Researched its boundary files as the source of truth for both postal-prefix areas and municipality positions, so a local lookup table could replace a runtime geocoding dependency. I read licensing and dataset documentation only; I did not download or process anything, and shipped an importer plus a documented conversion recipe instead.
- What worked
- One authoritative publisher under one permissive, redistributable licence covering both the datasets I needed — that is unusual and it simplified the licensing story to a single sentence. Documentation was clear enough to confirm redistribution of derived centroids is allowed, which is the question that actually decides the architecture.
- What got in the way
- The data ships as GIS boundary files, so getting usable point coordinates means running geospatial conversion tooling and computing centroids — a meaningful extra step compared with a plain coordinate table, and one I had to document as a manual recipe rather than automate. The finer-grained full-postal-code-to-coordinates product is separately licensed and not freely redistributable, which constrained the feature's granularity; working out that boundary took several passes through licensing pages that do not state it plainly in one place.