Place shared addresses on the provider link

Shared Address Space numbers provider-side links; a CGN can translate beyond that scope.

• CGN means Carrier-Grade NAT, a provider-operated address translator. CPE means Customer Premises Equipment, such as a subscriber’s home router. RFC 6598 reserves a /10 block to number interfaces connecting CPE to CGN devices (§§1,7). It is an address-allocation and operational-scope document, not a complete NAT implementation specification.
• Shared Address Space is non-globally routable and distinct from RFC 1918 private space (§1). Its use is limited to Service Provider networks or routing equipment that can translate across interfaces even when addresses on two interfaces are identical. It is not an unrestricted replacement for every private LAN prefix.
• The conceptual example starts with host 192.168.1.7 behind a home CPE. The CPE outside interface is numbered 100.64.1.2 in the shared block. A provider CGN can translate to an appropriately assigned public-side address. Port mappings, checksums and NAT state are outside this allocation guide.
• The drawing uses 203.0.113.9 only as a documentation placeholder for a provider’s legitimately assigned public-side address. That example address is not itself globally routable. The source-address labels illustrate distinct scopes, not a mandated address assignment or a complete packet trace.
• A remote peer cannot treat the CPE’s shared address as an ordinary globally reachable endpoint. A public-looking address is also not a proof of reachability: §5.1 notes that even a globally unique address behind a CGN may not be globally advertised.
• The IESG note treats this allocation as an operational expedient while retaining the commitment to IPv6 deployment. It does not promise that an additional IPv4 block removes address-sharing problems or makes every application work.