[Paper Review] On Negotiation as Concurrency Primitive

Javier Esparza, Joerg Desel|arXiv (Cornell University)|Jul 8, 2013

Business Process Modeling and Analysis12 references4 citations

TL;DR

This paper introduces negotiations as a concurrency primitive, modeling multi-party interactions with atomic negotiation steps. It presents a complete set of syntactic reduction rules that summarize sound, acyclic, deterministic negotiations into a single equivalent atom in polynomial time, enabling efficient analysis without state-space explosion.

ABSTRACT

We introduce negotiations, a model of concurrency close to Petri nets, with multiparty negotiation as primitive. We study the problems of soundness of negotiations and of, given a negotiation with possibly many steps, computing a summary, i.e., an equivalent one-step negotiation. We provide a complete set of reduction rules for sound, acyclic, weakly deterministic negotiations and show that, for deterministic negotiations, the rules compute the summary in polynomial time.

Motivation & Objective

• To formalize negotiations as a concurrency model with multiparty atomic interactions as a primitive.
• To address the soundness problem—ensuring negotiations complete without deadlocks or livelocks—and the summarization problem—computing a one-step equivalent of a multi-step negotiation.
• To develop reduction rules that preserve behavioral equivalence while avoiding explicit state-space exploration, thus mitigating the state-explosion problem.
• To establish polynomial-time algorithms for summarizing deterministic, acyclic negotiations using syntactic transformations.
• To extend existing reduction techniques from Petri nets to the novel context of negotiation-based systems with a focus on summarization, not just soundness.

Proposed method

• Introduces a formal model of negotiations using atoms defined by parties, outcomes, and state-transforming relations.
• Defines soundness as the absence of deadlocks or livelocks, and summarization as computing a one-step equivalent negotiation that preserves final state behavior.
• Proposes two core reduction rules: merge (for atoms with multiple outcomes leading to the same next atom) and d-shortcut (for atoms with a single outcome that unconditionally leads to the next step).
• Applies the rules in a systematic way, guided by the length of paths to the final atom and the shortest occurrence sequences involving each outcome.
• Uses structural properties such as acyclicity, determinism, and weak determinism to ensure completeness and termination of the reduction process.
• Proves that for deterministic negotiations, the number of rule applications is bounded by the total number of outcomes and shortest occurrence sequences, leading to polynomial-time complexity.

Experimental results

Research questions

• RQ1 Can a complete and sound negotiation be reduced to a single equivalent negotiation atom using syntactic rules without state-space exploration?
• RQ2 What is the computational complexity of the summarization problem for acyclic negotiations, and how does it vary between deterministic and non-deterministic cases?
• RQ3 Are there complete sets of reduction rules that preserve both soundness and summarization equivalence for specific classes of negotiations?
• RQ4 Can the reduction process be guaranteed to terminate in polynomial time for deterministic, acyclic negotiations?
• RQ5 How do the proposed rules compare to existing reduction techniques in Petri net theory, particularly in their applicability to summarization rather than just liveness or soundness?

Key findings

• A complete set of syntactic reduction rules—merge and d-shortcut—is presented for sound, acyclic, weakly deterministic negotiations.
• For deterministic negotiations, the reduction process terminates in polynomial time, bounded by the number of outcomes and shortest occurrence sequences.
• The number of rule applications is at most Out(N) for merge and Shoc(N) for d-shortcut, where Out(N) counts outcomes (excluding the final atom) and Shoc(N) sums shortest path lengths.
• The reduction rules preserve behavioral equivalence, ensuring that the final one-step negotiation correctly captures the final state distribution of the original negotiation.
• The summarization problem is co-NP-hard for arbitrary acyclic negotiations, but solvable in polynomial time for deterministic ones.
• The approach avoids the state-explosion problem by operating purely on the syntactic structure, making it scalable for complex but well-structured negotiations.