IPv4 and IPv6 are the two versions of the Internet Protocol, the addressing scheme that identifies every device on a network and routes packets between them. IPv4, introduced in 1981, uses 32-bit addresses written as four dotted-decimal octets, such as 192.0.2.10, giving roughly 4.3 billion possible addresses. IPv6, specified in the late 1990s, uses 128-bit addresses written as eight groups of hexadecimal digits, such as 2001:db8::10, expanding the pool to about 340 undecillion addresses. The two are separate, non-interoperable protocols: an IPv6-only host cannot talk directly to an IPv4-only host without a translation or dual-stack mechanism sitting in between.
In practice the choice is rarely a clean either/or. IPv4's address space was effectively exhausted at the regional-registry level years ago, which is why carrier-grade NAT and private addressing became universal — and why IPv6 was created in the first place. Yet IPv4 still carries the majority of traffic to many destinations, and a great deal of installed equipment, firmware, and third-party services speak only IPv4. Most organizations therefore run both at once (dual-stack) rather than switching wholesale, and the real question during a network refresh is how far to lead with IPv6 while keeping IPv4 reachable for everything that has not yet caught up. For U.S. federal agencies that calculus is set by policy: OMB Memorandum M-21-07 directs agencies to move most networked assets to IPv6-only, with 80 percent targeted by the end of fiscal 2025.
At a glance
Side by side
| Factor | IPv4 | IPv6 |
|---|---|---|
| Address length | 32-bit — about 4.3 billion addresses | 128-bit — about 340 undecillion addresses |
| Notation | Dotted decimal, four octets (192.0.2.10) | Hexadecimal, eight groups (2001:db8::10) |
| Address configuration | Static or DHCP; ARP resolves MAC addresses | SLAAC or DHCPv6; Neighbor Discovery replaces ARP |
| Header | Variable 20–60 bytes, with options and a header checksum | Fixed 40-byte base header, no checksum, optional extension headers |
| Address delivery types | Unicast, multicast and broadcast | Unicast, multicast and anycast — no broadcast |
| NAT | Widely used to stretch a scarce address pool | Not needed for addressing; end-to-end reachability restored |
| Fragmentation | Performed by the sending host and by routers along the path | Performed by the source host only, via Path MTU Discovery |
| DNS record type | A record | AAAA record |
Choose IPv4 when
- The endpoints, firmware, or upstream ISP path only speak IPv4 and cannot be upgraded in the near term
- You are extending a small, isolated, or air-gapped segment where the larger address space delivers no practical benefit
- A required third-party service, appliance, or SaaS endpoint still publishes no IPv6 (AAAA) address
- Operational tooling, monitoring, and staff familiarity are built around IPv4 and a phased transition is safer than a cutover
Choose IPv6 when
- Public-facing services must remain reachable to clients and mobile networks that are increasingly IPv6-only
- A federal IPv6-only mandate applies and networked assets must meet the OMB M-21-07 transition targets
- You are addressing at large scale — IoT fleets, dense sensor or device deployments — where NAT adds fragility
- You are building greenfield and want end-to-end addressing without carrier-grade NAT between you and the internet
Bottom line
Neither protocol is simply better; they solve the same problem at different points in a long transition. IPv6 is where the internet is heading — its vast address space ends the scarcity that forced NAT, and federal mandates now require it on public-facing and internal assets alike. But IPv4 is not going away on any near horizon: too much equipment, too much software, and too many external services still depend on it. For nearly every organization the practical answer is dual-stack — run both, lead with IPv6 for new and public services, and keep IPv4 reachable for everything that has not yet followed. Let the reachability of the systems you actually need, not a preference for the newer number, decide how fast you move.
Shop it at Uniqcli
FAQ
Common questions
- Is IPv6 faster than IPv4?
- Not inherently — throughput is dominated by bandwidth, latency, and congestion, not by protocol version. IPv6's simpler fixed header and its removal of router-based fragmentation can make forwarding marginally more efficient, and avoiding carrier-grade NAT can cut some overhead, but on a healthy dual-stack network users rarely see a meaningful speed difference. Any gap usually comes from the path traffic takes, not from the protocol itself.
- Can IPv4 and IPv6 talk to each other directly?
- No. They are separate protocols with incompatible address formats, so an IPv6-only host cannot reach an IPv4-only host without help. The common solutions are dual-stack, where a device runs both at once and picks whichever the destination supports, and translation gateways such as NAT64 with DNS64, which map traffic between the two. Tunneling can also carry one protocol across a network that only routes the other.
- Do I have to switch to IPv6 now?
- For most private networks there is no hard deadline — dual-stack lets you adopt IPv6 gradually while IPv4 keeps working. U.S. federal agencies are the exception: OMB Memorandum M-21-07 sets targets to move the majority of networked assets to IPv6-only over several fiscal years. Even without a mandate, enabling IPv6 on public-facing services is increasingly worthwhile as more clients arrive over IPv6-only paths.
- Does IPv6 remove the need for a firewall?
- No. IPv6 restores the end-to-end addressing that NAT previously hid, but NAT was never a security control — a stateful firewall is. Because every host can hold a globally routable address, an explicit inbound policy becomes more important, not less. Plan IPv6 firewall rules, prefix delegation, and segmentation deliberately rather than assuming that changing the address format protects anything on its own.