Guidelines for OSI NSAP Allocation in the Internet
obsoleted by
- rfc-1629 — Guidelines for OSI NSAP Allocation in the Internet
Extracted elements (25)
Areas are permitted multiple area addresses to support graceful transition operations: renaming (assign B, then remove A), merging two areas (add each other's address then unify), and partitioning one area into two (introduce new addresses before removing the old one). This avoids hard cutover disruptions.
If areas within a routing domain draw their addresses from unrelated allocation authorities, no common prefix exists for summarization at routing domain boundaries. This is analogous to flat routing in the IP Internet and cannot scale to internets with tens or hundreds of thousands of routing domains.
Multi-homed routing domains face a fundamental tradeoff: a single independent prefix (solution 1) allows one internal addressing scheme but forces all worldwide TRDs to maintain an explicit routing entry; per-connection prefixes (solution 2) scale better but require address changes when connectivity changes and do not automatically maintain backup routes.
NSAPs must be assigned along topological rather than purely administrative or organizational boundaries. Administrative and organizational boundaries often diverge from network topology, and misalignment prevents routing data abstraction, resulting in flat inter-domain routing that does not scale to very large internets.
Private point-to-point links between two routing domains may be ignored for NSAP allocation purposes because their routing information is not propagated throughout the internet. A domain with one public TRD attachment plus multiple private links is treated as single-homed for NSAP allocation.
Regionals should not derive their NSAP prefix from a backbone's address space because regionals and backbones are independent—regionals may change backbone providers (analogous to switching long-distance carriers). Tying a regional's addresses to a backbone prefix is structurally inconsistent with this independence.
A third multi-homing solution assigns a single prefix from one TRD and selectively propagates the route to a limited set of other TRDs rather than globally. This is particularly appropriate for special-interest group backbones whose routing prefixes should not be advertised to the general internet.
Address transition when network topology changes must be gradual rather than instantaneous. During the transition period both old and new addresses must remain accepted by end systems and both must result in correct packet delivery. DIS10589's support for multiple area addresses per area is specifically designed to enable this.
For non-U.S. parts of the Internet, local NSAP formats based on national DCC assignments may be preferable to GOSIP Version 2 format. The technical principles of hierarchical NSAP assignment and routing data abstraction apply equally regardless of which IDP format is used.
All TRDs should be explicitly involved in address administration for single-homed leaf routing domains attached to them. This reduces inter-domain routing overhead and is recommended as a valuable service offered by the TRD to its customers.
Each TRD should develop explicit policy on whether to accept non-local NSAP address prefixes from attached leaf routing domains and how such prefixes will be treated in routing exchanges with other TRDs. A conservative policy accepts non-local prefixes but does not advertise them to other TRDs.
For zero-homed and single-homed routing domains expected to remain so, all NSAP addresses within a single routing domain should use a single address prefix assigned to that domain, enabling the full set of reachable addresses to be described by that one prefix.
It is strongly recommended that addresses in the U.S. Internet be assigned hierarchically based on prefixes assigned to individual TRDs. For GOSIP addresses, AA identifiers should be assigned to all TRDs; single-homed leaf routing domains should receive an RD value from their TRD's address space.
The area address (concatenation of IDP and HO-DSP) must be globally unique. If the area address of an NSAP matches one of the area addresses of a system, the destination is considered to be in that system's area and is routed to by level 1 routing.
The ID field must be unique within an area for ESs and level 1 ISs, and unique within the routing domain for level 2 ISs. The ID field is treated as flat (non-hierarchical) by the routing protocol.
Within a routing domain conforming to DIS10589, the lower-order octets of the NSAP should be structured as ID and SEL fields. End systems with non-conforming addresses may require additional manual configuration and be subject to inferior routing performance.
Zero-homed routing domains (internally networked organizations with no public internet connection) must use globally unique NSAP addresses. Even across private links, the connected RD must be able to unambiguously distinguish the zero-homed domain's NSAPs from all other NSAPs it routes to.
DIS10589 allows each area to have up to a maximum of three area addresses (an architectural constant per Clause 7.2.11 and Table 2 of Clause 7.5.1). This enables graceful area renaming, merging, and partitioning by temporarily maintaining multiple addresses before removing old ones.
DIS10589 implements two-level hierarchical routing within a routing domain. Level 1 ISs know only the topology within their area and route to the ID field; Level 2 ISs know the level 2 backbone topology and route to address prefixes using longest-match, preferring internal routes over external routes.
End systems can autoconfigure their area address by extracting it from a neighboring IS's ISO9542 IS-Hello packets. If the ID portion uses IEEE-style burned-in hardware identifiers, an ES can reconfigure its entire NSAP automatically on topology change; routers still require manual area address reconfiguration.
Level 1 ISs forward packets for destinations outside their area to the nearest level 2 IS that indicates it is attached to the level 2 backbone. Level 1 ISs refuse to become neighbors with ISs whose area addresses do not overlap their own.
OSI NSAP addresses are structured as the Initial Domain Part (IDP) and Domain Specific Part (DSP). DIS10589 further subdivides the DSP into High Order DSP (HO-DSP), System Identifier (ID), and NSAP Selector (SEL). The combination of [IDP, HO-DSP] constitutes the area address.
The ID field may be one to eight octets in length but must have a single known length within any particular routing domain. The SEL (NSAP Selector) field is always exactly one octet and is always the last octet of the NSAP address.
The requirement for globally unique NSAP addresses (even for zero-homed and private routing domains) is a foundational security property: without global uniqueness, a routing domain cannot reliably distinguish packets addressed to it from those addressed to another domain with a colliding address, undermining correct packet delivery and routing isolation.
GOSIP Version 2 NSAP structure: AFI (1 octet, value 47), IDI (2 octets, value 0005), DFI (1 octet, value 80h), AA (3 octets), Rsvd (2 octets), RD (2 octets), Area (2 octets), ID (6 octets), SEL (1 octet), totaling 20 octets. AFI+IDI+DFI+AA = administration prefix; adding Rsvd+RD = routing domain prefix; adding Area = area address.