Traceability Link Correctness: Project-Specific or Generic?
To enable proper maintenance of a software product, it is important to discover new and maintain existing traceability links between artifacts. Correctness of trace links is important as it builds trust in existing trace links and improves software maintenance. Neglecting traceability in a project has negative impact on the software quality and increases project development cost and time. Additionally, properly set trace links in a project provide flexibility within a development team as the knowledge of understanding inter-dependencies in the system does not rely on a domain expert only. The research community opinion on trace link correctness differs, which results in contradictory solutions on how to evaluate trace link correctness. In this paper, we identify notions of trace link correctness applicable on the data model level and examine if the uses of the defined notions are generic or project-specific. Additionally, the paper examines if the evaluation of trace links using the defined notions requires a domain expert or if they can be evaluated by software engineers lacking domain expertise. The study is conducted in two iterations. The first iteration focuses on identifying notions of trace link correctness and examines if they are generic or project-specific. The second iteration focuses on understanding if the evaluation of trace link correctness using the identified notions requires a domain expert. Five notions of trace link correctness are identified: Versioning, Lifetime/Lifespan, Non-Duplicated Trace Links, Unique Artifact Identification and Mandatory Artifacts & and Mandatory Trace Link. Non-Duplicated Trace Links and Unique Artifact Identification are identified as generic while Versioning, Lifetime/Lifespan and Mandatory Artifacts & and Mandatory Trace Link are identified as project-specific. Furthermore, it was found that not only a domain expert, but also experienced software engineers can evaluate trace link correctness using the identified notions.