By default, we rely on the ISO 3166-1 classification for country codes. This taxonomy is not up to date and some inaccuracies are currently hidden because we use common names instead of the official one (see commit 8d1753c ).
CLDR unicode data seems like a good candidate: