The modern Internet as we know it started January 1, 1983 when the NCP protocol was completely replaced by TCP/IP as the official communication protocol.  It was a time when computer ownership was rare and the ones connected to the Internet measured in the thousands.  The architects of TCP/IP created a 32-bit addressing scheme called IP version 4 (IPv4) with 4.3 billion Internet Protocol (IP) addresses which seemed more than sufficient at the time.  A decade later, it became obvious that 32 bits wasn’t going to be enough because of the rapid expansion of the Internet so the Internet standards body created a 128-bit addressing scheme called IPv6 (don’t ask me what happened to version 1, 2, 3, and 5).

The creators of IPv6 assumed that this shouldn’t be a problem because the Internet has made some major changes before.  When the entire Internet abruptly switched from NCP to TCP/IP in 1983, there were only 100 nodes and computers measuring in the thousands which made abrupt changes feasible.  In 1987, the entire Internet was dealing with congestion collapse due to the inadequate congestion control system in TCP/IP so everyone upgraded to a new version of TCP/IP with a fix devised by Van Jacobson.  But there were only around 30,000 computers compared to the billions today.  That was quite an accomplishment considering that people were voluntarily giving up their own network performance for the sake of the entire network while the attitude today is to grab as much bandwidth as possible at the expense of everyone else.

A more efficient congestion control protocol called Explicit Congestion Notification (ECN) with higher performance has yet to be adopted by the masses despite its ratification in 2001.  Even though the switch over to ECN was meant to be fully transparent to legacy systems which were supposed to ignore ECN, some legacy systems were incorrectly implemented and they freaked out when they saw ECN packets.  This forced Microsoft to disable ECN by default in their shipping Windows Vista product because it triggered bugs in some legacy home routers which resulted in customer complaints during the beta testing stage of Windows Vista.  That doesn’t portend good things for IPv6 which deliberate breaks legacy support when people won’t even adopt something designed to be as backward compatible as possible.

Why users don’t care about IPv6

Today, there are billions of computers or devices running IPv4 and none of them are facing the brutal congestion storms of the mid 1980s.  Hardly anyone is upgrading to IPv6 despite 15 years of encouragement to upgrade to IPv6 and we will likely run out of IPv4 addresses in two years.  When that day comes, nothing on the Internet would actually break other than the fact that the price of IP addresses will go up and the Internet registries will get stingy and possibly start reclaiming IP addresses.  Every home and business already on the Internet will keep working as is but new home and businesses customers will have harder time getting IP addresses.  With no doomsday looming for the existing user base, there is no benefit for that user base to adopt IPv6.

Benefits of IPv6 such as native Internet Protocol security (IPsec) are moot because IPsec has been fully ported to IPv4 for anyone who cares, and it turns out few people in the home or business actually care to deploy it on their computers.  What about the cleaner routing of IPv6?  Only nerdy network engineers like me care about that.  And eliminating the need for DHCP?  It turns out that DHCP is ubiquitous and well understood in the business and home and it became a nice centralized place to manage IP addresses.  Just about every broadband consumer has a router in the home and every router has built in DHCP that makes network addressing simple.  Need to figure out which infected computer is attacking the network?  Look up its name on DHCP.

Then there’s the supposed “benefit” of eliminating Network Address Translation (NAT) which broke the end-to-end architecture of the Internet and caused many problems for communication applications like VoIP, but it turned out to be a case of technology designers being out of touch with what their users wanted.  Not only do homes and businesses value NAT for its security benefits, they also value the ability to take a single IP address and share it with multiple computers which reduced the cost of their Internet connectivity.  As for the problems associated with NAT?  They’re so well understood that all the popular applications have worked around NAT and users don’t even need to think about it.  This user affection for NAT took the designers of IPv6 by surprise and it wasn’t until recently that the architectural purists within the IETF gave way to the pragmatists and started developing NAT standards for IPv6.

What happens when we run out of IPv4 addresses?

Home broadband users who get dynamically allocated IPv4 addresses might have to be allocated IPv6 Internet facing addresses and they’ll share an IPv4 address with many other homes.  That will introduce problems for some applications such as P2P or VoIP which will have to be upgraded, and it will no longer be possible to have a visible IPv4 presence unless a user buys a static IPv4 account.  This is already the norm for many wireless Internet services and it might be adopted in the home broadband market if ISPs run out of addresses.  But there are only a fixed number of homes (around 100 million in the United States) so it’s possible that each homes will continue to get real IPv4 as well as IPv6 addresses.

Businesses looking to obtain a fresh block of IPv4 addresses will have to buy or lease a block of existing addresses.  The Internet registries that manage IP allocation might have to start taking back massive blocks of class A with 16.8 million addresses or class B blocks with 65.5 thousand addresses that were allocated to individual companies at a time when the going was easy.  There’s simply no reason those companies need those IP addresses since they’re not using them on the Internet anyways because they’re using Network Address Translation (NAT) due to security considerations.  It’s hard to argue that a new company should be denied 32 addresses while another company sits on 16.8 million.  Those companies could easily be compensated for the migration with funds derived from the leasing of those IP addresses.

Adopting IPv6 doesn’t eliminate IPv4

The worst part of IPv6 is that implementing IPv6 still requires you to continue supporting IPv4 since there’s no interoperability.  This lack of backward compatibility is what killed the Itanium processor because consumers want to upgrade existing solutions over a dump everything and start over approach.  Furthermore, adding IPv6 visibility today causes a number of problems (Iljitsch van Beijnum has a great explanation of all the problems).  One example is that users experience lengthy delays reaching websites when they can get to it over IPv6 but they can’t.  Google resolved this by using DNS white listing where they will only advertise IPv6 addresses of their websites to ISPs that participate in the Google over IPv6 program, but this illustrates how fragmented and broken IPv6 is.

The architectural purists of IPv6 wanted a fresh start by eliminating compatibility with IPv4, but that clearly won’t fly with the billions of users on the Internet so now they’re starting to work on standards that allow IPv6 to talk to IPv4.  Until there is a simple and standardized transparent solution to make IPv4-only devices talk to IPv6-only devices and vice versa, I can’t see anyone in the home or business voluntarily deploying IPv6.  Even the government which continues to set more deadlines for incremental IPv6 capability won’t completely switch over.  With the eventual exhaustion of IPv4 even with aggressive reclamation and NAT, we will have to switch over.  That transition won’t start happening until the process is made simple.