Summary

  • NIST measured entanglement distribution over 62 kilometres of difficult commercial fibre at 1,500 entangled photons per second and 92.8% availability during a 24-hour run. That proves a demanding link result, not a deployed quantum application.
  • RFC 9583 is an Informational IRTF catalogue and dependency map. Its six stages do not supply dates, costs, comparative advantage or a receipt that an application produced the promised outcome.

The photons made it from Gaithersburg to College Park. That sentence is both impressive and dangerously easy to overextend.

NIST reported in August that researchers had sent entangled photons through 62 kilometres of commercial fibre, much of it hanging from poles and exposed to heat, wind and physical movement. Real-time polarization correction countered the changing fibre. The system distributed 1,500 entangled photons per second and maintained distribution for 92.8% of a 24-hour period; the remaining 7.2% was used for correction. A statistical test confirmed that detections at the two ends remained entangled.

Those are unusually useful numbers. They identify the object tested, the environment, the rate, the observation period and the correction burden. They are not, however, a receipt for blind quantum computing, distributed computation, networked sensing or an “ultrasecure” end-to-end service. NIST itself said the rate would need to improve for practical networks. The experiment established that a fragile quantum resource could survive one difficult link under active stabilization. It did not establish memory, multi-hop swapping, application execution, service economics or a user outcome.

RFC 9583 supplies the missing discipline if it is read as a map rather than a promise. Published in June 2024, it is an Informational product of the IRTF Quantum Internet Research Group, not an Internet Standards Track specification. Its status language is explicit: IRTF results might not be suitable for deployment. The document offers an initial, non-exhaustive catalogue of applications and describes selected scenarios, including secure communication setup, blind quantum computing and distributed quantum computing.

Its most useful device is a six-stage ladder inherited from the 2018 quantum-Internet roadmap. Stage 1 uses trusted repeaters. Stage 2 adds prepare-and-measure capability. Stage 3 distributes entanglement. Stage 4 adds quantum memory. Stage 5 introduces fault-tolerant few-qubit networks. Stage 6 reaches quantum-computing networks. RFC 9583 places basic QKD, entanglement-enabled QKD, blind quantum computing, higher-accuracy clock synchronization and distributed computing at different points on that ladder.

A stage number is not a maturity badge. It describes a bundle of prerequisites. It does not say when the bundle will be available, at what scale, at what usable rate or fidelity, under whose operational control, or at what price. Nor does it show that the proposed application beats a classical alternative. A diagram can correctly place an application above quantum memory while saying nothing about whether memories last long enough for the workload, whether entanglement arrives faster than it decays, or whether control traffic completes before the resource is gone.

RFC 9340 makes the constraints harder to ignore. Entanglement is the network’s fundamental service, but fidelity is part of that service. Time is expensive because memory lifetimes are short, generation rates are low and even a small queue can destroy usefulness. Quantum and classical planes must operate together: qubits have no packet headers, so identification, routing, measurement results and orchestration travel over classical channels. The control record and the quantum resource must be correlated without confusing one for the other.

The application chain therefore needs several independent receipts. The physical layer establishes distance, loss and environmental stability. The resource layer establishes entanglement rate, fidelity, availability and correction overhead. The network layer establishes memory, swapping, routing and scheduling across more than one link. The control layer establishes identity, timing, authentication and coordination. Only then can an application test establish correctness. Comparative testing must still establish advantage over existing methods, and operating evidence must establish recovery, staffing, cost and a delivered outcome.

The dispute over quantum clock synchronization shows why this separation matters. RFC 9583 maps higher-accuracy clock synchronization to Stage 5. In January 2026, an individual Internet-Draft argued that RFC 9583-type clock synchronization is not a valid quantum-Internet application. It cited slow clock transport, the absence of time information in the relevant photon model, circular initial conditions and the adequacy of current methods.

That draft is not a verdict. It is revision 00, expired in July, and has not become an RFC, an erratum or QIRG consensus. The original 2000 proposal, a 2001 comment and later work form a longer technical argument. But the draft is still decision-relevant: an application can appear in a respected roadmap and remain contestable at the mechanism and comparative-value layers. Investment governance should preserve that uncertainty rather than convert a citation into readiness.

The correct reading of the NIST result is therefore stronger, not weaker, when its boundary is kept. A hard link worked under measured conditions. That advances one layer of a difficult system and gives future teams a real baseline. It does not make every use case above that layer real. Running code is primary only for the thing that ran.

Sources