Process Enactment and Coordination

Jean-Marc Andreoli, Jean-Luc Meunier, Daniele PaganiEWSPT 1996 · DOI

TL;DR

This paper introduces the Coordination Language Facility (CLF), a novel system for modeling and enacting complex software development processes. CLF offers a flexible, decentralized approach to coordinating distributed, heterogeneous software components, enabling dynamic process adaptation and robust execution in wide-area environments.

What Problem Does This Paper Address?

The paper addresses the challenge of effectively supporting and automating complex knowledge work and software development processes. Traditional workflow management systems often struggle with the dynamic, distributed, and heterogeneous nature of these processes. They typically rely on rigid Input-Process-Output (IPO) models and struggle to adapt to changing requirements or environments. The authors highlight the need for a coordination system that can handle distributed objects, allow for dynamic process modification, and operate reliably in wide-area computing settings, bridging the gap between low-level operating system primitives and high-level application requirements.

Methods

The core of the paper is the introduction and detailed explanation of the Coordination Language Facility (CLF). CLF is presented as a declarative coordination system that uses a high-level scripting language to coordinate autonomous active objects (called "participants").

Key aspects of CLF's methodology include:

  • Resource-Based Object Model: CLF objects are viewed as resource managers that accept operations for inserting and removing resources. These resources are accessed through properties defined in the object's interface.
  • CLF Protocol: A client/server protocol involving Inquiry, Reservation, Confirmation/Cancellation, and Insert operations. Inquiry is "defered synchronous," Reservation is synchronous, and Confirmation/Cancellation/Insert are asynchronous. This protocol forms the basis for elementary transactions using a two-phase commit.
  • CLF Scripting Language: A rule-based language used by "coordinators" to orchestrate operations on participants. Rules consist of left-hand sides (resources to be removed) and right-hand sides (resources to be inserted), separated by <>-.
  • Signature Resources: Used to specify implicit sequentialization constraints for inquiries, managing dependencies between tokens.
  • Modular Architecture: CLF objects and coordinators can be distributed, migrated, and replicated, supporting a highly decentralized and fault-tolerant architecture.
  • Dynamic Rule Manipulation: CLF rules can be manipulated within a process, allowing for dynamic modification of process models at runtime.

The paper contrasts CLF with traditional approaches, emphasizing its flexibility in handling dynamic changes to both the core process model and application-specific templates.

Datasets

Not discussed in the paper. The paper focuses on a conceptual model and its application to a sample process, rather than empirical evaluation on specific datasets.

Key Insights and Ideas

  • Decoupling Process Logic from Implementation: CLF separates the coordination logic (expressed in CLF rules) from the actual implementation of the participants (which can be written in any language). This promotes modularity and reusability.
  • Dynamic Process Adaptation: The ability to dynamically change CLF rules at runtime is a significant advantage, allowing for flexible adaptation of processes to evolving requirements or unexpected situations, a common challenge in software development.
  • Distributed Coordination as a First-Class Citizen: CLF is designed from the ground up for distributed environments, providing mechanisms for coordinating objects across a network without relying on a central database or heavy infrastructure.
  • "Process Assembly Language" Analogy: The CLF language is described as a "process assembly language," suggesting a low-level yet powerful way to construct and manipulate processes, offering fine-grained control.
  • Unified View of Actors: CLF treats all actors (human or software) as active objects to be coordinated, providing a more uniform view and enhancing modularity.

Key Results

The paper demonstrates the feasibility and benefits of using CLF to model and enact software development processes.

  • Modeling IPO Processes: The paper shows how traditional IPO process models and their hierarchical decomposition can be directly translated into CLF rules.
  • Implementing Flow Control: CLF rules are used to model fundamental flow control primitives like sequential execution, parallel execution, branching, and decomposition.
  • Handling Activity States: The paper illustrates how activity states (e.g., inactive, active, done) and transitions can be represented and managed using CLF rules.
  • Role Resolution and User Assignment: CLF rules are shown to effectively model mechanisms for assigning tasks to users, including extensions for manager-controlled selection.
  • Software Defect Report Process Example: A detailed walkthrough of modeling and enacting a software defect report process demonstrates CLF's practical application in a realistic software engineering scenario. The paper outlines the CLF rules and participant interfaces required for this process.
  • Benefits Highlighted: The paper emphasizes the benefits of CLF, including its modular distributed architecture, flexibility in changing process models, and robust support for wide-area computing.

Limitations and Future Work

  • Complexity of Rule Writing: While flexible, writing CLF rules can be complex, especially for intricate processes. The paper mentions that graphical editors could generate CLF rules to mitigate this.
  • Performance in Highly Distributed Environments: While CLF supports distribution, the paper doesn't delve deeply into performance benchmarks or optimizations for extremely large-scale, highly distributed scenarios.
  • Integration with Existing Systems: The paper discusses encapsulating legacy applications but doesn't provide extensive details on the practical challenges or specific strategies for integrating diverse existing systems.
  • Future Work: The authors suggest that further work could explore more sophisticated extensions to the basic mechanisms, such as advanced user interaction models, more complex exception handling, and further investigation into dynamic distribution and migration of CLF rules. The paper also implies that the CLF could be extended to support more complex forms of negotiation and collaboration.

Summary updated

BibTeX
@inproceedings{andreoli1996process,
  author = {Jean-Marc Andreoli and Jean-Luc Meunier and Daniele Pagani},
  title = {{Process Enactment and Coordination}},
  booktitle = {EWSPT},
  year = {1996},
  doi = {10.1007/bfb0017745},
}
Figure