Handling schematic differences between encodings

[m-mohr]

How do we handle schematic differences between the encodings, both in terms of documentation, data types and validation? While the mapping for fields like area are relatively straightforwards (GeoParquet float vs GeoJSON number), it is less obvious for e.g. bbox and geometry.

bbox:

• For GeoParquet it is still in discussion: https://github.com/opengeospatial/geoparquet/pull/191 But it looks like it's either 4 top-level fields or an object with keys such as xmin, xmax, ymin, ymax.
• In GeoJSON it's usually an aray of 4 (or 6) numbers.

geometry:

• In GeoParquet it's currently a WKB
• In GeoJSON it's an object with a type field and coordinates.

This all describes pretty much the same, but schematically they are different and as such need different schemas for validation.

It's not quite obvious how we define the data type in the core metadata. Do we leave the technical details open and just describe the semantic of the field?

cc @cholmes

[m-mohr]

Started working on a generic schema language in #8