IPv6 Features

IPv6 is not just 'longer addresses'. It's a redesign of the network layer: autoconfiguration without DHCP, no broadcast, mandatory multicast, extension headers instead of options. Understanding these differences is critical for a modern network engineer.

• **Mobile operators:** T-Mobile, Verizon, Jio use IPv6-only with NAT64. Your app must work in such an environment
• **Cloud native:** Kubernetes supports dual-stack. Service mesh (Istio) actively uses IPv6 for sidecar communication
• **IoT:** Billions of devices get global IPv6 addresses via SLAAC without needing DHCP infrastructure

Предварительные знания

• IPv6: The Future Is Already Here

SLAAC and DHCPv6

IPv6 has two ways to automatically obtain an address: **SLAAC** (Stateless Address Autoconfiguration) and **DHCPv6**. They can work together, complementing each other.

**RDNSS:** RFC 8106 added DNS to Router Advertisement. Now SLAAC can work entirely without DHCPv6. But not all OSes support RDNSS.

DHCPv6 works differently from DHCPv4:

Neighbor Discovery Protocol (NDP)

**NDP (Neighbor Discovery Protocol)** replaces ARP and does much more: router discovery, autoconfiguration, neighbor reachability detection, duplicate detection.

**Neighbor Solicitation/Advertisement** - analogous to ARP request/reply:

**NUD (Neighbor Unreachability Detection):** NDP periodically checks that a neighbor is still alive. If a host doesn't respond - the entry is deleted. ARP has no such mechanism.

ICMPv6 - The Heart of IPv6

**ICMPv6** is critically important for IPv6 to function. Unlike IPv4, where ICMP can be partially blocked, blocking ICMPv6 **breaks the network** - NDP, PMTUD, and SLAAC all stop working without it.

**Packet Too Big:** In IPv6, routers do NOT fragment packets (only the sender can). Without ICMPv6 Type 2, Path MTU Discovery is impossible - large packets are silently dropped.

The IPv6 header is simpler than IPv4 - rarely used fields were removed:

**Extension Headers** replace Options and are processed only by the end host (not routers).

Dual-Stack and Happy Eyeballs

**Dual-Stack** - the main transition strategy: a device has both IPv4 and IPv6 addresses simultaneously. Applications choose the protocol based on availability.

**Problem:** If IPv6 exists but is broken (misconfiguration, firewall), the connection will hang for 20-30 sec before falling back to IPv4.

**Happy Eyeballs (RFC 8305)** solves this problem - it tries both protocols in parallel:

All modern browsers and OSes implement Happy Eyeballs. The fallback delay is configurable (usually 200-300ms).

Tunneling and NAT64

When dual-stack is not possible, **tunneling** (IPv6 inside IPv4) or **translation** (NAT64/DNS64) is used. This allows IPv6-only clients to communicate with IPv4-only servers.

**HE.net Tunnel Broker:** A free service for getting IPv6 via tunnel. Useful if your ISP doesn't provide IPv6.

**NAT64 + DNS64** - the modern solution for IPv6-only networks (mobile operators):

Key Ideas

• **SLAAC** lets hosts get an address automatically from Router Advertisement without a server. DHCPv6 supplements with DNS and other options
• **NDP** replaces ARP and adds: Router Discovery, SLAAC, DAD, Neighbor Unreachability Detection. Uses ICMPv6
• **ICMPv6** is critically important - must not be blocked! Without Packet Too Big, PMTUD breaks; without NDP messages - the entire network fails
• **Happy Eyeballs** ensures fast connections in dual-stack. **NAT64/DNS64** connects IPv6-only clients with IPv4 servers

Related Topics

IPv6 features relate to core concepts:

• IPv6 Introduction — Basic address formats and types (GUA, link-local, ULA)
• ARP — NDP replaces ARP in IPv6, adding new capabilities

Вопросы для размышления

• Why did IPv6 designers decide not to carry broadcast over from IPv4?
• How does Happy Eyeballs affect performance for users with unstable IPv6?
• What security issues does NDP create without SeND (Secure NDP)?

Связанные уроки

• sec-01