nabu-lects

A small, curated registry of lects — language varieties identified by genealogical anchor × historical stage × variety × orthography — with a universal mapping from the standard language codes (ISO 639, Wiktionary) onto them.

^{xkcd #927, “Standards” by Randall Munroe,
CC BY-NC 2.5.}

Yes — we know. So this registry is deliberately not the fifteenth standard: it mints no new language codes. Every identifier is anchored in a code you already use, and the registry’s whole job is to write down how the fourteen existing standards line up — and precisely where each one stops.

The problem, in four codes

lat covers Plautus, Cicero, the Vulgate, and twelfth-century monastic charters alike — there is no ISO code for Old, Late, Medieval, or Neo-Latin.
grc (“Ancient Greek, to 1453”) spans Homer to the fall of Constantinople — a cataloguing convention, not a linguistic boundary.
lzh codes Literary Chinese, a written register used across two millennia, as if it were a historical stage.
la-vul names both attested Vulgar Latin (the Appendix Probi) and the reconstructed Proto-Romance of the comparative method — evidence and inference fused in one code.

The full argument — with a review of every prior system, from ISO 639-6’s rise and withdrawal to Glottolog, BCP 47, and Wiktionary — is in Prior art and the case for lects.

The answer, in one line

lect-id = anchor [ ":" stage ] [ "/" variety ] [ "~" script ] [ "@" ortho ]

Example	Meaning
`lat`	Latin, the whole span — honest coarseness
`lat:med`	Medieval Latin
`lat:late/ecc`	Late Latin, ecclesiastical register
`grc:koi`	Koine Greek
`zho/lit`	Literary Chinese (wenyan) — a register, not a stage
`roa:pro`	Proto-Romance — `mode: reconstructed` in the registry
`jpn:mod@kyu`	Modern Japanese in pre-reform (kyūjitai) orthography
`san~latn`	Sanskrit held as a romanized edition

Anchors are ISO 639 / Wiktionary codes, so every identifier stays legible to anyone who knows the standard codes. Stages carry date bands with their provenance recorded; genealogy is explicit parent: edges (roa: {parent: lat} — no pseudo-family nodes); reconstruction is a data field, never a naming convention; the ~ axis claims the script of the text as held (ISO 15924, lowercased, from a global table), never the artifact’s original writing system. Field-by-field semantics: the schema.

Integrating the registry in your tool

The registry is two plain YAML files, consumable with any parser:

https://raw.githubusercontent.com/arvicco/nabu-lects/main/lects.yml
https://raw.githubusercontent.com/arvicco/nabu-lects/main/codemap.yml

The identity default. codemap.yml lists only non-identity mappings. Any code not listed maps to itself as a bare anchor — so your tool needs no registry entry for a code to keep working, and adopting the registry cannot break codes it doesn’t mention.

Precedence. Layer your own knowledge above the universal defaults:

per-document override  >  per-collection override  >  codemap.yml  >  identity

The registry states what is true of a code for anyone; what is true of your holdings (this corpus’s lat is lat:med; that dictionary’s la-vul is roa:pro) belongs in your own override layer, keyed however your tool keys collections.

The one machine-load-bearing field. A stage with mode: reconstructed marks a comparative-method reconstruction. Key asterisk display, etymological-closure rules, or evidence filters on that field — never on the spelling of a tag or the suffix of a code.

Validating a vendored copy. If you vendor the files, run the bundled validator (plain Ruby, standard library only) against them:

bin/validate

It checks the identifier grammar, referential integrity (every mapping target and parent edge resolves), ord uniqueness, band sanity, Glottocode shape, and parent-graph acyclicity — the same script CI runs here.

Stability: pre-1.0 instability window — identifiers may be renamed or restructured without notice until the first tagged release. After v1.0.0, semantic versioning applies to the identifier space itself; see the README.

Attribution. CC BY 4.0: use freely, credit “nabu-lects” with a link to this repository.

The family

This registry was born from the holdings of Nabu — a personal research library of ancient-text corpora, roughly a million documents across more than 120 language codes — and is the frame that library describes its holdings with. Sister projects: nabu-data, the datasets the library publishes, and the School, the scribal-school reading environment built on top.

Contributions and corrections: issues and pull requests on github.com/arvicco/nabu-lects — every change passes bin/validate, and band or genealogy corrections are welcome with a source note.

This site is open source. Improve this page.