Source files, hashes and version history

A source pointer and a captured file differ

A URL identifies a location which may change. A saved file captures bytes obtained at a particular time. A note describing the page is a third object. A reproducible notebook links these together without treating them as identical. Retain the original location, retrieval time, file type and the portion relevant to the question. A failed retrieval is useful operational evidence but not a successful inspection of the desired source. Do not convert an error page into a financial record merely because it can be saved.

What a digest can establish

A cryptographic hash gives a compact identity for input bytes under a specified algorithm. Matching digests can help detect whether two copies have the same bytes. NIST publishes the Secure Hash Standard. A digest does not itself establish that the bytes are truthful, were published by a particular person or describe an actual trade. Keep byte identity separate from authenticity and financial validation. A fabricated document has a perfectly usable hash too, so the source and its meaning remain essential.

Version the interpretation as well as the artifact

A reviewer can change a conclusion after noticing a different event section, receiving a corrected statement or learning the intended accounting basis. Preserve the earlier source and explain the new interpretation. Do not silently replace a supposedly point-in-time note with today's knowledge. Where a comparison key changes, keep the old and new values with the reason. A version history is most useful when it records the actual operation performed, not a polished narrative of checks which never happened.

File time is not publication time

A local creation date can reflect when a file was copied, downloaded or restored. It is not automatically the date the source first existed. A dated page can also be edited later. Historical-publication claims therefore need their own appropriate evidence. In an ordinary reading notebook, distinguish source-stated publication date, inspection time and artifact-save time. Our new guides use the actual addition date and leave the inherited archive dates intact. They do not invent historical snapshots or backdated source captures.

Make exports travel with their scope

A downloaded worksheet can be detached from the page that explained it. Include input assumptions, classification and unperformed validation fields inside the JSON. The fictional packet explicitly says that its records are invented teaching inputs. Its calculations are reproducible on those inputs, not authenticated financial results. The export should retain IDs and missing values so another reviewer can see why a comparison was withheld. A portable number without the rule and source context is much easier to misinterpret.

Protect access while preserving traceability

A private statement may require lawful restricted access. Public documentation should not expose account credentials, access tokens or confidential attachments just to appear transparent. A controlled reference can identify the record to an authorised reviewer while public prose states the actual scope. Our browser tools run on entered data and do not upload the records to a service. A completed software calculation is not a certificate, an external timestamp verification or a financial audit. Those are different operations with separate evidence requirements.

An authored reading exercise

Save an authored teaching file, calculate its digest, and change one character in a copy. Different digests identify different byte content. They do not tell you which wording is correct or who authored it. Keep the original, the revision and the actual reason for change as separate records. If you later discover that you saved an error page instead of the desired document, preserve that retrieval failure accurately. Do not describe the digest as authentication of the unavailable financial source.

Primary reading

• NIST: Secure Hash Standard