Concepts and vocabulary#

The product’s first interaction is a place and a time. The richer vocabulary below exists to make research state precise, not to require a visitor to learn an ontology before starting.

Research objects#

Reference / event: A location and time around which an investigation is organized, with known coordinate conventions and any temporal/spatial uncertainty. A user-defined analytical reference need not be a catalogued mission event.
Research draft: Editable intent in the Lab: reference, windows, source-intersection mode, method choices, and related exploratory state. It has no scientific admission authority.
Observation: A source product with a durable identity and available acquisition metadata. Missing metadata is part of the object, not a value to invent.
Temporal role: PRE, POST, overlap, unknown, or outside the relevant window under the current reference. The same observation may have a different role in another investigation.
Candidate pair: A relation between one eligible before observation and one eligible after observation under a declared context. Pair construction precedes scientific screening.
Protocol: Explicit rules for query scope, partition, missing metadata, metrics, objective sets, and screening. The current Python family uses Pareto minimization; the broader product is not intrinsically limited to that family.
Scientific claim: A reviewable statement with Standing, Basis, limitations, and a defined evidence ceiling. It is not merely a number on a screen.

Source and population objects#

Source query: The exact archive scope and parameters requested. Changing the location or source time window requires a new query. Sorting a table does not.
Discovery snapshot: In the Lab, a bounded exploratory retrieval with response receipts and normalized rows. It is not automatically a canonical Python ArchiveSnapshot.
Canonical archive snapshot: A Python model tying source query, raw-response identities, normalized rows, counts, capture time, and closure. Compiler acceptance requires consistency across those values.
Query closure: Whether the retrieval witnesses establish completion within a bounded source query. COMPLETE, PARTIAL, and UNKNOWN refer to retrieval, not physical reality.
Currentness: What is established about the archive’s indexed-through horizon relative to the event. Query time and HTTP success cannot substitute for that horizon.
View population: The subset or projection currently rendered after filtering, paging, or axis selection. It must not masquerade as the complete source or scientific population.
Included population: The observations deliberately chosen for exploratory analysis. Selecting a row to inspect it does not automatically include or exclude it.

Evidence and authority#

ContextBasis: Information used for orientation or inspection, such as a lunar basemap. Its appearance in the interface does not make it an input to a historical Study.
StudyBasis: The retained evidence and assumptions on which a scientific record depends. New contextual information cannot silently rewrite it.
Artifact integrity: Agreement of supplied bytes with an expected identity and internal contract. It does not independently prove source authenticity or scientific truth.
Scientific Standing: The justified status of a particular claim under its evidence and limitations. A single global verified badge is inadequate.
Historical aggregate: A record preserving reported aggregate findings without the original row-level inputs needed for full replay. It must not gain synthetic rows in presentation.
Compiled candidate: A result produced by the canonical compiler from consistent typed inputs. Execution alone is not scientific admission.

The distinctions to remember#

Do not collapse	Why
Selected and included	Inspection and method input are different actions
View filter and source query	Presentation cannot silently change retrieval scope
Source response and scientific claim	Returned records still require normalization and interpretation
ContextBasis and StudyBasis	Better present-day context does not rewrite an earlier record
Hash match and scientific truth	Byte identity is a narrower proposition
Frontier membership and best image	Nondominance expresses tradeoffs, not universal quality
Fresh query and replay	A changing archive may return a different population
Public preview and published research	Deployment is not peer review or scholarly publication