open-civic-data.json

{
	category : "category_name", // Required Single category from (http://opencivicdata.com/#categories-wg)
	meta : { // Optional: Simple Dublin Core Metadata Element Set
		title : "",
		creator : "",
		subject : "",
		description : "",
		publisher : "",
		contributor : "",
		date : "",
		type : "",
		format : "OpenCivicData.org",
		identifier : "",
		source : "",
		language : "",
		relation : "",
		coverage : "",
		rights : ""
	}
	geometry : {}, // Optional Geojson location feature (http://geojson.org/geojson-spec.html)
        // any other optional properties may follow. These may include both common and shared properties
        // from (http://opencivicdata.com/#properties-wg). some illustrative examples:
        historic: true|false, // if the object is considered historic
        last_maintenance_date: RFC3339DateTime,
        manufacturer: "ACME Corp.",
        starts_at: RFC3339DateTime, // for events such as road closures or live music
        ends_at: RFC3339DateTime
}

90% of the data I'm dealing with has data-set-level metadata. object-level metadata properties are rare, but possible, and typically override data-set-level metadata properties.

Could you give examples of the difference between dataset level metadata and object level metadata just so I am clear on where the line is?

E.g. our bike facilities layer.

dataset datasource: MAPC
some trails are based on DOT or town data, so they would have different datasource entries on the object.

ODbL http://www.opendatacommons.org/licenses/odbl/ was developed to enable databases with different licenses on an object level for instance.