Keeping the change. Keep the reason.

Use file digests to detect changed bytes while keeping their narrow integrity role separate from source authenticity or scientific correctness.

The acquired rheology files retain repository checksums and independently calculated SHA-256 digests. NIST’s Secure Hash Standard describes algorithms for producing message digests used to detect changed bytes (NIST, 2015). A matching digest can support the custody of a particular file; it does not establish authorship, licence or scientific truth. The proposed manifest would therefore retain the source URL, repository version and rights record separately. A buyer inspecting the export could see which source bytes supported the calculation and which transformation version was applied, without mistaking a cryptographic check for validation of the experiment itself.

(NIST/SEMATECH, n.d.)

Separate extraction corrections from interpretation changes so that a revised comparison explains exactly which scientific quantity or assumption was affected.

A header correction may alter units without changing the source bytes. A revised interpretation may change which comparison is permitted without changing any numerical value. The proposed release record would classify those changes separately and record their consequences. A reader could then determine whether a prior result is numerically affected, scientifically narrowed or simply more clearly documented. The method would also preserve unresolved discrepancies instead of resolving them through presentation. That distinction matters to a buyer who has used a comparison in a development meeting: it identifies what needs reconsideration and prevents a harmless wording change from being confused with a new experiment.

Trace a correction through dependent outputs before deciding whether it requires a revised calculation, a notice or another experimental check.

A correction process should identify which outputs depend on the changed record and whether the buyer’s decision could be affected. The proposed system would retain a dependency map between raw inputs, derived values, plots and passports. It would not claim that every historical user can be notified without an agreed service and contact process. Public corrections could be documented in a visible version note; a commissioned delivery would define its correction responsibilities in scope. This makes curation credible while keeping operational commitments realistic. A historical dataset does not refresh itself, and a checked date should not imply continuous monitoring or automated scientific review.

Release a revised record for a stated source or processing change, preserving the earlier version that supported the original decision.

The next commitment should test the custody process on a deliberate parsing correction or a recoverable source-version difference. The team can then inspect whether the affected observations, calculations and exports are identified correctly. If that task fails, expanding the collection would multiply the problem. If it succeeds, new records can enter through the same finite review process with explicit versioning. A buyer should be able to answer what changed, which decision it could affect and how the earlier version remains reproducible. That makes source custody a useful scientific and commercial capability rather than an invisible implementation detail.