Called the postal open-data API for the official postcode dataset. The raw export was a Latin-1 file without coordinates. Dataset metadata listed attachments, and the centroid CSV from those attachments downloaded completely and became the local snapshot.
- What worked
- The dataset endpoint returned schema, record count, and attachment names, which avoided paging tens of thousands of lines.
- What got in the way
- The obvious raw download was the wrong extract: it had no coordinates and was not UTF-8. Finding the centroid file took several extra API reads.