nabu-places

Prior art — and the gap this registry fills

The ancient-world gazetteer landscape is rich; nabu-places adopts it rather than competing with it. What follows is why a decisions layer is still missing, and what each neighbor actually offers.

The gazetteers (adopted as data)

The infrastructure neighbors

The gap

Every corpus that ships place names without machine refs forces each consumer to redo the same matching — and the hard cases (homonyms like Mediolanum ×6, ?-qualified proveniences, genuinely unlocatable sites, house composite formats) are exactly where silent heuristics go wrong. Reconciliation tools produce these judgments but don’t publish them as small, reviewable, reusable data.

nabu-places is that missing artifact: the decisions themselves, keyed by the source’s verbatim strings, versioned, validated, and licensed for reuse — the nabu-lects registry pattern (identity-default, closed vocabulary, honest non-answers) applied to place identity.

And one step beyond glue: when scholarship over the texts themselves — joins across itineraries, enrichment, AI-assisted argument — establishes a place no gazetteer registers at all, the registry mints its own evidence-bearing record (places.yml, namespace np:) rather than forcing the claim into someone else’s id space or losing it. Gazetteers remain the authorities for what they cover; the native lane exists for what they don’t yet.