3.4 KiB
3.4 KiB
collapsed:: true type:: REVIEWS tags:: year:: 2026 venue:: ECMFA full-title:: Endotrophic Models: Formalizing Inter-Model Consistency and Semantic Overlap date-start:: 05-02-2026 - 21:41 date-submitted:: external-links:: status:: DOING deadline-submission:: file:: parent:: todoist:: https://app.todoist.com/app/task/30-colin-atkinson-mattias-ulbrich-romain-pascual-and-erik-burger-endotrophic-mod-6frhCMrvj87rRx9c
- ### [[Comments]]
- Summary: The paper addresses the problem that large systems are described through multiple heterogeneous artefacts (e.g., different models, schemas, constraints, rule fragments) that may partially overlap and evolve independently, making it hard to reason uniformly about what each artefact implies, where constraints fail, and what happens when artefacts are combined. To address this, the authors propose Endotrophic Models (EMs) as a small rule-based "semantic back-end" based on function-free Horn rules plus denial constraints. An EM computes (1) a least-fixpoint closure of what can be derived, (2) explicit witnesses that show which constraints are violated (with concrete bindings), and (3) a query interface that can behave in an open-world or closed-world way for facts that are not derived. The paper then defines several analysis notions for single models and for composed models (syntactic union), and illustrates them with a small running example.
- Comments: The paper is about a relevant problem, and the core idea of separating “what you can derive” from “which constraints are violated” is interesting because it makes violations explicit. However, my main concern about the paper is about the presentation as discussed below:
- The paper becomes technical very early (atoms, fixpoints, witnesses, policies), and the reader has to wait until later to see a concrete multi-model example. Even though Section 4 provides a running example, I suggest to add a very small, intuitive example in Section 1 or at the start of Section 2 involving two or three recognisable modelling artefacts (e.g., a small UML-ish domain model + OCL-like constraint + an access-policy rule set, or a schema+view+constraint combination). Then reuse this example to explain (i) what becomes a fact, (ii) what becomes a rule, (iii) what becomes a denial, (iv) what composition means, and (v) how a composite-only conflict/witness looks.
- The paper claims heterogeneous artefacts can be mapped into the EM kernel. A short, end-to-end walkthrough that maps one familiar modelling artefact (e.g., a diagram plus an invariant) into facts, derivation rules, and denials would clarify what is assumed from the user/tooling and how semantic overlap is established in practice. Table 1 aligns "modeler-facing" terms with formal terminology, but it is still not fully clear how this alignment relates to typical modelling stacks (model/metamodel, transformations/weaving, trace links, etc.), nor how semantic correspondences/overlap are obtained or assumed.
- The formal development and operators are well-structured, but the paper would benefit from stronger evidence that the operators (overlap, amplification, redundancy exposure, conflict sets) answer practitioners' questions effectively beyond the running example. Even a small "experience report" style discussion (e.g., what scale is feasible, typical numbers of predicates/facts/denials, what kinds of conflicts are found) would improve the overall quality of the paper.
-