Choose strict, feasible, or loose uRPF for asymmetric routing paths

Use this document to resolve one bounded operational decision: Choose strict, feasible, or loose uRPF for asymmetric routing paths. Only the official source and traced locations below supply facts. Confirm applicability before acting.

Source fact:

The official RFC 3704 — Ingress Filtering for Multihomed Networks from RFC Editor / Internet Engineering Task Force supports the following bounded statements:

• Strict reverse-path forwarding accepts a source only when the packet arrived on the interface selected by the best forwarding path back to that source. The research record locates this support at Section 2.2 (Strict Reverse Path Forwarding).
• Feasible-path reverse-path forwarding also treats routing-protocol-derived alternative paths as valid, accommodating multihoming and asymmetric routing better than strict mode. The research record locates this support at Section 2.3 (Feasible Path Reverse Path Forwarding).
• Loose reverse-path forwarding checks only that some route to the source exists, which tolerates asymmetry but rejects only unrouted or martian sources when no route is present. The research record locates this support at Sections 2.4 (Loose Reverse Path Forwarding) and 4.1 (Use Loose RPF When Appropriate).

The source support ends with the statements listed above. Use them to examine address plans, interfaces, routes, peers, protocol roles, timers, middleboxes, and intended failure domains in the applicable environment, not to imply a wider guarantee.

What the source does not establish

This RFC evidence supports only the named network-protocol decision; it does not select vendor settings, topology, capacity, or an acceptable failure mode. Do not read the source as proof of implementation or permission to change production. Its guidance remains conditional on DNS, Active Directory authentication, PKI, time, routing policy, transport reachability, and monitoring and the environment’s recorded constraints.

Applicability questions

• For source statement 1 at Section 2.2 (Strict Reverse Path Forwarding), which observable configuration, record, or test can confirm applicability here?
• For source statement 2 at Section 2.3 (Feasible Path Reverse Path Forwarding), which observable configuration, record, or test can confirm applicability here?
• For source statement 3 at Sections 2.4 (Loose Reverse Path Forwarding) and 4.1 (Use Loose RPF When Appropriate), which observable configuration, record, or test can confirm applicability here?
• Within address plans, interfaces, routes, peers, protocol roles, timers, middleboxes, and intended failure domains, which versions, roles, and configuration states define the review population?
• Could DNS, Active Directory authentication, PKI, time, routing policy, transport reachability, and monitoring invalidate the test, hide a failure, or change applicability?
• Who owns the decision, and which observation requires stopping, escalation, or rollback?

DSE recommendation:

DSE recommends using the cited source as the evidence anchor for this decision. Make the source, asset scope, owner, and expected outcome explicit in the review record. Record the source location, examined part of address plans, interfaces, routes, peers, protocol roles, timers, middleboxes, and intended failure domains, observed and expected states, owner, and reason for deviation.

An implementation decision needs an owner, approved window, prechecks, observable outcome, stop authority, and rollback path. Validate DNS, Active Directory authentication, PKI, time, routing policy, transport reachability, and monitoring before and after the test, and store only sanitized operational evidence.

Verification and evidence

A reviewer should be able to retrace the decision from Section 2.2 (Strict Reverse Path Forwarding); Section 2.3 (Feasible Path Reverse Path Forwarding); Sections 2.4 (Loose Reverse Path Forwarding) and 4.1 (Use Loose RPF When Appropriate) through configuration snapshots, route or neighbor state, packet captures, counters, topology records, and controlled failover results. Record what was collected, where, when, by whom, and which system or role it represents.

Retain the starting state, authorization, execution record, outcome, deviation, and final state as one review package. Move disruptive checks to an approved test path. Reopen the decision when versions, design, dependencies, ownership, or official guidance changes.

Official references

• RFC 3704 — Ingress Filtering for Multihomed Networks — RFC Editor / Internet Engineering Task Force