Summary

  • A second carrier, invoice or circuit identifier does not demonstrate a second physical or logical path.
  • UK guidance tells providers and users to examine end-to-end dependencies, distinct geographic routes and points of presence rather than assume that supplier diversity creates resilience.
  • A circuit should count as a resilience control only after its shared failure domains are recorded at a disclosure level that can be maintained and tested.

The resilience file looks convincing. There are two contracts, two service identifiers and two support numbers. One circuit is sold by Carrier A and the other by Carrier B. The facilities manager colours the pair green.

Outside the building, both access tails enter the same chamber. They continue through the same street duct before separating deeper in the network. One excavation can therefore remove both services, despite the commercial separation on the invoices.

That scene is hypothetical, not a finding about a named provider. It describes the evidence gap that matters: the buyer has proved supplier plurality but has not yet proved path diversity.

What the contracts establish

Two service contracts can still have real value. They may create separate support channels, billing relationships, escalation paths, customer-premises equipment or contractual remedies. A second circuit may also add capacity or protect against a failure limited to one provider's service platform.

None of those benefits demonstrates that the circuits are independent at the failure domain that concerns the buyer. The access tails may use the same building exit. Wholesale arrangements may place both services on one underlying local network. Separate access routes may converge at one aggregation site. Two provider networks may depend on the same power, interconnection, field-maintenance contractor or upstream facility.

The claim being tested is therefore narrow: does one realistic failure remove both services? Supplier names alone cannot answer it.

What the UK guidance supports

Ofcom's current Network and Service Resilience Guidance addresses the architecture, design and operational models that underpin resilient public communications networks and services. Its detailed guidance treats responsibility as an end-to-end matter, including relevant third-party operations and interconnections. That framing matters because a retail product can inherit dependencies that sit outside the buyer's contract and outside the selling provider's own physical estate.

Cabinet Office guidance is even more direct for the purchasing problem. It warns that apparently different technical solutions can contain inherent dependencies. It also says that buying telecommunications services from more than one supplier may not improve resilience, because different suppliers can use shared infrastructure. For the last mile, it advises obtaining assurance that geographic cable routes and points of presence are physically separate; without that assurance, a buyer should not assume more than one path.

The Department for Science, Innovation and Technology separately summarises providers' resilience duties and Ofcom's role. The EC-RRG material adds established best-practice context, while expressly not presenting itself as current regulatory guidance. Together, these sources support a disciplined inquiry into dependencies. They do not certify any customer's particular circuits.

The shared-fate register

A buyer does not necessarily need a sensitive engineering map. It does need evidence at the level required to test the resilience claim. A useful shared-fate register can ask the following questions.

Layer Evidence question
Building access Do the services leave through different entry points, risers or internal pathways?
Access tail Are the local cables on physically distinct routes, or do they share a pole line, chamber or duct section?
Aggregation Do the paths reach distinct aggregation sites and points of presence before they can fail together?
Interconnection Do both services depend on the same handoff, exchange, carrier or upstream facility?
Power Are critical active elements exposed to the same grid supply, room, battery system or generator?
Operations Can one maintenance change, contractor action, configuration error or escalation failure affect both?
Recovery Has failover been tested under a scenario that removes the primary path rather than merely disabling an interface?

The answers do not have to be binary. A provider may confirm physical separation to a named boundary while withholding precise route coordinates. A wholesale reseller may only be able to attest to the access tail and not the upstream core. The correct response is to record the boundary of the assurance, not fill the unknown area with a resilience claim.

Diversity can decay

Path diversity is not permanent. Network rearrangements can move circuits onto common infrastructure. An acquisition can consolidate operations or points of presence. A planned maintenance change can create a temporary common path. Building works can force both cables through one entrance. A new managed-service layer can introduce a shared control or support dependency.

For that reason, the evidence has a date even though this article is not about ownership of dates. The date only records when the diversity statement was observed. It does not create or control the underlying resilience. The operating question remains whether the paths and dependencies are separate now, and whether a material change requires revalidation.

What buyers should not claim

The retained sources do not show that any named operator has sold falsely diverse circuits. They do not provide outage probabilities, prove the condition of a customer's route or establish that every shared component is unacceptable. Some services are deliberately designed with shared elements and can still meet the buyer's risk tolerance.

The sources also do not create a standard Ofcom route-diversity certificate. A shared-fate register and attestation are operational recommendations derived from the guidance. They should be presented as buyer controls, not as statutory forms.

The acceptance decision

The procurement record should distinguish three states.

First, additional service means a second commercial product exists, but its failure relationship to the first is unknown. Second, bounded diversity means separation is evidenced through a stated boundary, with remaining dependencies disclosed. Third, tested resilience means the claimed failover has been exercised against a scenario that removes the relevant primary path and the result has been recorded.

This vocabulary prevents a second invoice from silently becoming an engineering assurance. It also gives regional providers a fairer way to sell resilience: not as an absolute adjective, but as a documented scope of separation that can be priced, maintained and retested.

Two fibre circuits become a resilience control only when the route and shared-fate evidence supports that classification. Until then, they are two services with an unresolved common-mode risk.

Sources