This document discusses the challenges of correctly reusing model transformations. It begins by noting that while transformations can increase productivity, they are often not reused due to difficulty and cost. The opportunity of reusable transformations is highlighted, but it notes that correct reuse must be ensured. Existing reuse mechanisms like model subtyping and concepts are examined, but found to not guarantee intent preservation. The difficulty of correct reuse is that transformation semantics and intents depend on both the syntax and semantics of source and target languages, which can be in arbitrary relation. Special cases like families of transformations parametrized with language-specific predicates or model types are proposed to enable more sound and complete reuse, but implementations may not be efficient. The conclusion is that semantics are key to correct