Wikidata: Wikipedia's shared facts, in the public domain
On 29 October 2012 Wikimedia Deutschland opened Wikidata. At first it could only create items and attach Wikipedia articles in different languages to them; every item got an identifier beginning with Q, and all the data got the CC0 licence, which is the public domain.
Why it matters
In the same year a knowledge graph became one company's property and stopped being reachable from outside, an open twin appeared: facts that can be downloaded whole, queried and reused without permission. It is why most later open systems rest on Wikidata rather than on the closed graph.
The launch was preceded by a log entry of 26 October: the database is created but not yet editable. The entry for 29 October reads "Wikidata is launched". The first phase solved a specific misery. Wikipedia's language editions kept links to one another inside article text, quadratically: 207 articles on Rome carried 206 links each, 42,642 lines in all. By the end of 2012, 66 of the 287 editions held more of these links than actual content. Wikidata moved them to one place, a page per concept, called an item. The rollout was staged: the Hungarian Wikipedia on 14 January 2013, Hebrew and Italian on 30 January, English on 13 February, the remaining 282 editions on 6 March. Phase two, properties and statements, began on 4 February 2013 and reached all editions on 24 April. The data model was described later by the founders themselves: every claim can carry qualifiers and a list of references, and contradictory claims are both kept - Wikidata "intends to represent all views rather than choose one 'true' claim", marking one preferred or another deprecated. What this record does not claim. It does not say Wikidata records provenance automatically: the article says the opposite - the system provides a structure for references but does not track provenance itself.