The most visible—and most frequently cited—difference between IPv4 and IPv6 is the expansion of the address space from 32 bits to 128 bits. But this quantitative change triggered a cascade of qualitative transformations that fundamentally altered how we think about IP addressing.

IPv4's 4.3 billion addresses (~2³²) seemed astronomically large in 1981. Today, with over 20 billion connected devices worldwide and projections of 75+ billion by 2025, that address space is not merely exhausted—it has become a constraining factor requiring complex workarounds like NAT, CGNAT, and address market trading.

IPv6's 340 undecillion addresses (~2¹²⁸) represent such an enormous expansion that new addressing paradigms become possible. Every device can have multiple globally unique addresses. Subnets can be profligate. Address conservation becomes unnecessary. This abundance fundamentally changes network architecture.

Numbers at the scale of 2¹²⁸ are difficult to conceptualize. Let's put the IPv4 and IPv6 address spaces in perspective with concrete comparisons:

Why 128 Bits?

The choice of 128 bits for IPv6 was deliberate and carefully reasoned:

• 64 bits for network prefix: Allows for an effectively unlimited number of networks (2⁶⁴ possible /64 subnets)
• 64 bits for interface identifier: Enables automatic address generation from MAC addresses (EUI-64) or privacy extensions
• Hierarchical allocation: Enables deep hierarchy for routing efficiency (RIR → LIR → Enterprise → Site → Subnet → Interface)
• Future-proofing: Even with wasteful allocation, exhaustion is effectively impossible

The design explicitly prioritizes simplicity and aggregation over address conservation. Unlike IPv4 where address conservation is a constant concern, IPv6 encourages liberal allocation for cleaner routing tables.

The notation systems for IPv4 and IPv6 reflect their different scales and usage patterns. Understanding both notation styles is essential for network configuration, troubleshooting, and documentation.

Both IPv4 and IPv6 categorize addresses by their delivery semantics: how many destinations receive packets sent to that address, and how those destinations are selected. However, the specific types and their implementations differ significantly.

Anycast: From Informal to Formal

In IPv4, anycast is implemented informally by assigning the same IP address to multiple servers and relying on routing to direct packets to the nearest instance. This works but isn't explicitly supported by the protocol.

IPv6 formally recognizes anycast:

• Syntactically identical to unicast: Anycast addresses cannot be distinguished from unicast by their format
• Subnet-router anycast: Every subnet has an implicit anycast address (all-zeros interface identifier) pointing to all routers on that subnet
• Application-defined anycast: Services can allocate anycast addresses for load distribution (e.g., root DNS servers)
• Routing-based selection: Packets reach the topologically nearest instance based on routing metrics

Key Constraint: A source address cannot be anycast—senders must use unicast addresses for their source address field.

IPv6 introduces a formal concept of address scope—the topological span within which an address is unique and routable. This scope hierarchy enables more nuanced and secure network architectures than IPv4's binary public/private distinction.

Multicast Scope Field

IPv6 multicast addresses (ff00::/8) include an explicit 4-bit scope field in their structure:

ff0s::/16 where s = scope value

ff01::/16 - Interface-local (node-internal)
ff02::/16 - Link-local (same link)
ff05::/16 - Site-local (deprecated with unicast site-local)
ff08::/16 - Organization-local (defined by policy)
ff0e::/16 - Global (Internet-wide)

This structure enables routing-enforced scope boundaries. A router receiving a link-local multicast packet (ff02::) knows it must not forward that packet to other links—the scope is encoded in the address itself.

Contrast with IPv4: IPv4 multicast has no standardized scope field. TTL is sometimes used as a proxy for scope, but this is a weak and overloaded mechanism.

Both protocols reserve certain address ranges for special purposes. The following table provides a comprehensive comparison of these reserved addresses:

How hosts obtain IP addresses differs significantly between IPv4 and IPv6. IPv6 introduces stateless autoconfiguration as a core feature, reducing dependence on centralized servers.

SLAAC: Stateless Address Autoconfiguration

SLAAC is a fundamental IPv6 capability that enables hosts to self-configure globally reachable addresses without any central server:

1. Host generates link-local address from its MAC address (EUI-64 format) or randomly (privacy extensions)
2. Host performs Duplicate Address Detection (DAD) by sending Neighbor Solicitation to its own address
3. Host listens for Router Advertisements (RAs) which contain network prefix(es)
4. Host combines prefix with interface identifier to form global unicast address(es)
5. Host validates global address with DAD before activation

Result: A host can join a network and obtain a globally routable address in seconds, with zero manual configuration or server infrastructure.

One of the most significant conceptual shifts from IPv4 to IPv6 is the expectation of multiple addresses per interface. In IPv4, a host typically has one address per interface (with exceptions for virtual hosts and aliases). In IPv6, multiple addresses per interface is the norm, not the exception.

Practical Example: Typical Desktop Interface

eth0: flags=4163<UP,BROADCAST,RUNNING,MULTICAST>
      inet6 2001:db8:1:1::10          scope global        # Static global
      inet6 2001:db8:1:1:a2c3:d4e5:f6g7:h8i9  scope global dynamic  # SLAAC
      inet6 2001:db8:1:1:temp::1       scope global temporary  # Privacy temporary
      inet6 fe80::1a2b:3c4d:5e6f:7a8b  scope link          # Link-local

This host has four IPv6 addresses on a single interface—all simultaneously valid and usable:

• Static global: Manually configured, used for stable services
• SLAAC dynamic: Auto-configured from Router Advertisement
• Temporary: Privacy extension address, rotated periodically
• Link-local: For neighbor discovery and local communication

The structure of IPv6 addresses was designed with hierarchical allocation and route aggregation as primary goals. This differs markedly from IPv4's historically fragmented allocation.

The /64 Boundary: A Fundamental Constraint

IPv6 mandates that subnet prefixes must be /64. This is not merely convention—it's a protocol requirement that enables several features:

• SLAAC compatibility: Hosts need 64 bits for the interface identifier portion
• EUI-64 format: MAC addresses expanded to 64 bits fit exactly in the interface ID
• Privacy extensions: Random 64-bit interface IDs require the full space
• Solicited-node multicast: Computed from the low-order 24 bits of the interface ID

Practical Impact:

• A /48 allocation provides 2¹⁶ = 65,536 /64 subnets
• A /56 allocation provides 2⁸ = 256 /64 subnets
• Even small organizations receive enough subnets for any conceivable expansion

The addressing differences between IPv4 and IPv6 extend far beyond the simple increase in address length. They represent a fundamental reimagining of how network identification and communication should work at Internet scale.

What's Next:

With addressing differences understood, we'll next examine how security differs between IPv4 and IPv6—from built-in IPsec requirements to new attack surfaces introduced by IPv6's features and the security implications of running dual-stack networks.