On the Enactability of Business Protocols

Nirmit Desai, Munindar P. SinghAAAI 2008 · OpenReview

Abstract

Protocols specifying business interactions among autonomous parties enable reuse and promote interoperability. A protocol is specified from a global viewpoint, but enacted in a distributed manner by (agents playing) different roles. Each role describes a local representation. An ill-specified protocol may yield roles that fail to produce correct enactments of the protocol. Existing approaches lack a formal and comprehensive treatment of this problem. Building on recent work on declaratively specifying a protocol as a set of rules of causal logic, this paper formally defines the enactability of protocols. It presents necessary and sufficient conditions for the enactability of a protocol as well as a decision procedure for extracting correct roles from enactable protocols.

Introduction: The Global-Local Gap

In modern e-commerce, business interactions such as supply chain orders or foreign exchange are governed by "protocols." Designers typically specify these protocols from a global viewpoint, assuming a shared state where every message is observed in a specific, consistent order. However, in practice, these protocols are enacted locally by autonomous, asynchronous agents (e.g., Buyer, Seller, Shipper). These agents only observe the messages they directly send or receive.

This discrepancy leads to two major classes of failure:

  1. Nonlocal Choice: Two agents might simultaneously perform actions that are globally mutually exclusive because neither is aware of the other's immediate action.
  2. Blindness: An agent may be required to act based on information they have not yet received, or they may be "blind" to the overall progress of the interaction, leading to protocol violations.

The paper formalizes these failures to determine if a protocol is "enactable" before implementation.

Methodology: Causal Logic and Awareness

The authors utilize the action description language $C+$, based on causal logic, to model protocols. This approach goes beyond simple message sequencing by incorporating "fluents" (conditions that hold true or false) and "commitments" (legal or business obligations between parties).

Key methodological contributions include:

  • Formal Definitions of Awareness: Distinguishing between "direct" and "indirect" awareness of messages or state changes.
  • The Five Properties (P1–P5): Necessary and sufficient conditions for enactability, covering choice localization, message verifiability, and commitment discharge.
  • Role Extraction Algorithm: A procedure to automatically derive local rules for an individual agent from the global protocol specification.

Validation through Representative Protocols

The framework is validated using several business logic models that illustrate various enactment challenges.

Figure 1: Representative business protocols Ord and Rec used for validation.
  • Ord: A standard order placement protocol between a buyer and a seller.
  • Rec: A claim reception and verification protocol involving three parties: Reporter, Call Center, and Provider.
  • BL (Blindness) Family: Synthetic protocols designed to isolate and illustrate specific failure modes.

Key Insights: A Taxonomy of Failure

A central contribution is the categorization of "blindness" and its impact on protocol execution.

Figure 2: Illustration of nonlocal choice and various types of blindness in distributed enactment.
  • Taxonomy of Blindness:
    • Benign: The agent is unaware of a state, but the protocol remains functional (though local verification is difficult).
  • Malignant: The protocol fails because an agent's lack of information leads to incorrect message ordering.
  • Fatal: A fundamental modeling error where no local implementation can satisfy global requirements.
  • Commitment-Centric Design: The meaning of messages—expressed as business commitments—is as vital as their order for determining enactability.
  • Local vs. Global Consistency: A protocol is correct only if the local knowledge bases of all participants eventually agree on the status of mutual commitments.

Key Results

The paper establishes a rigorous "litmus test" for protocol designers:

  • Proposition 3: A protocol is enactable if and only if it satisfies five formal properties (P1–P5).
  • Algorithm 1: An automated role extraction algorithm that prunes rules invisible to a specific role and replaces global fluents with local message-based evidence.
  • Identification of Nonlocal Choice: The authors demonstrate that many existing standards (e.g., TWIST) contain inherent nonlocal choices, requiring additional synchronization layers for distributed implementation.

Limitations and Future Work

While the paper provides a robust logical foundation, some areas remain for future exploration:

  • Complexity of Choice: Current definitions focus on binary choices. Extending this to multi-party choices increases computational complexity, often necessitating heavy synchronization like two-phase commits.
  • Reliability Assumptions: The model assumes a reliable messaging infrastructure. Future research could address enactability in the presence of lossy or malicious networks.
  • Tooling: Integrating these formal checks into mainstream business process modeling IDEs is a key next step for practical adoption.

Summary updated

Abstract

Protocols specifying business interactions among autonomous parties enable reuse and promote interoperability. A protocol is specified from a global viewpoint, but enacted in a distributed manner by (agents playing) different roles. Each role describes a local representation. An ill-specified protocol may yield roles that fail to produce correct enactments of the protocol. Existing approaches lack a formal and comprehensive treatment of this problem. Building on recent work on declaratively specifying a protocol as a set of rules of causal logic, this paper formally defines the enactability of protocols. It presents necessary and sufficient conditions for the enactability of a protocol as well as a decision procedure for extracting correct roles from enactable protocols.
BibTeX
@inproceedings{desai2008enactability,
  author = {Nirmit Desai and Munindar P. Singh},
  title = {{On the Enactability of Business Protocols}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {2008},
  url = {https://openreview.net/forum?id=rkZKSk-uWS},
}
Figure