Just because you don't need to be addressable today doesn't mean it won't become important in the future. If the average internet user can't be addressed, no one will write good server apps for the average internet user. We need to preserve this capability.
Give me Classic Slashdot or give me death!
ISP level NAT breaks: Skype, Blizzard's patching system(well, really really slow), Bad Company2, Modern Warfare 2, Sending pics/files to friends over [insert chat app], Punkbuster, VOIP, Starcraft1/2, Warcraft3, Diablo2, Civ3/4/5, many other games on 360/PS3. Small list for examples. Incoming ports are required mostly for hosting games and certain DRM features.
But hey, end users won't notice this at all.
You may think... but all of that works fine behind NATs.. yeah, because uPNP, which doesn't work at the ISP level.
You may also think.... most cell phones use an ISP level NAT.. yeah, but 500MB/month avg for a cellphone is not the same as a broadband connection that can tear through 2GB just watching a movie on Netflix.
Just offering another point of view.
Or do they intend to use 'ipv6 is too hard / expensive' as an excuse to convert every broadband users' comp into dumb terminal playing 'Mother May I', where DNS = ISP and every request is logged and shared with the FBI and the MAFIAA.
The reason IPv6 hasn't taken off is because it is an insane, stupid, standard and everybody knows it. It's like the network god people were taking way, way too much lithium and mixing it with their crystal meth that day. 10^38 addresses is enough to give every single gram of mass in the planets of the solar system their very own IPv4 address space! Gosh, sure, why not? Nanites are coming, after all, and they'll all need their own unique non-NAT'd IPv6 address.
The good news is that they skipped 5. Here, I'll do a better job of inventing the next Internet protocol. IPv5:
48 bit addresses (add two bytes to the left of existing IPv4 addresses, otherwise use precisely the same packet header, four whole bytes longer, six if somebody wants to add more checksumming or the like while we're at it).
Oh, wait, I'm done. That gives us 65,536 IPv4 address spaces, which is enough for every country on earth to have one, bigger countries to have 2 or even 3. It's enough to trivially provision every human on the planet with their own block of 256 addresses, with enough left over for gorillas, chimps, cetaceans, and dogs to get their own as well after we're done uplifting them, and that is without NAT.
Existing routers can probably be damn near hacked in firmware to manage the longer addresses. Existing route tables continue to function with a similarly trivial hack. The US gets the block 0.0.x.x.x.x, so all existing addresses in the US don't need to change, they just need a script to be run to prepend a couple of zeros (sorry Europe and China, but we invented the Internet and have the most addresses already assigned so by either measure this must be so). After that, we can just give countries their own block and encourage migration to the same IPv4 address(es) their hosts have now, but with their very own country code prepended. They can make up their own internet authority to manage their own internal addresses.
Naturally, this extension should come with plenty of room for NAT. We can establish one half of the space for NAT. In fact, we can make the first bit the NAT bit. No packet with the leading bit set is externally routed. Sure, we sacrifice 32,278 4-billion-address IPv4 subnets that way, but we make it REALLY EASY to make a home network address normal humans (or small business admins) can remember: 128.x.x.x.x.x, fill in whatever you like for the x's and we're done. We can either make: 255.0.0.0.0.1, loopback or make loopback x.x.127.0.0.1 at block 4 (for any address blocks), just as it is now, or both. I like both. Why not?
See how easy? See how extensible? Everybody does fine managing four byte addresses already, and basically in IPv5 one will now manage the same four byte addresses plus a country code. We can PROBABLY even make the country codes for IPv4 address spaces and the country codes for countries MATCH! What an idea! That will make them REALLY easy to look up!
Of course, that means the US starts out with both 0.0.x.x.x.x and 0.1.x.x.x.x, but that's fine, we invented the telephone too and we're most likely to exhaust the whole IPv4 space on our own even after other countries get their own and move out, so we probably will need two from the very beginning anyway, and why not 0.1.x.x.x.x? It even makes sense there.
Why only 48 bit addresses? Given NAT, it is actually very unlikely that we will really need more than two extra bytes -- truthfully we could probably make things work forever with only one. And y'know, if we ever do need another byte or two beyond this, in a century or so we can add it then and just as painlessly extend to 64 bit addresses. But honestly, I doubt that we ever will.
rgb
Even when the experts all agree, they may well be mistaken. --- Bertrand Russell.
If everyone is placed behind ISP-level NAT, which is the way things appear to be going, particularly in Asia, BitTorrent would go away. You can't do peer-to-peer communication if you can't receive incoming connections.
ISPs would love to get rid of BitTorrent, because it's more than half the traffic their customers use. ISPs would also love to get rid of people running servers off their home machines, something also prevented.
It would not surprise me at all if the movie and music industries would bribe^W contribute to the campaigns of congresscritters to make IPv6 illegal or inhibit its adoption. It would certainly be in their best interest.
"Screw Sun, cross-platform will never work. Let's move on and steal the Java language." - Visual J++ Product Manager
Doubtless ISPs will offer a real, honest-to-FSM IPv4 address, but they won't offer it to domestic subscribers.
Maybe i'm an optimist but I'd hope that ISPs would make public v4 addresses available for a small extra charge. It would be a new revenue stream for them and they really only need to get the lusers behind NAT to free up more than enough IPs for their other uses.
note: i'm known as plugwash most places but i screwd up registering that here somehow in the past and now can't register
That should motivate change.
So you are suggesting that going to 128 bit rather than 48 bit is the reason that there is no takeup of IPv6? I don't think it is the case. Whether you add 16 bit or 96, the code for routers and computers needs to be rewritten. This is the reason why prefixing 0.0 still makes your IPv5 addresses a different address (so all addresses in the US still need to change). Since everything needs to be changed anyway, it makes sense to throw in a few useful extra features. By the way, one of the reasons to go to 128 bit is to allow efficient routing tables and to minimise fragmentation of the address space, which would still happen with your IPv5.
They don't have much choice. It is pretty clear that v4 addresses will run out for at least some ISPs before we reach a situation where a v6 only connection is acceptable for clients. Therefore it is pretty much inevitable that ISPs will have to deploy some sort of ISP level NAT regardless of whether they also deploy IPv6.
note: i'm known as plugwash most places but i screwd up registering that here somehow in the past and now can't register
"I've heard we have been running out of IPs for years now and it never happened so I don't believe you now."
But when did they say we would run out? It was never tomorrow or the next day..a decade ago the predictions were always sometime in the next decade.
"ISPs benefit from selling IPs to consumers or selling off unused blocks in the exchange"
Except for all the ISPs who have no more addresses to hand out and now must purchase large expensive boxes to keep connection state for protocols..(or buy IPs at ripoff rates from third parties). Routers that can forward packets via an ASIC at wire speed are expensive enough. What happens when they also have to understand the content of data at much higher levels using general purpose processors? What about CALEA compliance hell caused by no longer being able to associate addresses with users without long term storage of massive amounts of state? What about commercial web sites that need an IP for an SSL site or a colo facility or a large business customer? Yea you might be able to make a few dollars more from selling static IPs to the few people who ask for them but in the bigger picture IPv4 shortages suck for operators.
"IPv6 adoption is a failure and will never happen"
Companies who in total are responsible for more than half of total US bandwidth utiliziation will be posting AAAA records on IPv6 day AND the largest ISPs in the US all have active trials. IPv6 is uptake is currently following an exponential curve.
You never said what your problem with 128bit addresses is. Hacking support for 16 more bits into v4 isn't fundamentally different than adding 96. You can't add 16bits without breaking pristine v4 devices, and any such hack would be ugly and messy. If you have to touch/hack/upgrade every single v4 device to support your scheme, how about we just go all the way and upgrade to a totally new address scheme? Its not hard, it just works, and its easier to dual stack (or run straight v6) than you'd imagine.
The slow part is- up until now there has been zero benefit, so there is no reason for ISPs to provide v6 support, and extremely broken dual stacks fail when a query returns both A and AAAA, so content providers are understandably nervous. Its changing though- I think by the end of the year most popular sites will be available in v6. For the larger folks its simple to implement, there are only a few devices on the edge of the network that need v6.
For home use I'm loving v6. Suddenly every device I have is individually addressable from the internet instead of NAT + port forwarding. Even my 3y/o network printer picked up a v6 address, and it just works. I'm dual stacked, so my internet experience hasn't degraded, most v6 aware clients will prefer v6 if a AAAA is offered, but will use v4 otherwise, its seamless.
Name one thing NAT breaks that an ISP actually cares about.
Most of the stuff, like home servers and bittorrent, that NAT breaks are things ISPs would actually be quite happy to be rid of, or at least charge an outrageous premium for a static IP.
So you are suggesting that going to 128 bit rather than 48 bit is the reason that there is no takeup of IPv6?
No, the main reason for NOT adopting IPv6 is that it is inconvenient and carries disincentives for many of the big telcos.
Since the main reason FOR adopting IPv6 is the putative exhaustion of address space, why not create a standard that addresses THAT problem alone, and that does not create a squirming mass of additional incompatible and incomprehensible features and headaches?
This was not what the great-grand parent was saying, so this was what I addressed. But I wouldn't say that there is a "mass of incomprehensive features and headaches". Generally, an IPv6 packet is much simpler than an IPv4 packet. Of course, if there are IT guys are too lazy to learn IPv6, I will look elsewhere.
- Got IPv6 connnectivity set up on my computer
- Set up all my servers to support it myself.
- Been in my firewall scripts for about 3-4 years now.
So I'm WAY ahead of the game compared to most people BUT:
- No ISP gives you it by default.
- No server host gives you it by default (mine let me add it by clicking a button in the control panel).
- Most dedicated / virtual server hosting places won't set it up for you in their default images (ALL IPv4)
- It is of precisely zero use, even to me, because nobody uses it, there's no reason to use it and nobody supports it.
When Slashdot (a highly-geeky site) publishes their AAAA records, then we can discuss the future of the rest of the Internet. In the meantime - NOTHING and I don't even get IPv6-originated spam or automated access attempts.
That's not even *mentioning* the technical problems and amount of upheaval and management necessary (ever changed a complicated iptables script to support ip6tables?) that I decided to do "just because".
The IPv6 net is largely silent, except for techno-geeks pinging it because they can. And it *literally* takes a day to enable it, even for the largest sites.
A long time ago I worked in a data center where we had a Novell server that was consistently maxed out on connections. I made a snarky remark about kicking managers off the server to make room for analysts and developers. One manager who was known as POD or the Prince of Darkness did get kicked off by one of his developers and he went off on me. I replied "So it's not a problem until *YOU* can't log on." I think the same mindset is slowing the adoption of IPv6. As long as the vast throngs can access the net to get their email and the latest news there won't be pressure to force the migration.
So the top 64bits, or as we like to call it in the modern age "one native integer" (on now-default platforms) is the network part, and "one native integer" is the customer part of the address.
(Okay, its a "long long" on a commodity PC as of the date of standardization, but they were kind of planning for the future).
Oh the horror.
It's actually quite elegant. No real netmasks, just two integers. Very easy to detect the special addresses by examining the first byte (easily done in FPGA and embedded CPUs). Zero crowding.
The cost savings of being able to load a raw integer into a register and do math on it (a fast lookup table won't even care about the "Real Byte Orderingâ" because there is no promise of "ranges" so no question of adjacency.
So a natural size (two integers), natural math in a platform agnostic way, easily automated and easily aligned (as opposed to shifting bytes etc.)
Your annalysis is shallow. In trying to save bytes you add significant instruction/processing overhead. The bandwidth will shrink the presumed burden of communicating the extra octets to approximately nil, while the fast hardware can be faster by virtue of the natural ordering.
Innocent people shouldn't be forced to pay for inferior software development.
--"Code Complete" Microsoft Press
The summary badly mischaracterizes the article. The article reports that the volume of ipv6 traffic is growing.
IPv6 does appear to be accelerating down the run-way, if not actually taking off at this point.
Upcoming we have world ipv6 day on 8 June where it is hoped all of the problems with publishing AAAA dns records can be found, and unreasonable fears put to rest. With a little luck the browsers will cope gracefully on that day and the major sites will decide that offering AAAA records on their main domains just makes sense.
Once it is realized serious content is available on ipv6 networks with the choice of PNAT in the ISP network (because they don't have enough ipv4 addresses) vs ipv6 can start deploying ipv6 to get better service.
There are no alternatives as far along as ipv6 to deal with the ipv4 address shortage. Their is a complete ecosystem of gear that can support IPv6. You probably have to search for it, and you might have to deal with bugs, but that is 100% better than vaporware. Admittedly there are weak spots in the provide edge and in the consumer edge where it really matters. But in practice all any of that means is that it is time to be testing and getting out the bugs in the equipment.
Eric
You asked the wrong question...
I tried to join the Comcast IPv6 testbed. My ownership of an XboX 360 eliminated my potential participation (even though I know how to do a 4to6 translation endpoint).
Its not the home routers etc, though I am sure that the large number of NAT-presumptive home routers is no small drop in the bucket, the bulk of systems rolling out are plagued by the IPv4 Assumption.
Gaming, Mobile Phones, Clearwire Routers, the entire cell-phone data support infrastructure, none of these sections of the "internet base" have exerted one forward-thinking jot of effort to join the IPv6 reality.
"Bit-Wise, Byte-Foolish" is the new "Penny-Wise, Pound Foolish" for the information age.
Innocent people shouldn't be forced to pay for inferior software development.
--"Code Complete" Microsoft Press
I think your 80% figure is wildly optimistic. Between websites that don't offer v6, non-v6 capable routers and non v6 capable end systems (IIRC XP doesn't support IPv6 unless manually enabled on the command line) I would expect far more than 20% of traffic to remain v4 in a scenario where the ISP offeres IPV6+Natted IPv4.
The ISP could try to push things along with new routers and tutorials but unless they offer some largecarrot and/or used some large stick (for example blocking IPv4 acess to youtube) I doubt many users would take them up on the offer unless they had no choice.
note: i'm known as plugwash most places but i screwd up registering that here somehow in the past and now can't register
The issue isn't the rewriting of the code -- anybody can do that. The code has even long since been written and is in e.g. the linux kernel at least. Machines could care less if the addresses are 32 bit or 1024 bit -- you could make the entire packet header, allow every atom in the Universe to be individually addressed, and machines would still parse it once the code was written (and it wouldn't be that difficult to write, even making the code route hierarchically to distant galaxies to get at their atoms). Nor is processing power an issue, in routers or elsewhere. Nor is memory. I really do think that it is the extra complexity -- even people who could turn it on, don't turn it on. Why should they? They don't need the extra complexity. They (for the most part) don't know how routers function in the first place, and could care less as long as they route. They find it much easier to use a private internal space and NAT rather than try to deal with a backwards incompatible new standard.
I think you are missing that point altogether. The standard I proposed is not only backwards compatible, it is trivially so. If an IPv4 packet comes into a router, you just rewrite it "IPv5" with the default country code. You could even make a router that transparently strips the extra bytes of address space and writes IPv4 header packets out to connected hosts from which they receive IPv4 packets. All of this would actually be pretty simple to write and one could probably write it so that one doesn't even change the IP header layout if one used some of the "options" space in IPv4 for the extra four bytes. From the four extra bytes (two each for source and destination) you get what amounts to unlimited address space, especially when further eked out with non-routed NAT.
I absolutely agree that 128 bits allows one to do all sorts of clever things at the routing level, but it did so at the cost of making it incomprehensible and unwieldy at the human level and backwards incompatible. The history of computing is littered with the wreckage of good ideas (in principle) that ignored those two things, and one can easily count the success stories of modest kludges that were easy, and hence became universal (kludgy or not). Consider, for example, Motorola's flat address space vs Intel's segmented address space back in the 80's. Motorola's was clearly superior, right? Especially when the 8088 (that really needed it) became the 8086 (that still needed it) became in rapid succession 80[1,2,3,4,5,6...]86. Intel processors could handle segmented code in compatibility mode long after their processor was flat because of the vast "inertia" of humans who possessed legacy segmented code.
Now, count the number of Motorola vs Intel processors in use in the world today. Hmmmmm.
Hell, we don't really need a two byte extension to IPv4 address space. One would do fine for a few more decades and would easily fit into the options part of the IPv4 header. IPv5 could refer to five byte addressing, IPv6 (if it were ever needed) could refer to six byte addressing, etc. One could (of course) extend IPv5's header so that it can be backwards compatible but so that it allows for longer header lengths. Since one has 8 possible values in the version field, only one of which is ever used, that leaves 7 more bit patterns to differentiate and/or facilitate a seamless transition from IPv5 to IPvN for any N you like or need (starting by allowing for IPv5 to go ahead and "reserve" room in header-space for a more seamless transition for all future N+1 transitions.
That would have made (and would still make, if anybody just boldly went ahead and adopted it) legacy hardware useable with a FIRMWARE upgrade. That's probably impossible for IPv6. And right there you have it. Hardware inertia. User/admin inertia. No backwards compatibility. No human readability, even in xxxx:xxxx:xxxx:xxxx:: notation (which more or less acknowledges, BTW, that they went insanely overboard
Even when the experts all agree, they may well be mistaken. --- Bertrand Russell.
See other reply. I don't have a problem with it at all per se (aside from the vast, vast overkill), I only think it would have made more sense to build a bridge interpolating 4 and 6 and made the bridge a lot more backwards compatible. I also think that there is a bit of difference for an end user that simply plugs an IPv6-enabled machine or device in and it "just works" (because the administrator of that network has it set up to just work) and the administrator of that network. There is a larger barrier, I think, than you are acknowledging.
rgb
Even when the experts all agree, they may well be mistaken. --- Bertrand Russell.
Didn't the '.xxx' suffix get approved for pr0n sites, recently?
Make access to .xxx IPv6-only and you'll see millions of consumers really, really needing to connect to *cough* "ipv6.google.com", and the ISPs will finally have to cave and provide it.
That'll spark traffic and equipment sales!
Well, apparently, you only have to fool the majority of people for a little while.
i checked my roughter and its ipv6 ready and its a few years old however my isp does still hand out ipv4 addys. its just ipv6 will be used for new thing wile ipv4 will say where it is on old stuff and for everyone to play nice we will need to use hacks.
Sorry, but your scheme simply does not work as it still does not allow a legacy IPv4 device to connect to a "IPv5" device. As you mention, they need an upgrade in the firmware. And if you need an upgrade you can just as well upgrade to IPv6 to take advantage of the other features. I also don't think that IPv6 is much more difficult to grasp; ok, it needs to be understood by few people that deal with the lower level network. All the applications (and their authors) don't need to know how it works exactly, just that it might exist and deal with it this way (i.e. also doing a lookup for an AAAA record).
Now, count the number of Motorola vs Intel processors in use in the world today. Hmmmmm.
Both pale compared to the number of ARM chips with flat address spaces. Oh, snap.
And yet you have no idea how such a thing should work. You could build such a thing for users right now - NAT64 exists, it could be implemented. Just no one's doing it, because no ISPs are handing out only IPv6 etc. etc.
If there's one thing about IPv6 which grates its every single time it comes up someone stamps their feet and throws a tantrum about how stupid it is that it doesn't work seamlessly with IPv4, while immediately going on to demonstrate they have no real idea why you can't do this in the first place, and thus no idea about how to fix it either.
The moment you put out the words "well IPv4 devices will need a firmware upgrade..." *bzzzt* game over. That's what IPv6 needs. But a device with a 32-bit ASIC for handling IPv4 is just as broken whether you're adding 12 bits or 96 bits of address, and anything which can be firmware flashed to support an extra 12-bits can do 96-bits as well (and under IPv6 will have an easier time of it as well).
Since neither subby nor the self-serving linkfarm reblog site they submitted bothered to either link to the Arbor Networks article, or read it beyond the first few paragraphs, here it is.
A better summary might be that native IPv6 usage has "more than doubled" in the past six months, while tunneled IPv6 has declined. This is exactly what we'd hope to see, but maybe not as catchy a headline?
Even the ones that do, we should expect bugginess.
I mean, Google and Princeton can't agree on what a non-buggy DHCP implementation looks like, and that's small potatoes next to a new network stack.
My God, it's Full of Source!
OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
I phoned my ISP last month to ask if my internet connection through them supported IPv6. I know my PC does, and my router does, but when I asked my ISP I got told that "IPv6 is not a fully ratified standard yet, and when it becomes a completed standard then we we look at supporting it". Just goes to show you that one link can break the whole chain.
- James
How many people do you suppose are hosting FPS matches right now?
There are other categories of online computer games in which regular users are hosting other users, but I'd guess that FPS games are the most common case, and the one in which small differences in ping time make the largest difference.
As soon as comcast allows any computers to sign up for IPv6, the switch will happen QUICKLY (as in 2-4 years). The reason is that all of the other companies, client and server ISPs, will be under extreme pressure to either dual stack, or to switch.
I prefer the "u" in honour as it seems to be missing these days.
That sounds like a good description of Carrier Grade NAT. My understanding is that it's the presumptive solution to keep IPv4 alive until IPv6 is fully deployed.
The downside is that it will mean that you've got NATs inside NATs, and that users will be competing for ports, both making for a poor experience for end users. I think it's a mistake to assume that users running servers is unusual -- hosting FPS deathmatches is quite common, for instance.
An upside is that the moderately technically inclined -- those who do want to offer services on the Web, or use peer-to-peer services, will have a practical reason to get IPv6.
I think your suggestion makes a lot of practical sense. I doubt many people will actually understand and appreciate it though. Judging from the other comments they don't - it is actually quite shocking how negative most of the replies are. That is why it has zero change of being adopted. Oh, well, at least a couple of years from now, when we are fucked and IPv6 is still nowhere, you can tell everybody "I told you so" :-)
I find it curious that you are against "artificial bandwidth restrictions and bullshit like that", yet in favor of artificial scarcity of network addresses (which is what sticking with IPv4 means).
Score: i, Imaginary
They are mutually exclusive, and I also never said I was for "artificial scarcity of network addresses." I'm saying intelligently allocate what we do have and come up with a better system overall... IPv6 is not a panacea. There's a reason it is less than 1% after 15 *years*.
http://teasphere.wordpress.com - A little spot of tea
You can't, it already exists. See http://www.apps.ietf.org/rfc/rfc1819.html
Doesn't work with my existing 2wire home router nor does it even work with a router running IOS. I had the unique advantage of testing various compatibility tricks with ipv4 when I was writing my own TCP/IP stack for AROS. Since it requires a software update with the routers, may as well go IPv6 instead, specification was first released in 1998 and has a much broader widespread support already. Phones support it, every PC OS in the last ten years appears to support it. It's just network providers that need to sufficiently upgrade their infrastructure.
I have been behind double NATing already. No.
Change is certain; progress is not obligatory.
I think what some people are worried about is greedy ISPs charging per device.
As it is now, you've got one IPv4 address hooked to an el cheapo Netgear router, and the ISP is none the wiser.
With IPv6, you've got a cell phone, touchpad, laptop per family member plus Xbox/Playstation, Boxee, TV, microwave oven.
Don't think it could happen? Look at how they treat smartphone usage vs. "dumb" phone usage.
I'm not a lawyer, but I play one on the Internet. Blog
In personal computers? Are you even paying attention to the idea of what I said? I'm not talking about processors used in mobile phones that didn't exist at all in the 80's and 90's when the anecdote I was referring to took place. I'm talking about a specific battle between two -- well, really more than two -- CPUs that were in direct competition for desktop space during the "personal computer revolution". My point wasn't that "segmented architectures are good" -- it was that good, bad, indifferent -- segmented CPUs were the first to really cover the desktops of the US and then much of the world. During the time of the original 8088-based IBM PC, a large amount of software was created that used segmentation, and that software created an inertia that lasted for decades and continues on in a small way today. New processors created for completely different markets didn't have the legacy issue so of course they were flat -- the technical constraints that caused processors to be segmented in the 80's no longer really existed by the 90's, certainly not by the mid-90's. However, during the 80's, if Intel had announced that (say) the 80386 was going to be completely incompatible the 16-bit 64K memory segmentation model of the previous generations of the processor family, they would have opened the door for Motorola's competing (and arguably better) processor because they would have lost all of the legacy software that gave their system inertia.
The point still being that in a developed market, it helps to provide a compatibility bridge. It probably took a decade for segmented code to mostly disappear, given that all of the major PC applications (Lotus, DOS itself, the compilers, Wordstar, and so many more) were written for the 8086 segmented family and cost to rewriting them.
To be honest, the exact same human tendency is being discussed in the eternal "why hasn't the US gone metric" thread in play at the moment. After all, SI metric has many advantages. It is faster, easier to use (when both the imperial and metric units themselves are equally intuitive, the arithmetic involved with metric is doable in your head, where that involved with imperial is not). It is used by far more people worldwide. It is important to use a single system of measurement worldwide in a world-spanning marketplace. Any sane person, when asked which system makes the most sense and should be universally adopted, would choose "metric". Indeed, they'd almost certainly strongly suggest that we redo our clock into metric and dump this 1440 minutes or 86400 seconds per day nonsense (and might wish that we could move the earth back in its orbit to where it was e.g. 400 days long:-).
Yet they don't. Why not? Because imperial works "well enough" as is for them in their everyday life, and they don't really care about global problems brought about by their decision. They'd have to change, and learn. It would take work. I teach physics and would love to live metric, but I can't even get my own family to change. If I set the thermometer in the kitchen that measures indoor/outdoor temperatures to centigrade, it gets reset to fahrenheit. If I talk about speeds in kilometers per hour or meters per second, I get blank stares. I have the same problem myself -- I should think of all mundane velocities in meters per second (SI units) but my experience with mundane velocities driving is all in miles per hour. I can manage the mph to kph conversion in my head, but I still have to resort to the google units converter to go to m/sec. It takes too much energy for me to convert completely to a sane set of life units, and of course even if I do I still have to function in a backward compatible way for the huge number of legacy humans using the old units (including those I live with).
More examples -- how about religion? Talk about a "legacy application" -- it has been perfectly obvious for several hundred years now that things like Chr
Even when the experts all agree, they may well be mistaken. --- Bertrand Russell.
Efficient for machines, perhaps, but not for users.
I concede your points, well argued, sir.
/etc/hosts, nameservers, SSL, netmasks, local routing tables, port filters on routers now, but who will more or less have to go back to school and relearn networking all over for IPv6 because it is nothing like IPv4. They are every bit as excited about this as they would be if they were asked (once upon a time) to learn to administer an Appletalk network, a Novell network, Digital's old network, ATM. It is that different. It isn't a matter of altering their existing addresses in simple ways; it is a matter of completely altering their entire set of address tables. Worse, it is a matter of doing so in all probability with both tables in place, managing both kinds of networks at once.
/etc/hosts and all other tables and use a simple set of software tools to handle IPv4 addresses and IPv5 addresses. In fact, it could handle all the old software at the kernel level in a compatibility mode. Set your kernel to prepend and remove IPv5's country code and otherwise communicate with a legacy app in on-the-fly converted IPv4 packets. I've written networking code before, and have a fair bit of it I still use. It is (for better or worse) 32 bit specific in a variety of ways. It is utterly incompatible with IPv6. I'm just one single instance of the vast, vast problem you are ignoring. If/when Duke finally converts to IPv6, I either have to retire my network application entirely or completely rewrite its network stack. Worse, if I want to still be able to have it function on an IPv4 network (and there are bound to be lots of them for years) I have to make it transparently function with both stacks. Rewriting it will take me how long, exactly? Let's see:
However, I still think you are missing certain points. It isn't just the "IPv4 devices" that are a problem. It is all of the applications explicitly written for IPv4 and 32 bit addresses. It is all of the sysadmins who understand and can deal with
You are doing it at home and that's great. Doing it in an environment with a few hundred (department), or a few thousand (institution), or a few tens or hundreds of thousands (ISP) of addresses is every bit as attractive to the admins of those organizations as elective heart transplant surgery. The fact that the heart in question has sixteen chambers instead of four and is proven scalable to the point where it could pump radioactive blood from Godzilla's enormous toes right up to his teensy brain isn't really the point, is it?
On the other hand, getting a pacemaker for the heart you've got -- sure, it isn't as cool and certainly won't help Godzilla -- might just keep you alive until you die of other causes. A metaphor multiplied by a large population of developers, administrators, and end users all of which have to pay in time, money, education, effort, and pain for any change. You are quite right, in one sense any change requires new software so why not go for the Godzilla solution?
The solution that I proposed, however, would let everybody keep their same old
* First, I can learn all about IPv6. I'm pretty smart, and know networking in general fairly well, so perhaps this will only take me a week or two of intensive, self-directed effort.
* Second, I can learn how to program a network application in IPv6. Will the old socket systems calls still work? Are ports handled the same way? If I want to display addresses (my application does that as it actually is among other things a network activity monitor) how do I parse them, how do I print them out in human-readable form? For IPv4, I have things like Stevens and several other books that are precious like platinum for the network programmer who wishes to succeed. What similar books exist for IPv6 -- yet? Perhaps I'm lucky and I find one, or find other resources that let me template it and build a new stack next to my old one. Again, it's hard to i
Even when the experts all agree, they may well be mistaken. --- Bertrand Russell.
The problem is that IPv6 really is a lovely piece of engineering, and as I pointed out in one of my other replies to some of the objections, about as attractive as elective open heart surgery to put in a really, really technically superior artificial heart to just about everybody involved. If their primary design goal was to make the transition maximally expensive in human time (in order to optimize performance in pieces of hardware instead, for pete's sake) they could hardly have done, um, "better". Forget just changing the sizes of a few objects in the network stack, changing a few loop limits from "4" to "6" or "8" -- oh no. Rewrite everything from scratch, all of your old tables are obsolete in form, everything you've learned about managing TCP-IP network is now wrong. Back to school with you! Or, in the case of a small cap company trying to write an application with a network stack, back to the VC peoplewith you! I'm sure you won't mind giving up another 10% of your company, or your new company's profits spent rewriting your networking, or adding some more sweat equity, or taking time away from fishing or your family, or just plain going broke because everybody knows technically superior IPv6 is worth it! It's better for routers and that is clearly what really matters, right? We'd hate for machines to have to work hard at the expense of human time, after all...
:-)
rgb
Even when the experts all agree, they may well be mistaken. --- Bertrand Russell.
May I be the first to express shocked admiration that "minuscule" was spelled correctly?
To have a right to do a thing is not at all the same as to be right in doing it
An ISP can't NAT a customer without breaking stuff. Once an ISP runs out of IPs, they will have to stop accepting new customers and stop deploying any IP based devices. They will risk a massive lawsuit if they sell an "internet" connection that knowingly will not work correctly.
They will risk a massive lawsui
Do you have any evidence to back up that claim? I wouldn't think it would carry any higher risk of lawsuits than many of the things big ISPs are already doing. Further many 3G ISPs are already doing it yet I haven't seen any evidence of a lawsuit against them.
And to start with they can always give the customers who bitch too much a public IP back leaving those who don't know and/or don't care with private IPs.
note: i'm known as plugwash most places but i screwd up registering that here somehow in the past and now can't register