Summary

  • The base special-purpose space contains values 0 through 15. Label 15 is the XL: it cannot be bottom of stack and gives special meaning only to the next label.
  • The IETF and IANA registry establish interoperable meaning. A sender selects an already defined value; it cannot invent meaning, establish an LSP or route, authenticate an identity, grant access, or override receiver support and policy.
  • An unsupported top-of-stack XL or an unknown following eSPL must be dropped. Logging is optional and must be rate-limited.
  • eSPLs are not a load-balancing mechanism. Legacy parsing can split one flow across paths and reorder packets.

Sixteen base values are scarce because special-purpose processing is not merely a larger ordinary label table. An implementation may need forwarding-hardware behavior that is expensive to change or impossible to add. RFC 7274 therefore created an extended space and set Standards Action as the allocation policy, with early allocation possible case by case. It did not reserve base values for experimental or private use: the base space is too small for that promise.

The compound rule is exact. An XL must be followed by another label and cannot be bottom of stack. Only the immediately following label receives the eSPL interpretation. Under the RFC 9017 update, values 0 through 15 must not appear immediately after an XL; an LSR seeing that sequence at the top of stack must drop the packet. An LSR that does not understand a top-of-stack XL, or that sees an unknown following eSPL, must also drop the packet. It may log the event, but logging must be rate-limited.

This saves scarce base values where minimizing stack size matters, but the saving is not free. The compound form consumes an additional stack entry and can meet legacy parsers, reordering hazards, fragmentation, or implementation limits. RFC 7274 warns that a larger stack can make packets unprocessable by some implementations. The standards do not establish how widely current hardware supports every eSPL, how often these failures occur, or whether a registered value is implemented on every path.

Do not confuse namespaces. RFC 6790 adds entropy for load balancing; eSPLs must not be used for load balancing. RFC 3443 addresses TTL propagation and tunnel visibility, while RFC 4950 discloses an arriving label stack in ICMP. Neither changes the XL parsing authority described here. RFC 9017 standardizes current terminology—Base Special-Purpose Label, Extension Label and Extended Special-Purpose Label—and clarifies handling; it does not establish an LSP or alter forwarding through a control plane or centralized controller. “Reserved labels” is historical terminology, not the current generic name for allocated base SPLs.

Retirement is intentionally slow. Deprecation requires IETF consensus and an RFC. A survey may begin after twelve months; release for future assignment can be requested only after the defined no-use finding and at least twenty-four months. Registry allocation is not operator permission to inject a label, a deployment guarantee, or a commercial support commitment.

Verification fixtures

  1. Parse 15, eSPL-X as XL followed by the immediately interpreted eSPL; never treat XL as bottom of stack.
  2. Put 15 at top of stack with no following label: drop.
  3. Put 15, 0 through 15, 15 at the top: drop under RFC 9017.
  4. Use an unsupported top-of-stack XL or unknown following eSPL: drop; any log must be rate-limited.
  5. Run the same flow through a legacy parser fixture and verify that eSPL is not selected as a load-balancing input.

Operator decision path

First identify the registry value and the required implementation behavior. Next confirm that every relevant device and path policy supports it; do not infer that from allocation alone. Then test stack depth, fragmentation exposure, legacy parsing and fail-closed behavior. If support is absent or uncertain, do not inject the label: choose a supported design or obtain the relevant operational authorization. Finally monitor drops and rate-limited diagnostics without claiming that the standards predict a current incident rate.

Sources