Research method
Evidence before availability claims.
A public webpage, an accessible endpoint, and openly reusable data are three different things. This catalog records them separately and distinguishes source freshness from the freshness of every returned field.
Evidence labels
A read-only request returned usable data during this research session.
A public interface visibly exposed the data, but no reusable endpoint was validated.
A catalog or layer listing was inspected without retrieving representative records.
Primary documentation describes the data, but this research did not retrieve it.
The source or endpoint failed current checks.
Field freshness
Current operations or observations: vehicle position, outage status, sensor measurement.
Relatively stable identity or geography: route name, station location, category label.
Calculated for a request: minutes to arrival, route plan, current index classification.
A completed past period: census result, procurement award, annual statistic.
The source does not establish cadence or the field semantics remain unclear.
Field dictionaries
Tables record the exact field or path, observed/documented type, nullability, example, human meaning, freshness, units, enums, anomalies, and localization behavior. Nested entities receive separate tables instead of being flattened into an unreadable list.
Licensing boundary
Technical accessibility is not treated as permission to reuse. When terms are unclear, the catalog says so. A community wrapper's software license never substitutes for the underlying provider's data license.
Representative samples
Samples preserve public values, nulls, localized text, and anomalies that teach something about the schema. Credentials, cookies, personal information, and tokens are excluded.
Interpretation limits
Geographic proximity is not population impact. Missing crowdsourced records are not absence. One current vehicle snapshot is not long-term transit equity. Each source page records the boundary between the data and any inference.