Why is IPv6 so complicated? A former IETF chairman explains.

On April 14, 2026, a draft of IPv8 (Internet Protocol Version 8), the next-generation network protocol, was submitted to the IETF (Internet Engineering Task Force), the standardization body for internet protocols. As of the time of writing, the draft has not yet been approved by the IETF, but IPv8 has attracted attention because, unlike IPv6, it aims for complete backward compatibility with IPv4. In this context, Brian E. Carpenter, a former CERN engineer and former IETF chairman during the development of IPv6, has provided an explanation of the complexity of IPv6.

misc/why6why.md at main · becarpenter/misc
https://github.com/becarpenter/misc/blob/main/why6why.md

The very reason IPv6 was proposed was the depletion of IPv4 addresses. ICANN allocated all IPv4 addresses to its regional internet registries (RIRs) in 2011. From that point onward, it became impossible to obtain any new unallocated IP addresses.

Fourteen years after it was said that 'IPv4 addresses have been exhausted,' what has actually become of IPv4 in 2025? - GIGAZINE

Subsequently, IPv4 addresses were exhausted in North America in September 2015, and by November 2019, all IPv4 addresses in Europe , including reserves, had been used up . In response, a global transition to IPv6 is underway.

Carpenter stated, 'There is no doubt that IPv6 is more complex than IPv4, and some people will wonder why that is. Some argue that it would have been much simpler to just add 32 bits to the IPv4 address and not change anything else.' He then mentioned that alternatives to IPv6, such as this IPv8 draft, are regularly submitted.

However, Carpenter argues that 'such proposals are a waste of time for everyone,' and suggests the following three answers:

1: Adding bits to an address is not as simple as it looks.
Carpenter first explains that the idea that 'it should have been as simple as adding 32 bits to the IPv4 address' is actually not feasible. IPv4 implementations, both in 1994 and today, are based on a 32-bit address format, so if you change the address length to 33 bits, 64 bits, or 128 bits, existing IPv4 implementations will discard that packet. Therefore, expanding the address length ultimately requires changing the protocol itself, which necessitates a new version number and new code to handle it.

Moreover, introducing a new protocol doesn't mean that all the old IPv4 devices will disappear at once. Therefore, a mechanism is needed for the old and new versions to be interconnected, but Carpenter says there are practically only two ways to do this. One is dual-stack, where the new devices support both IPv4 and the new protocol, and the other is translation, where the address and protocol are converted somewhere. In other words, the complexity associated with the coexistence and transition of IPv4 and next-generation IP is not so much a result of IPv6's unique design, but rather an unavoidable condition when bringing a new IP that exceeds 32 bits into the existing internet.

Furthermore, Carpenter stated that while the idea of simply adding a few bits to the beginning of IPv4 for IPv6 was actually tried, the ' IPv4-Compatible IPv6 address ' proved largely useless for coexistence or migration and was later abandoned. The related ' IPv4-Mapped IPv6 address ' also retained a role in the POSIX socket API, but it was not a definitive solution. Technologies like 6to4 and Teredo were also used, but these ultimately remained only temporary migration techniques.

In addition, Carpenter states that the current IPv4 itself is no longer a simple protocol, as it incorporates many mechanisms such as NAT, carrier-grade NAT, firewalls, IPsec, VPNs, Differentiated Services, link-local addresses, and CDNs.

2: In 1994, IPv4 was not the only network layer protocol in the world; other protocols had superior features that IPv4 lacked.
Carpenter then states that to understand IPv6, we need to look at the situation in the early 1990s. At that time, the internet had not yet conquered the world, and the World Wide Web hardly existed before 1993. Many network layer protocols other than IPv4 were in use, and especially among governments and large corporations, it was widely believed that the OSI protocol suite, the official international standard, would be the future standard. Various other proprietary protocols also coexisted, and the IETF had presented only several IPng proposals that had not yet been finalized in terms of implementation.

In this situation, what was needed in the next-generation IP was not simply IPv4 with a larger address range. Other existing protocols had several convenient features that IPv4 lacked, and IPng also needed to have at least some new features to meet expectations.

Carpenter explains, 'Looking back, it may have been an unfortunate situation, but the reality at the time was that IPng had to be better than IPv4, better than DECnet and Netware , and above all, better than the OSI protocol suite.' In other words, the fact that IPv6 didn't become 'just IPv4 with a larger address range' was largely influenced by the technological competitive environment at the time.

3: Have the IPv6 designers gone mad?
The third reason Carpenter cites is the view that 'the IPv6 designers went crazy and over-engineered it.' However, Carpenter himself does not necessarily support this view.

Rather, the question that should be re-examined is whether IPv6, as a so-called Second System Syndrome, or 'successor to the first successful system,' has fallen into the trap of being over-designed. To this, Carpenter replies, 'Not really.' He assesses IPv6 as a fairly conservative design, as it does not change the basic IP model of packet switching that does not require connections and topological addresses.

Furthermore, IPv6 added flow labels, refined the fragmentation mechanism, and replaced IPv4's 'options' with IPv6's 'extension headers.' Even more importantly, it introduced Stateless Address Autoconfiguration (SLAAC) , its closely linked Router Advertisements (RAs), and Interface Identifiers (IIDs) as part of the address.

SLAAC was inspired by DECnet, Netware, and AppleTalk , which did not require manual address configuration. The reason why it appears to have some overlap with DHCPv6 is that at the time of design, DHCP itself was still new and not well established, and DHCPv6 was added later. Carpenter stated that these changes were by no means meaningless additions and were not the main cause of the problems in IPv6 deployment, and that most of the problems still arose from the coexistence of IPv4 and IPv6.

However, it wasn't entirely without confusion. Carpenter reflects that the decision to make IPsec support mandatory for IPv6 was a politically motivated tie-in of a technology that was still immature at the time, and as a result, it hindered the widespread adoption of IPv6.

Furthermore, Carpenter warns that many of the 'IPv8' type proposals that have repeatedly appeared in later years have actually worsened the situation. For example, he says that addresses with embedded geographic information, prefixes based on AS numbers , and addresses with meaning in the bit sequence itself, while seemingly easy to understand, pose risks of breaking internet routing, making it even more difficult to change site addresses, causing another shortage of addresses, and facilitating widespread surveillance. He also points out that while geographic addresses may seem useful when location information is needed, they are incompatible with the internet's inter-domain routing mechanism.

Furthermore, Carpenter argues that even if a new proposal other than IPv6 were to be seriously adopted now, it would still take a long time to become widespread. He points out that it took more than 25 years for IPv6 to reach a penetration rate of approximately 50%, and the situation would be no different with a different proposal.

In fact, the retirement of Frame Relay , the replacement of ATM , the deployment ofDNSSEC , and the deployment of RPKI have all progressed across the internet not in years, but in decades. Carpenter states that a complete overhaul of the network infrastructure is such a time-consuming task.

Finally, Carpenter concludes that the primary and sole true reason for IPv6's existence is 'larger addresses.'

The coexistence of IPv4 and IPv6 is unavoidable, and any new IP address exceeding 32 bits will create the same kind of transition problems we have experienced before. Dual-stack architecture and protocol-to-address translation were mathematically unavoidable. Therefore, Carpenter argued that it is more constructive to understand IPv6 based on these premises and operate it realistically, rather than expecting the existence of a 'simpler next-generation IP' separate from IPv6.