On Payment Channels in Asynchronous Money Transfer Systems

TL;DR

This paper investigates the feasibility and limitations of implementing payment channels in asynchronous money transfer systems, providing bounds and conditions for their construction in such environments.

Contribution

It establishes a quadratic lower bound on on-chain transfer complexity and characterizes when unidirectional and bidirectional payment channels are possible in asynchronous systems.

Findings

Quadratic lower bound on on-chain transfer message complexity.

Conditions under which payment channels can be implemented asynchronously.

Impossibility results for certain types of payment channels in asynchronous settings.

Abstract

Money transfer is an abstraction that realizes the core of cryptocurrencies. It has been shown that, contrary to common belief, money transfer in the presence of Byzantine faults can be implemented in asynchronous networks and does not require consensus. Nonetheless, existing implementations of money transfer still require a quadratic message complexity per payment, making attempts to scale hard. In common blockchains, such as Bitcoin and Ethereum, this cost is mitigated by payment channels implemented as a second layer on top of the blockchain allowing to make many off-chain payments between two users who share a channel. Such channels only require on-chain transactions for channel opening and closing, while the intermediate payments are done off-chain with constant message complexity. But payment channels in-use today require synchrony, therefore they are inadequate for asynchronous money transfer systems. In this paper, we provide a series of possibility and impossibility results for payment channels in asynchronous money transfer systems. We first prove a quadratic lower bound on the message complexity of on-chain transfers. Then, we explore two types of payment channels, unidirectional and bidirectional. We define them as shared memory abstractions and prove that in certain cases they can be implemented as a second layer on top of an asynchronous money transfer system whereas in other cases it is impossible.