Isolate claims
List numbers, dates, attributions, comparisons and promises that may influence a decision.

GUIDE / PROTOCOL
A strong claim must remain verifiable after publication and after reuse by search engines or AI.
Direct answer
It preserves the origin of each public fact, its context, validator and review date. It prevents hypotheses becoming certainties through repetition.
01 / Review method
List numbers, dates, attributions, comparisons and promises that may influence a decision.
Keep URL or document, author, date, relevant passage and known reuse rights.
Separate primary source, independent corroboration, interpretation and unconfirmed information.
Assign an owner, deadline and visible correction or withdrawal process.
02 / Evidence to keep
One unambiguous statement per row.
Durable link, title, author, date and verifiable passage.
Confirmed, corroborated, provisional, disputed or obsolete.
Publish, qualify, recheck or withdraw.
PRACTICAL CHECK
A frequent misreading and the check that helps avoid it.
A cited source does not prove an entire sentence. Its date, scope and wording may support a narrower fact than the published claim.
Link each important claim to the exact passage, keep a dated copy and separately qualify the source, inference and remaining unknowns.
AFTER THE REVIEW
Find the primary source and independent confirmation. If neither is available, qualify the claim as uncertain and state its scope.
Keep a record of the cited version, review dependent passages and publish an explicit correction if the conclusion changes.
Decision and review trail
What the check must help decide and what to retain so the choice can be reviewed.
Link each material claim to its exact source passage, date, saved copy and confidence level. A register home page does not let another reader repeat the check.
Publish only the precision the evidence supports. If a source is absent or conflicting, keep the hypothesis open and state the missing check.
OSINT / Official references
Tie each claim to a dated record, a version and an explicit reading limit.
Keep the series code and method behind a public statistic.
Limit to rememberKeep the series code, unit, coverage, method and revisions; different scopes are not directly comparable.
Read source entryCite the case, decision text and relevant paragraphs.
Limit to rememberRead the ruling and its reasoning; a summary or press release cannot replace them.
Read source entryIdentify the edition and page of a digitised historical document.
Limit to rememberA digital copy does not validate every statement in the document. Check provenance, OCR errors and resource-specific reuse terms; a BnF building photograph’s license does not apply to those documents.
Read source entryTypical situations for preparing a check. They do not describe completed assignments or actual observations.
Open the old URL and inspect destination, version and date of the recovered document.
Update the consultation link without erasing the historical URL; a redirect does not prove identical content.
Both URLs, inspection date and differences between versions.
Find the query, filters, date and passages surrounding the retained excerpt.
Add missing context; narrow the conclusion if original settings cannot be recovered.
Screenshot, reconstructed settings and remaining unknowns.
Compare the retained passage with the new publication and check its scope.
Revise or suspend the claim if evidence no longer supports it.
Versions, difference, decision and recheck date.
These questions complement the method. They help establish what is actually supported while retaining qualifications in the record.
Can a reviewer recover the document, version and used passage from the claim? Are repetitions of the same document marked as dependent?
Fictional example: the state applies only to the criterion and scope described.
Does the conclusion retain contradictions, covered period and missing evidence? Reopen the record when a source is corrected or replaced.
Fictional example: the state applies only to the criterion and scope described.
Does the digest identify the retained copy? Check origin and quoted passage separately: equal bytes do not validate a claim.
Documentation consulted : NIST · Secure Hash Standard ·
Fictional example: the state applies only to the criterion and scope described.
TOOLS / REGISTERS
Use one row per precise claim. Link a document supporting several claims to several rows, retaining the relevant passage separately for each claim.
03 / FAQ
Sometimes for a fact directly reported by its author. Disputed or decisive claims require suitable cross-checking.
No. Shareable evidence may be published; confidential material remains in a controlled perimeter.
It preserves a state, but also record the URL, date, author or publisher, context and verification method.
Keep both versions, date observations and explain the disagreement before reaching a conclusion.
When it no longer supports the current claim or when context, method or source has changed.
04 / Official references
DOCUMENTED REFERENCE
ETSI EN 319 142-1 V1.1.1 describes these PDF signature levels; European Commission DSS 6.5 documents augmentation. They distinguish material available for validation over time. A level name does not replace its validation result.
| Reference | Value or form | What it identifies |
|---|---|---|
| B-B | PAdES-BASELINE-B | Basic signature and attributes required by the profile. A signer-claimed time alone is not third-party trusted temporal evidence. |
| B-T | PAdES-BASELINE-T | Adds a trusted time indication establishing signature existence. Verify the token and its coverage rather than just the visible signature-box time. |
| B-LT | PAdES-BASELINE-LT | Incorporates long-term validation material, including certificates and revocation information. Presence does not remove the need to check validity and temporal relevance. |
| B-LTA | PAdES-BASELINE-LTA | Adds timestamps protecting validation material for continued use. Preservation may require further timestamps; it is not an unlimited one-off guarantee. |
Retain detected format and level, validator and version, policy, validation time and detailed report. Legal qualification is a separate analysis. LTA in a filename proves neither the actual level nor successful validation.
Not necessarily. Augmentation adds validation or preservation material to an existing signature. Retain original and augmented files and validate signatures and revisions rather than counting two content approvals.
Documentation consulted on .
DOCUMENTED REFERENCE
Start and end offsets are reproducible only when the text and counting unit stay the same. TextPositionSelector counts Unicode code points in logical order, including for right-to-left text.
| Reference | Value or form | What it identifies |
|---|---|---|
| start / end | [5,10) | The first position is zero. Start is inclusive and end is exclusive: [5,10) extracts five code points. |
| Unicode / UTF-16 | 🧭 · 1 / 2 | In “🧭 A: alpha; B: alpha.”, the initial symbol occupies one code point but two UTF-16 units. The first alpha is [5,10) in code points and [6,11) in UTF-16. |
| grapheme | e + U+0301 | W3C recommends avoiding grapheme splits. A combining accent or emoji sequence can contain several code points; the tool flags boundaries using the browser’s Intl.Segmenter. |
| resource changes | offsets ≠ version | Inserting text before a quotation shifts offsets. Preserve the extraction and resource state, and reread the actual selected text rather than trusting numbers alone. |
Original example with no real document: 🧭 A: alpha; B: alpha. The two explorers are independent.
Nonempty text: at most 2,000 code points; quotation: 1–200; each context: at most 100. No spaces, markup or accents are removed. Text fields use LF line endings.
Offsets follow logical text order. Right-to-left display does not reverse indexes.
These tools calculate in your browser and transmit no text.
Selection calculated
Text code points : 21 · Text UTF-16 units : 22
Showing the first 100 matches; refine the context to shorten the list.
| Start | End | Selected passage | Grapheme boundaries |
|---|---|---|---|
| 5 | 10 | alpha | Boundaries preserved |
Boundaries preserved
{"type":"TextPositionSelector","start":5,"end":10}Finding or rereading a passage validates neither the source nor the claim.
Pasted text is assumed to be extracted already. Browser text fields normalize line endings to LF; no other normalization is added. Empty ranges are explicitly shown. This reader is not a full Web Annotation validator.
String.length and normal JavaScript indexes use UTF-16. Array.from splits code points, not graphemes. The correct unit depends on the contract; Web Annotation offsets cannot be used directly as JavaScript indexes.
Documentation consulted on .
DEEPER READING / PRACTICE
Documents are stored in a file, but their connection to the conclusions has been lost. A register makes it visible why a piece was collected and what it allows to establish.
Updated on .
Attribute an identifier, indicate origin, date, version and supported proposal. Separate the stored file, the researcher's notes and the changes that occur after consultation.
A speaking file name is not enough to find the collection conditions. An identical hash demonstrates a content identity, not the authenticity of the document.
The register includes reference, path or URL, context, verification status and use. Another person must be able to find the room and understand his place in reasoning.
PRACTICAL CASE / WORKED EXAMPLE
The example below is pedagogical. It shows how to move from a document to a qualified proposal, preserving the context, the contradictory elements and the information that the research does not yet allow to establish.
A file contains many correctly named files. Several months later, no one knows what proposals they supported or under what conditions they were obtained.
Attribute identifiers and link the documents to the questions of the file. Information about origin, version and observation. Separate the notes written after collecting the content of the room.
The register returns the role of each element. The documents unnecessary to reasoning remain identified as an abandoned context or track, rather than being presented as indistinctive evidence.
This can be useful in explaining a distant track or a contradiction. Its status and role must be clear so that it does not believe that it supports the final conclusion.
DETAILED WORKSHOP / APPLICATION
Collect files, references and collection notes. Choose an identification agreement and the fields useful to find the pieces and understand their role.
Attribute an identificator, indicate source, version, date and supported proposal. Separate document, observation and comment. Check whether the routes or links lead to the right item.
If a piece has been removed, keep its status and the useful reason. It may document a contradiction or a track without being part of the evidence of the conclusion.
The register is used when someone can find a piece and explain its use. Ranking by file name alone is not enough for this traceability.
Transmit index, conventions and conservation limits. The version changes are linked to the old pieces and the conclusions that used them.
ANNOTATED WORKING DOCUMENT / REVIEW GRID
The index keeps the identification, origin, version and proposal supported. The researcher's notes remain separate from the pieces to explain the place of each element in the file.
On a small screen, scroll the grid horizontally to read all three columns.
| Observation | Supported interpretation | Next step |
|---|---|---|
| The room and its use are found. | The evidence can be reviewed with its collection context and its contribution. | Keep the references and check the version changes. |
| The file exists without any identified reason for use. | Storage is not enough to rebuild the reasoning. | Connect the room to a question or qualify it as a context. |
| A dismissed piece remains presented as support. | The register may give a higher evidence impression than the actual file. | Indicate his status and the reason for his withdrawal from reasoning. |
The final verification verifies that each item can be found and understood without depending on the editor's memory. Provisional uses and replaced parts must remain visible in the history.
REPORTING / EXAMPLE WORDING
A preserved file must keep the reason for its presence in the file.
Relate file, source and use without depending on the name of the document.
The document supports the proposal described and can be found by its identifier. The contextual parts and distant tracks retain a distinct status.