Data collection and provenance

A screenshot can look stable while its provider changes labels or confirmations. A timestamp without a time basis can make an old event look like a new observation.

Treat each retrieval as a small evidence package: request method and exact URL, network, record identifier, block or slot reference, response status, actual UTC collection time, response bytes and a cryptographic checksum. Keep the provider’s event-time field separate. Store the data first, then derive a table; otherwise later readers cannot distinguish provider content from analyst edits. Never include credentials or private session material in a public package.

An identifier checksum reproduces bytes, not truth. A locally recomputed transaction ID proves that the supplied raw serialization matches that identifier; it does not alone prove canonical inclusion, finality or human identity. An indexer’s confirmed label has its own source boundary. Record missing pages, truncation, chain reorganization policy and unsupported event types rather than silently normalizing them away.

For derived values, state the unit and operation. Keep raw integers before decimal display. Wallet Intelligence can help inspect records, but this lesson creates no actual report and does not infer tool coverage from the interface. Provenance follows each claim all the way from stored response to calculation.

Retrieval time: when a response was collected.

Event time: timestamp declared by the event’s chain record.

Checksum: fingerprint of stored bytes.

Coverage: which records and event types a provider includes.

ILLUSTRATIVE: response R names block 500 with header time 12:00Z. It is fetched at 12:12Z and lists 2500000 raw units with six decimals. Provider P reports page 1 of 2; page 2 is unavailable. Derived display is 2.5 units, not 2,500,000 whole tokens. The fixture stipulates these values; no request was made.

Transaction timeline with two lanes: declared event time at 12:00Z and retrieval at 12:12Z. Attach raw-unit conversion and a conspicuous page-2 gap.

Caption: Illustrative provenance separates event time from the time a provider response was collected.

SPECIFICATION_ONLY — the rendered visual has not been produced or independently reviewed.

The fixture supports a reproducible unit conversion and a clearly bounded page-1 observation. Collection time and event time have different recorded meanings.

A matching checksum does not certify provider correctness. Missing page 2 prevents a complete-event-count claim; a header time does not establish exact wall-clock arrival.

Hashing a screenshot instead of preserving the response; replacing event time with retrieval time; stripping raw units; reporting a complete count from one available page.

Write a provenance row for R, compute its display amount, and rewrite “all events occurred at 12:12Z.” Include the unavailable page and one independent verification step.

• Keep exact identifiers and raw units.
• Separate event and retrieval timestamps.
• Save bytes before derivation.
• Disclose incomplete pagination and indexed finality.

Provenance binds records, retrievals and derived claims without converting a checksum into a truth certificate.

• Preserve response bytes and request context.
• Time basis and unit basis travel with each value.
• Partial coverage cannot support a complete count.

P10-L03

Use the named ZECOIN tool only as an evidence-reading context. This lesson creates no tool output, account session or entitlement. Offline exercise; do not sign, deploy, approve or fund anything.

• API: Esplora public API specification — GET routes, integer satoshis and indexed confirmation/outspend fields. Boundary: Mechanism documentation only; original illustrative values are stipulated, not observations. Historical records have separately frozen provenance and explicit collection gaps.
• BTC: Bitcoin developer guide — transactions — Input references identify previous transaction outputs; scripts do not identify a human. Boundary: Mechanism documentation only; original illustrative values are stipulated, not observations. Historical records have separately frozen provenance and explicit collection gaps.

ILLUSTRATIVE — entirely fictional offline inputs; no current observation.

null — no real observation

1

null

needs_review