The dynamic-PD problem
A network may be internally stable while its externally usable IPv6 prefix changes for reasons entirely outside that network.
1.1Where the instability enters
Residential and small-business Internet connections commonly obtain IPv6 space through DHCPv6 Prefix Delegation. The mechanism itself is not the problem: it gives a customer router enough address space to number multiple internal networks. The operational problem appears when the ISP-provided prefix is not persistent.
A reconnect, equipment replacement, service change, lease event, or provider-side reconfiguration may result in a different usable prefix. The LAN topology has not changed, but every internally used address derived directly from the old prefix has.
Addressing
Static host addresses and subnet identifiers inherit the external change.
Policy
Firewall objects, ACLs, monitoring targets, and allow-lists may become stale.
Naming
DNS records and documentation must follow the renumbering event.
Key observation
Nothing inside the network changed. Only the externally usable prefix changed.
1.2Why automation alone is incomplete
Dynamic DNS and configuration automation can reduce the labour associated with renumbering. They do not make the internal plan stable; they merely make repeated instability easier to process. Systems with embedded addresses, human-maintained rules, locally significant DNS, or cross-site policy may still experience avoidable churn.
1.3NPTv6 as the separation layer
NPTv6 creates a stateless, algorithmic mapping between an internal prefix and an externally usable prefix of corresponding length. Internal hosts retain stable addresses. When the ISP-side prefix changes, the edge mapping is updated while the internal topology, subnet plan, and host identifiers remain intact.
1.4Design objective
The following chapters examine two ways of selecting the stable internal prefix: ULA, and an intentionally non-routed GUA.