Table of Contents
Fetching ...

Circumventing the FLP Impossibility Result with Open Atomic Ethernet

Paul Borrill

TL;DR

This essay argues that FLP is not a law of physics but a theorem about a particular system model -- and that Open Atomic Ethernet (OAE) circumvents it by rejecting the asynchronous model at its foundation, achieving deterministic atomic coordination without violating any impossibility result.

Abstract

The Fischer--Lynch--Paterson (FLP) impossibility result is widely regarded as one of the most fundamental negative results in distributed computing: no deterministic protocol can guarantee consensus in an asynchronous system with even one faulty process. For forty years, the field has treated this as an immovable constraint, designing around it with randomized protocols, failure detectors, and weakened consistency models. This essay argues that FLP is not a law of physics but a theorem about a particular system model -- and that Open Atomic Ethernet (OAE) circumvents it by rejecting the asynchronous model at its foundation. We introduce the term bisynchronous to describe OAE's key property: bounded-time bilateral resolution in which both parties reach common knowledge of outcome at every round boundary -- a strictly stronger guarantee than synchrony alone. By constructing a bisynchronous, swap-based protocol at Layer 2, OAE sidesteps the load-bearing assumptions of FLP's asynchronous model, achieving deterministic atomic coordination without violating any impossibility result.

Circumventing the FLP Impossibility Result with Open Atomic Ethernet

TL;DR

This essay argues that FLP is not a law of physics but a theorem about a particular system model -- and that Open Atomic Ethernet (OAE) circumvents it by rejecting the asynchronous model at its foundation, achieving deterministic atomic coordination without violating any impossibility result.

Abstract

The Fischer--Lynch--Paterson (FLP) impossibility result is widely regarded as one of the most fundamental negative results in distributed computing: no deterministic protocol can guarantee consensus in an asynchronous system with even one faulty process. For forty years, the field has treated this as an immovable constraint, designing around it with randomized protocols, failure detectors, and weakened consistency models. This essay argues that FLP is not a law of physics but a theorem about a particular system model -- and that Open Atomic Ethernet (OAE) circumvents it by rejecting the asynchronous model at its foundation. We introduce the term bisynchronous to describe OAE's key property: bounded-time bilateral resolution in which both parties reach common knowledge of outcome at every round boundary -- a strictly stronger guarantee than synchrony alone. By constructing a bisynchronous, swap-based protocol at Layer 2, OAE sidesteps the load-bearing assumptions of FLP's asynchronous model, achieving deterministic atomic coordination without violating any impossibility result.
Paper Structure (11 sections, 1 theorem, 2 figures, 1 table)

This paper contains 11 sections, 1 theorem, 2 figures, 1 table.

Key Result

Theorem 1

No deterministic protocol can guarantee consensus among $n \geq 2$ processes in an asynchronous system if even one process may crash.

Figures (2)

  • Figure 1: OAE bilateral swap as Petri net. Alice and Bob hold symmetric EPI (epistemic) knowledge states. The Ping/Pong heartbeat drives ONT (ownership) transitions. 16 transitions, 8 states; the symmetry break resolves to a single committed outcome.
  • Figure 2: Alice and Bob perspectives of the OAE Petri net. Each party has three places and three transitions under their control. The Spekkens epistemic restriction ensures neither party can distinguish all states---only bisynchronous swap resolves the joint state.

Theorems & Definitions (1)

  • Theorem 1: FLP, 1985