Interesting point...maybe he or she was thinking of it from Dyn's perspective of remote access. Each of those devices probably connect to a hostname with their own IP (using port forwarding) or something. Good point though..the second half of the graphic is what I thought was interesting. The breakdown per region is pretty interesting. It's hard to imagine that years ago people they didn't expect this to happen.
They realised this was an issue in the 1980s, back when it wasn't even clear that the internet would spread into a general usage network (rather than military/academic).
It took until 1995 to get agreement on what IPv6 would look like - but that's still 18 years ago. Plenty of time for a phased changeover :->
In my home I have: Nest, Sonus Playbar, Sonus Sub, 2 Wemo light switches, 2 Wemo outlets, 1 Wemo activity sensor, AppleTV, Ruku, ChromeCast, PS3, Apple Time Capsule, Apple airport, Work router for VPN, Work IP phone, Printer, Ooma home phone, 2 ipads, 1 kindle HD, and 2 iphones.
I have 1 public IP address and everything works.
We have an an IPv4 allocation problem, TONS of IPv4 space allocated to companies and not used. What we don't have is IPv4 usage problem, or the need for everything "on the internet" to be publicly addressable.
I've seen the math done on the unused IPv4 space by some institutions. In the end, even if an organization gave up their entire /8, you get something like 1 to 2 months of extra time at best.
That is still allocations, not usage. If we had a public market for IPv4 addresses there would be incentive to use them like the limited resource they are.
I wouldn't doubt it. I suppose it comes down more to personal opinions at some point, but I'd rather see an Internet of mass availability and resources. I can start a web service from my house because I can rent a VPS (including IP) for a few $'s per month.
Only one or two months, yes, because there's an allocation problem. Unless that's fixed (and it won't be), just giving back extra addresses will only result in them being squandered ... again.
You have 1 public IP address and everything works because there are external servers for you to bounce things off, in order for external signals to reach the internal devices.
If we want proper peer-to-peer communications to be possible, we need to avoid, where possible, the need for a routable middle-man to help out.
Also, Asia Pacific has long run out of IPv4 addresses - back in 2011. There are now services starting up over there that are IPv6 only. This will spread over the next few years.
IPv6 is the future. There is no point complaining that we can shuffle around a few addresses here and there to make up a bit of space.
I know how it works, but the point is it works. Let's not pretend that everyone getting Nest thermostats is a reason for IPv6. And Most applications need "man in the middle" services anyway even if everything was publicly addressable. Nest HQ needs it public servers, There is no need for my thermostat to have a public IP.
> And Most applications need "man in the middle" services anyway
For all instant communication (VoIP, chat, file transfer, setting your thermostat), a man in the middle schema is a hindrance. The communication where you publish, and then somebody reads data need a man in the middle.
If your thermostat does not have a public IP, you must buy it as a service, and it'll never really be your property. One can live with that in a thermostat (but let's not pretend it's a good thing), just think twice before automating more of your home...
I also need my ISP, my power company, etc. I'll never be an island.
This IPv6 utopia where everything is peer2peer just isn't going to happen, there is too much value in central services.
If we want proper peer-to-peer communications to be possible, we need to avoid, where possible, the need for a routable middle-man to help out.
That is the very reason many people are nervous about the switch. Right now NAT/PAT give home users reasonable security without the need for a firewall. The obvious solution is just to make cheap firewalls that can route properly, and default to blocking all incoming traffic. However, that brings up the same concern of needing some way to dynamically poke a hole through the firewall. Which brings us to UPnP, and I for one do not trust that at all, but maybe thats my tinfoil hat.
I totally agree that we need to move forward with IPv6 though.
Just because your router supports IPv6 doesn't mean it should open connections from the outside world. I'd expect all home routers to continue to default to "only outwards connections allowed" and let you specify open ports as you need them - exactly as they do now.
This is getting close to my point. If:
1) outbound only is going to be the default,
and
2) most services already need 'man in the middle' (directly service at least)
and
3) IPv6 is going to need to NAT to get to the legacy IPv4 addresses anyway
What advantage is IPv6 for most of my devices giving me?
IPv6 doesn't NAT to get to IPv4 addresses - you'll need to run both in parallell to get to IPv4 addresses, until everyone is switched over.
And the advantage is that when you _do_ want to go direct you aren't stuck with being behind NAT on your router and then another NAT at your ISP, and then possibly another NAT level, and then the same in reverse at the other end, so that it's possible at all.
And sure, it might not give _you_ anything you particularly care about. But when you find that you can't do something, because there are several layers of indirection between you and someone else, and now you need to run your own server in order to hook two machines to each other over the internet, then you'll wish IPv6 was available.
Ironically, plenty of companies assign IPv6 addresses from the RFC4193 Unique Local IPv6 Unicast Address space - so no real advantage over IPv4 for them with regards to that. For security fixated companies, non-routability is considered a desirable feature, not a problem.
Nope. Everything does not work properly as you described. Yes, your devices can get online, but you cannot get to your devices. For example your Nest thermostat must connect to an external server to communicate with your phone when you are not at home. This is a bad thing, since it means that you are now relying on a third party to provide connectivity.
Also, let's say you want to connect to your Time Capsule when you are on the go. Wouldn't it be nice to be able to connect to its public static IP address directly (assuming you set your firewall to allow such connections).
Lastly, as the Internet goes IPv6 only, your IPv4 only devices will get cut off. This will happen sooner than you might imagine.
Yes, there are large swatches of unused IPv4 addresses, but the demand for new allocations is much greater. Why cling to an old system just to offset its death by a few months when we can spend the effort on making the new system work better?
"IPv6 is the future and the future is NOW!" Is what has been said for years. And everyone overlooks the value that IPv4 has ala Metcalfe's Law. And that value can be further mined if we had a public market for IPv4 addresses. IPv6 is not the only solution.
Let's start admitting that IPv6 offers no (or little) present day net benefits compared to the benefit of currently connecting to all of IPv4. And let's stop with the NAT booggy man argument. NAT works well, as much as any tech works and isn't perfect. And let's stop pretending that only thing stopping a peer 2 peer utopia is the lack of globally unique address. There many economic forces in play beyond that.
NAT works as well as it was designed to, but it is flawed by design. Simple things like "what is my IP" become complex thanks to it. It has it's uses, but not for consumer connections.
IPv6 does provide real world benefits, not the least of which is the simplification of allocations and a huge price drop in terms of address assignments.
And all of those things are wasting power, bandwidth and routing resources by continually connecting to/polling external services to see if they need to do anything. Instead of more passively listening and waiting for incoming traffic.
I'm a bit taken aback the utter lack of preachiness in that infographic. It seems most infographics like to tell you what you must do, right now, what the hell are you waiting for, why aren't you taking action right now, etc.
Quite refreshing that, no matter how simplified or slightly erroneous the data is, it's just kinda hanging out there.
To summarize: we'll run out of IPv4 addresses, and they don't work interchangeably with IPv6. Oh well.
I know that the standard way to state ipv4 is 2^32, as it does in the infographic, but isn't that inaccurate? since the 255 is not used, as well as many other reserved IP addresses, shouldn't an infographic actually get the exact IP addresses available, and not the theoretical number that anyone with a calculator can figure out?
10.0.0.255, 10.255.255.1, etc.. are valid ipv4 addresses. The only restriction in IPv4 is around not setting all of your host bits to 0 or 1. There are no restrictions regarding the number "255".
If I have to quibble with infographics like these, is that they obscure the fact that IPv4 was able to have variable length subnet masks, I.E. You could have networks that were 24 bits in length (/24) or 27 bits in length (/27) - depending on your requirements. With IPv6 you don't have that freedom - every network is a /64, so, if you are only going to use two hosts on your network (for a network link) - you are still assigning 2^64 addresses to that link, of which you will only use 2, resulting in a "wasted" 2^64 - 2 hosts.
It's odd the first time you do it, but eventually you just learn to live with the fact you have, for all practical purposes, unlimited address space in IPv6 - so wasting 2^32 times as many IP addresses than you had in ALL of IPv4 on just a single routed link turns out to be no big deal.
The reason, IMO, that IPv6 has not taken off is that it is far too complex. As a networking professional, it seems like an absolute nightmare, even if all equipment properly supported the protocol.
I still never understood why they couldn't just extend IPv4 by adding another octet, and then making all existing IPv4 addresses 0.{current_address}. This would give us about 1 trillion addresses and would make things dead simple.
As a networking professional you should understand that 4 octets = 32 bits. Five octets would = 40 bits, which is a pretty weird thing to be shuffling around all the time. It would require the same kinds of changes to hardware that IPv6 requires, so why not not go all the way to 128 bit so we don't ever have to worry about it again.
Ethernet addresses are 48 bits which is funky but we managed to survive. But I agree that if you're going to fork the Internet you might as well fix as much as you can without running afoul of the second system effect.
We've been using IPv6 for nine years now, and have about 17 million nodes. After you get over the hump of seeing fdc2:d343:216f:0001:0213:50ff:fe10:1234:5678 instead of 10.17.25.4, it really is more elegant in almost every way over IPv4.
The only glitch that we ran into in the last nine years was a (self created) configuration error, where we statically added default gateways on hosts/servers that pointed to the router. Those Router IP addresses had been auto-generated VLAN MAC addresses, which where then concatenated with the IPv6 network /64.
Unfortunately, the Cisco routers we were using, generated those EUI-64 addresses for the vlans based on the order they were created - and, after a power cycle, they were created in the order they appeared in the configuration - and all of a sudden our servers no longer had valid routes.
Lesson learned - don't use static gateways on an IPv6 network, let your routers advertise themselves, and, if for some reason (possibly a bad one), you really, really want a static gateway , make sure you don't rely on the auto-EUI-64 generation ability of your router.
Other than that - the protocol has been an easy drop-in replacement for IPv4.
The "Why" of not just adding 0. to IPv4 - is that IPv6 does much more than just increase address length. First of all, it eliminates CIDR - which is the source of many network misconfigurations. All networks in IPv6 are /64s. Further to that, you get stateless address autoconfiguration - nodes can now determine their own address without having to rely on a stateful registry somewhere. Just concatenate the network+mac address (or random number for the privacy sensitive), and you have a routable IPv6 address.
Broadcast is eliminated - and solicited nodes multicast is a very elegant replacement for services like neighbor discovery (rough equivalent of IPv4 ARP)
I could go on - but, I don't know very many IPv6 network operators who dislike the protocol.
40 comments
[ 4.0 ms ] story [ 100 ms ] threadIt took until 1995 to get agreement on what IPv6 would look like - but that's still 18 years ago. Plenty of time for a phased changeover :->
I have 1 public IP address and everything works.
We have an an IPv4 allocation problem, TONS of IPv4 space allocated to companies and not used. What we don't have is IPv4 usage problem, or the need for everything "on the internet" to be publicly addressable.
If we want proper peer-to-peer communications to be possible, we need to avoid, where possible, the need for a routable middle-man to help out.
Also, Asia Pacific has long run out of IPv4 addresses - back in 2011. There are now services starting up over there that are IPv6 only. This will spread over the next few years.
IPv6 is the future. There is no point complaining that we can shuffle around a few addresses here and there to make up a bit of space.
For all instant communication (VoIP, chat, file transfer, setting your thermostat), a man in the middle schema is a hindrance. The communication where you publish, and then somebody reads data need a man in the middle.
If your thermostat does not have a public IP, you must buy it as a service, and it'll never really be your property. One can live with that in a thermostat (but let's not pretend it's a good thing), just think twice before automating more of your home...
That is the very reason many people are nervous about the switch. Right now NAT/PAT give home users reasonable security without the need for a firewall. The obvious solution is just to make cheap firewalls that can route properly, and default to blocking all incoming traffic. However, that brings up the same concern of needing some way to dynamically poke a hole through the firewall. Which brings us to UPnP, and I for one do not trust that at all, but maybe thats my tinfoil hat.
I totally agree that we need to move forward with IPv6 though.
What advantage is IPv6 for most of my devices giving me?
And the advantage is that when you _do_ want to go direct you aren't stuck with being behind NAT on your router and then another NAT at your ISP, and then possibly another NAT level, and then the same in reverse at the other end, so that it's possible at all.
And sure, it might not give _you_ anything you particularly care about. But when you find that you can't do something, because there are several layers of indirection between you and someone else, and now you need to run your own server in order to hook two machines to each other over the internet, then you'll wish IPv6 was available.
Also, let's say you want to connect to your Time Capsule when you are on the go. Wouldn't it be nice to be able to connect to its public static IP address directly (assuming you set your firewall to allow such connections).
Lastly, as the Internet goes IPv6 only, your IPv4 only devices will get cut off. This will happen sooner than you might imagine.
Yes, there are large swatches of unused IPv4 addresses, but the demand for new allocations is much greater. Why cling to an old system just to offset its death by a few months when we can spend the effort on making the new system work better?
Let's start admitting that IPv6 offers no (or little) present day net benefits compared to the benefit of currently connecting to all of IPv4. And let's stop with the NAT booggy man argument. NAT works well, as much as any tech works and isn't perfect. And let's stop pretending that only thing stopping a peer 2 peer utopia is the lack of globally unique address. There many economic forces in play beyond that.
(edit spelling fix)
IPv6 does provide real world benefits, not the least of which is the simplification of allocations and a huge price drop in terms of address assignments.
Quite refreshing that, no matter how simplified or slightly erroneous the data is, it's just kinda hanging out there.
To summarize: we'll run out of IPv4 addresses, and they don't work interchangeably with IPv6. Oh well.
I like it : )
If I have to quibble with infographics like these, is that they obscure the fact that IPv4 was able to have variable length subnet masks, I.E. You could have networks that were 24 bits in length (/24) or 27 bits in length (/27) - depending on your requirements. With IPv6 you don't have that freedom - every network is a /64, so, if you are only going to use two hosts on your network (for a network link) - you are still assigning 2^64 addresses to that link, of which you will only use 2, resulting in a "wasted" 2^64 - 2 hosts.
It's odd the first time you do it, but eventually you just learn to live with the fact you have, for all practical purposes, unlimited address space in IPv6 - so wasting 2^32 times as many IP addresses than you had in ALL of IPv4 on just a single routed link turns out to be no big deal.
It needs routers to have native IPv6 support.
The real problem is getting ISPs to provide IPv6 support.
http://www.potaroo.net/tools/ipv4/index.html
I still never understood why they couldn't just extend IPv4 by adding another octet, and then making all existing IPv4 addresses 0.{current_address}. This would give us about 1 trillion addresses and would make things dead simple.
The only glitch that we ran into in the last nine years was a (self created) configuration error, where we statically added default gateways on hosts/servers that pointed to the router. Those Router IP addresses had been auto-generated VLAN MAC addresses, which where then concatenated with the IPv6 network /64.
Unfortunately, the Cisco routers we were using, generated those EUI-64 addresses for the vlans based on the order they were created - and, after a power cycle, they were created in the order they appeared in the configuration - and all of a sudden our servers no longer had valid routes.
Lesson learned - don't use static gateways on an IPv6 network, let your routers advertise themselves, and, if for some reason (possibly a bad one), you really, really want a static gateway , make sure you don't rely on the auto-EUI-64 generation ability of your router.
Other than that - the protocol has been an easy drop-in replacement for IPv4.
The "Why" of not just adding 0. to IPv4 - is that IPv6 does much more than just increase address length. First of all, it eliminates CIDR - which is the source of many network misconfigurations. All networks in IPv6 are /64s. Further to that, you get stateless address autoconfiguration - nodes can now determine their own address without having to rely on a stateful registry somewhere. Just concatenate the network+mac address (or random number for the privacy sensitive), and you have a routable IPv6 address.
Broadcast is eliminated - and solicited nodes multicast is a very elegant replacement for services like neighbor discovery (rough equivalent of IPv4 ARP)
I could go on - but, I don't know very many IPv6 network operators who dislike the protocol.