Providing structural metadata is a W3C Data on the Web Best Practice.
Providing information about the internal structure of a data distribution helps others understand, explore or query the dataset.
Structural metadata is usually supplied in a machine-readable format describing for each data column:
- name, title, description
- data type (e.g. number, and format
- any constraints (e.g. required, minimum length, unique)
- is it a primary or foreign key
- what represents a missing value.
Standards to specify structural metadata include:
- Model for Tabular Data and Metadata on the Web (limited implementations)
- XML Schema (broad support but open data is not often published as XML)
- Table Schema (a number of implementations and can be integrated into CKAN)
Do you provide structural metadata with your open data? If so, how do you do it?
- are your schemas hand-crafted, inferred or generated from the source database?
- what standard do you use and why?
- what tools do you recommend?