3 ms·

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 equi

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.

zrail 13y ago
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.
- wmf 13y ago
  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.
pyvpx 13y ago
you can understand why "they" can't do that by doing a little reading on how those routers and switches you work with actually route IP packets.
ghshephard 13y ago
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.

zrail 13y ago

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.

wmf 13y ago
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.

pyvpx 13y ago

you can understand why "they" can't do that by doing a little reading on how those routers and switches you work with actually route IP packets.

ghshephard 13y ago

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.