On characterising strong bisimilarity in a fragment of CCS with replication

We provide a characterisation of strong bisimilarity in a fragment of CCS that contains only prefix, parallel composition, synchronisation and a limited form of replication. The characterisation is not an axiomatisation, but is instead presented as a rewriting system. We discuss how our method allows us to derive a new congruence result in the $\pi$-calculus: congruence holds in the sub-calculus that does not include restriction nor sum, and features a limited form of replication. We have not formalised the latter result in all details.

💡 Research Summary

The paper investigates strong bisimilarity for a deliberately restricted fragment of Milner’s Calculus of Communicating Systems (CCS). The fragment contains only action prefix, parallel composition, synchronization (communication via complementary actions), and a limited form of replication written as !a.P instead of the full !P. This restriction forces replicated processes to repeatedly offer the same action a, thereby avoiding the full expressive power of unrestricted replication while still capturing essential concurrent behaviour.
The authors first recall the standard definition of strong bisimilarity as a binary relation on labelled transition systems that is preserved by every transition step. Traditional proof techniques rely on either constructing bisimulation relations directly (often infeasible for infinite-state systems) or on axiomatisations that equate process terms through a set of algebraic laws. Both approaches become cumbersome when replication is present because the induced transition graph may be infinite.
To overcome this difficulty, the paper introduces a rewriting system R that operates directly on process terms. The system consists of three families of rewrite rules:

1. Prefix‑merge rule – when two parallel components share the same visible action a, they can be merged into a single prefix:
     a.P | a.Q → a.(P | Q).
   This rule eliminates redundant parallel prefixes and brings the term closer to a normal form.

2. Replication‑propagation rule – a replicated process !a.P interacts with any parallel component that offers the same action a:
     !a.P | a.Q → !a.P | P | Q.
   The rule “unfolds” one instance of the replicated behaviour, allowing the replicated process to participate in the computation without expanding it indefinitely.

3. Synchronization rule – complementary actions synchronize into an internal τ‑step:
     a.P | ā.Q → τ.(P | Q).
   This captures the standard CCS communication semantics.

All three rules are shown to be strongly bisimulation‑preserving: applying a rule to a term yields a new term that is strongly bisimilar to the original. Moreover, the authors prove confluence and termination of the rewriting system on the considered fragment. Consequently, every process term reduces to a unique normal form (NF). The central theorem states that two processes are strongly bisimilar iff their normal forms are syntactically identical. This result reduces the bisimilarity decision problem to a straightforward syntactic comparison after normalisation, providing an algorithmic procedure that works even in the presence of limited replication.

Having established the rewriting characterisation for the CCS fragment, the authors turn to the π‑calculus. They consider the sub‑calculus that excludes restriction (ν) and choice (+) but retains input, output, parallel composition, and the same limited replication. Because restriction is absent, all channel names are globally free, and the synchronization rule above suffices to model communication without needing additional scope‑handling machinery. The omission of choice eliminates nondeterministic branching, ensuring that the rewriting system remains deterministic. Under these conditions, the same normal‑form construction applies, and the authors prove a congruence theorem: strong bisimilarity is a congruence for all operators of this π‑sub‑calculus. In other words, if P ≈ Q then for any context C