TL;DR
This paper introduces the "ViewPoints" framework, a novel approach to managing inconsistencies that arise naturally in large-scale, evolving software specifications. By treating different perspectives as distinct "ViewPoints" and explicitly managing their relationships and inconsistencies, the framework facilitates more robust requirements elicitation and development.
What Problem Does This Paper Address?
Developing large software systems is an inherently evolutionary process. As multiple participants contribute to a specification, they often hold different, and sometimes conflicting, views of the system. This leads to inconsistencies, which can be difficult to detect and resolve, and even harder to keep resolved as the specification evolves. Traditional approaches often enforce strict consistency, which can hinder development progress. This paper addresses the challenge of managing these inevitable inconsistencies in a way that tolerates them, uses them as a source of information, and ultimately leads to more robust specifications.
Methods
The core of the paper is the "ViewPoints" framework. This framework partitions software development knowledge into multiple, overlapping "ViewPoints," each representing a distinct perspective or actor's view. Key aspects of the method include:
- Explicit Relationship Representation: Inconsistencies are managed by explicitly defining and recording the relationships between different ViewPoints. These relationships can be syntactic, semantic, or process-oriented.
- Inconsistency Detection and Classification: The framework supports detecting inconsistencies by checking these defined relationships. Inconsistencies are then classified by their cause (e.g., mistake, misunderstanding) and type.
- Inconsistency Handling: Rather than immediate resolution, the framework supports various handling strategies:
- Ignore: For minor inconsistencies with low risk.
- Delay: When more information is needed.
- Circumvent: Temporarily disabling or modifying a rule.
- Ameliorate: Incremental steps towards resolution.
- Resolve: Taking direct actions to repair the inconsistency.
- Work Records: Each ViewPoint maintains a "work record" detailing detected inconsistencies, actions taken, and their outcomes. This record is crucial for tracking the evolution of resolutions and understanding the impact of changes.
- User-Defined Relationships: The framework allows users to define their own relationships between ViewPoints, enabling the capture of domain-specific interactions and nuances.
Datasets
Not discussed in the paper. The paper uses a scenario involving the behavioral specification of a telephone as its primary example for demonstrating the framework.
Key Insights and Ideas
- Inconsistency as a Feature, Not a Bug: The paper argues that inconsistencies, when managed properly, are not just problems but valuable indicators of missing information, misunderstandings, and differing conceptual models.
- ViewPoints as Building Blocks: The ViewPoints framework provides a structured way to manage complexity by breaking down a large specification into manageable, distinct perspectives.
- Explicit Management is Key: The success of the approach hinges on explicitly representing and managing the relationships between ViewPoints and the inconsistencies that arise from them.
- Evolutionary Nature of Resolution: Resolutions are not static. The framework acknowledges that changes in one ViewPoint can affect others, potentially re-introducing or resolving inconsistencies, and emphasizes the importance of tracking these effects.
- Conflict Resolution through Inconsistency Analysis: By identifying and resolving inconsistencies, the framework indirectly helps in uncovering and resolving underlying conflicts and mistakes in conceptual models.
Key Results
The paper demonstrates the effectiveness of the ViewPoints framework through a detailed scenario involving the specification of a telephone. The scenario illustrates:
- How the framework can detect inconsistencies arising from different interpretations of states and transitions between two ViewPoints (Anne's and Bob's).
- How various resolution actions can be suggested and applied, with some resolutions having unintended side effects that are also tracked.
- The importance of user-defined relationships in capturing domain-specific interactions that might not be caught by generic consistency rules.
- The ability to manage the evolutionary aspect of specifications, where resolutions can be affected by subsequent changes.
Limitations and Future Work
The authors acknowledge several areas for future work:
- Further Formalization: More formalization of the ViewPoints framework and inconsistency notions is needed to improve tool support for inconsistency handling and reasoning.
- Dependencies: Further investigation into the dependencies between elements of an evolving specification is required.
- Domain-Specific Relationships: While demonstrated, more examples and case studies are needed to validate the approach for eliciting, expressing, recording, and tracking domain-specific relationships.
- Conflict Detection: The paper notes that some conflicts may not manifest as inconsistencies detectable by the current set of rules, suggesting the need for new consistency rules to capture these.
- Tool Support: While a prototype "Viewer" and "Inconsistency Handler" are described, further development is needed to fully support the described inconsistency management tools, particularly for distributed and concurrent development.
- Development Guidance: Combining inconsistency handling with development guidance remains an area for further research.
Summary updated