Language Tag

BCP 47 identifies a language with a tag such as en, zh-Hant or sl-rozaj, composed from registered subtags: a primary language, optionally an extended language, a script, a region and variants. The IANA Language Subtag Registry is the source of truth for the subtags and for a small set of whole tags registered before the grammar settled (grandfathered) or registered redundantly. The language-tags package at codes.kanonak.org is that registry projected into Kanonak by its producer; this package is the schema the projection is typed with.

The two classes to know are Language Tag — anything that can stand as a complete tag, and therefore anything a language-tagged string may point at — and Subtag — a registered component. A Language Subtag is both, so en alone is a valid tag. A Script Subtag such as Hant, a Region Subtag such as US or 001, an Extended Language Subtag and a Variant Subtag are components only. Grandfathered Tag and Redundant Tag are the registry's whole-tag entries. Because the registry holds subtags, not the tags people compose from them, a document that needs zh-Hant-TW declares it in its own package as a LanguageTag composed via Primary Language, Script and Region from the registry's entries — the composition is validated, and the registry stays a finite list.

Lifecycle follows the code-lists floor: a retired entry carries owl.deprecated and Deprecated On, and the registry's preferred value becomes prov.supersededBy on the entry — so a validator can say "iw is deprecated; use he" with a real reference, not a string.