What we do not know

Heritage documentation fails in a particular way. A figure is estimated, then quoted, then cited and within three publications it has become a fact that nobody can trace back to a measurement. The damage is not that the first estimate was wrong. It is that the record lost the ability to say how it knew.

Vaayil’s answer is structural rather than editorial.

Unknown is a value

Every field in a record that can be unknown is nullable and an unknown field renders clearly as not recorded, not as a plausible guess and not as a hidden row. If a record does not state a height, it is because the height has not been established.

Confidence is recorded with the claim

Dating carries its own confidence: secure, probable, contested or unknown. A contested attribution is published as contested, with the reason, rather than resolved silently in favour of the more attractive one.

Condition is recorded part by part, each with the date it was assessed. A building assessed in December 2025 is described as it was in December 2025; condition is a reading and readings expire.

Every record declares a gap

The schema requires at least one entry in What we do not know before a record can be published. This is enforced by the build, not by good intentions: a record with an empty uncertainty list fails validation and never reaches the site.

In practice these entries say things like: the sanctum interior was not entered; the attribution is not established by inscription on this structure; the survey provenance is incomplete and is being recovered from the original project files.

Versions

Records carry a version. When a gap is closed, such as when a tolerance is established or an inscription is read, the record is revised, the version increments and the URL does not change. A citation made against version 1.0 remains resolvable.

Real content is more valuable than a complete-looking form.