Agreed! People are surprised but CSV files (while I would not use them for everything) work great in low-tech environment, for instance, or when files must be p

Agreed! People are surprised but CSV files (while I would not use them for everything) work great in low-tech environment, for instance, or when files must be produced by many different actors with different technical levels. They also work nicely as an interop between different stacks (Cobol <-> Ruby). One concrete example is the French standard to describe Electrical Vehicles Charge Points, which is made of 2 parts (static = position & overall description, dynamic = current state, occupancy etc). Both "files" are just CSV files: https://github.com/etalab/schema-irve https://github.com/etalab/schema-irve Both sub-formats are specified via TableSchema (https://frictionlessdata.io/schemas/table-schema.json https://frictionlessdata.io/schemas/table-schema.json). Files are produced directly by electrical charge points operators, which can have widely different sizes & technicality, so CSV works nicely in that case.