Static versus dynamic reversibility in CCS

Abstract

The notion of reversible computing is attracting interest because of its applications in diverse fields, in particular the study of programming abstractions for fault tolerant systems. Most computational models are not naturally reversible since computation causes loss of information, and history information must be stored to enable reversibility. In the literature, two approaches to reverse the CCS process calculus exist, differing on how history information is kept. Reversible CCS (RCCS), proposed by Danos and Krivine, exploits dedicated stacks of memories attached to each thread. CCS with Keys (CCSK), proposed by Phillips and Ulidowski, makes CCS operators static so that computation does not cause information loss. In this paper we show that RCCS and CCSK are equivalent in terms of LTS isomorphism.

Forward-Reverse Observational Equivalences in CCSK

“…In [19], CCSK has been proved equivalent to RCCS [9], the first reversible process calculus, thus, in principle, our results apply to RCCS as well. However, the mechanism of memories used in RCCS introduces much more redundancy than the key mechanism used in CCSK, hence a direct application of our results to RCCS is not easy.…”

Foundations of Reversible Computation

“…Indeed, the approach of [160] can be considered static, since the structure of processes does not change during computation, and the minimal history information needed to enable reversibility is kept in the processes themselves, while in [44] the process is consumed during execution (as standard in process calculi) and larger memories are added to store history information. Nonetheless the two methods are equivalent as hinted at by [130] and fully proved by [115], where a mapping from an instance on CCS of [160] to the reversible CCS of [44] and vice versa is presented.…”

Towards Bridging Time and Causal Reversibility

“…Note that, since dynamic operators such as prefixing and choice are forgetful by definition, making them static avoids information loss during a reduction. Despite these two approaches may seem different, they have been shown to be equivalent in terms of labeled transition system isomorphism [18]. The approach of [5] is more suitable for systems whose operational semantics is given in terms of reduction semantics, hence its application is to be preferred in the case of very expressive calculi [4,19] as well as programming languages [21,24].…”