Execution · NO — NOT EXECUTED
- Overview
- Discovery Clock
- Institutional Clock
- Authority Surface
- Reversibility
- Decision Boundary
- Known Risk Dwell
- Friction
- Defender's Window
- Case Studies
- Comparison
- Evidence
- Trace
- Accountability
- Counterfactual
- Validation
- Readiness
- Methodology
- Theory
Methodology
Discovery is not authority. Prediction is not evidence. Recommendation is not execution. An exploit attempt is not proof of compromise. Approval is not sufficient without scope, verification, and operational constraints. Public knowledge is not organizational awareness. Delay is not proof of negligence. Dependency is not proof of failure. Technical capability is not formal authority. Absence of evidence is not evidence of inaction. Disclosure is not organizational awareness. Patch availability is not patch deployment. Traceability is not accountability. Research reviewability is not a finding of legal liability. Possible is not available. Available is not feasible. Feasible is not preferable. An unobserved outcome cannot be directly verified. Counterfactual analysis is not a negligence finding. Technical feasibility is not execution authority. Agreement is not truth. Methodological reliability is not execution authority.
01
Observation versus inference
An observation is a retrieved source statement or a corroborated public fact. An inference is a derived or modelled claim. INFERENCE never silently enters a verified public clock. SYSTEM_INFERENCE remains labelled as such.
02
Public versus synthetic records
PUBLIC records are source-backed. SYNTHETIC records demonstrate institutional clocks. They share CVE identifiers only as inherited public facts. Synthetic timestamps cannot authorize a real environment and are never presented as observed metrics.
03
Date precision versus timestamp precision
A source that names a calendar day does not establish a time of day. Date-only instants stay date precision. They are not converted into midnight event times. Mixed date/datetime intervals are date-bounded, not confirmed measurements.
04
Discovery-to-disclosure intervals
Intervals are end minus start in UTC. Negative values are valid and labelled INVERTED rather than rewritten. Missing start or end remains NOT ESTABLISHED, never zero. Incomplete evidence cannot produce a confirmed measurement.
05
Exploitation attempt versus compromise
ATTEMPTED_EXPLOITATION, CONFIRMED_EXPLOITATION, and CONFIRMED_COMPROMISE are not synonyms. A honeypot attempt is not successful exploitation of a production system. Successful exploitation of a target is not confirmed compromise of a named organization.
06
Decision latency
Decision latency is organizational detection to authorization. On the public PHP record both ends are null, so the interval is NOT ESTABLISHED. Synthetic decision latency is labelled inference and is not an observed metric.
07
Authorization versus execution
Recommendation, authorization, and executionStatus are separate fields. ALLOW is an advisory classification. Execution status is always NOT_EXECUTED. VRB maps onto evaluateBoundary and does not ship a second engine or any execution pathway.
08
Reversibility versus recoverability
Reversibility asks whether an action can be undone. Recoverability asks whether service can be restored. UNKNOWN does not count as verified. Backup existence does not prove restore capability. Rollback existence does not prove successful recovery. Restore testing is recorded separately.
09
Human review responsibilities
A reviewer must keep missing evidence missing, refuse to treat severity or confidence as permission, and refuse to treat approval as sufficient without scope, verification, and operational constraints. Production-changing actions remain HOLD or DENY in v0.1.1 even when an approval record exists.
10
HOLD versus DENY
Missing evidence, unverified scope, missing approval, or unresolved reversibility is HOLD. Explicit prohibition, EXCLUDE evidence, or authorityGrant NONE is DENY. The adapter does not weaken global evaluateBoundary rules.
11
Evidence classes
Public claims are labelled VERIFIED_PUBLIC_FACT, SOURCE_STATED, INFERENCE, UNVERIFIED, or SYNTHETIC. These annotate existing EvidenceStatus values (OBSERVED, STATED, INFERENCE, VERIFY, EXCLUDE). A listed URL is not by itself proof of every claim.
12
Future CUSTOMER_STATED access control
Public routes and static generation must not ship private organizational incident data. CUSTOMER_STATED records, if ever ingested, require an access-control boundary that is not implemented in v0.1.1. Do not place customer evidence in public static bundles.
13
Known limitations
v0.3 is advisory. It does not scan networks, install patches, isolate hosts, or shut down services. GitHub Security Advisories referenced by PHP were not independently retrieved in v0.1.1. CERT-EU and PHP Foundation disagree by 30 minutes on release time; that conflict is preserved. No light theme is provided. Browser layout acceptance is documented separately.
14
Knowledge scope
PUBLIC_KNOWLEDGE is not ORGANIZATION_CONFIRMED_KNOWLEDGE. VENDOR_KNOWLEDGE describes a maintainer or vendor. INFERRED_KNOWLEDGE cannot authorize a dwell calculation. UNKNOWN stays unknown.
15
Known Risk Dwell
Dwell is confirmed awareness by a defined institution to a documented disposition. Disposition may be mitigation, formal risk acceptance, transfer, or retirement. Acceptance is not patch completion. Missing awareness or disposition is NOT ESTABLISHED. No disposition with a cutoff is RIGHT_CENSORED / STILL_UNRESOLVED. Public disclosure is never substituted for awareness.
16
Decision friction
Friction factors require supporting evidence or an explicit hypothesis label. Delay is not classified as negligence, RESOURCE_CONSTRAINT, or AUTHORITY_AMBIGUITY merely from elapsed time. Multiple factors and UNKNOWN are allowed.
17
Defender's Window
PUBLIC, ORGANIZATIONAL, and SIMULATED windows are separate. Negative values mean the warning arrived after the observed attempt. Date-only observations produce bounded or NOT ESTABLISHED results, not exact event times.
18
Comparative case methodology
Cases are compared only after an explicit comparability assessment. PUBLIC and SYNTHETIC records may appear in one labeled mixed view; they are not one empirical population. Right-censored durations are not treated as completed. Unknown is not converted to zero. Sorting is not a ranking of institutional performance.
19
Dependency direction
source → target means the source depends on the target for the specified decision or response activity. Graph proximity does not imply control. A dependency graph never grants permission.
20
Bottleneck classification
DOCUMENTED_BOTTLENECK requires supporting evidence. MODELED_BOTTLENECK is synthetic or hypothesized structure. POTENTIAL_BOTTLENECK is incomplete. NOT_ESTABLISHED is the default when graph evidence is missing. Topology does not prove causation, delay, or negligence.
21
Readiness verification
Readiness responses are VERIFIED, PARTIALLY_VERIFIED, UNKNOWN, or NOT_AVAILABLE. Self-attested VERIFIED is not independently audited. No numerical readiness score is computed. The assessment is local, synthetic in v0.3, and cannot authorize action.
22
Export restrictions
Research export schema vrb-research-export/0.3 is SYNTHETIC-only. Public cases are omitted with an explicit exclusion manifest. Secret-looking fields are refused. executionStatus is always NOT_EXECUTED. No telemetry, remote upload, or server storage. Schemas vrb-empirical-export/0.4A, 0.4B, 0.4C, and 0.5 remain buildable. Schema vrb-empirical-export/0.6 adds historical decision points, possibility dimensions, feasibility assessments, constraints, and counterfactual hypotheses. Synthetic counterfactual fixtures are labeled separately and are never mixed into the public empirical export.
23
Empirical source hierarchy
Vendor advisories, official notices, government advisories, maintainer disclosures, and first-party statements outrank third-party analysis. A retrieved URL does not verify every attached claim. Source credibility does not eliminate claim-level verification.
24
Claim verification and event reconstruction
VERIFIED requires retrieved supporting sources and a precise supported statement. SOURCE_STATED records what a source says. CONTESTED preserves contradictory statements. UNVERIFIED and NOT_ESTABLISHED remain unlabeled as facts. Publication date is kept separate from the described event date. Date-only evidence stays date precision.
25
Observability versus performance
The Observability Audit records whether public evidence exists for discovery, vendor awareness, public disclosure, organizational awareness, acknowledgement, authorization, mitigation start, mitigation completion, and recovery verification. Low observability is not a performance score. Absence of evidence is not evidence of inaction.
26
Empirical comparability
Incident class, disclosure model, observation completeness, and precision must match before comparison. Public empirical candidates are not merged with synthetic scenarios. Descriptive differences are preferred over numerical ranking. Causation is not inferred from temporal association.
27
Source retrieval versus chronology
v0.4B records HTTP status, retrieved title, and failure reason. A DNS or 404 failure cannot be marked retrieved. A marketing page is not an incident postmortem. A date in a URL is not a publication timestamp. Email Date headers, when retrieved, are exact timestamps of the message, not invented discovery times. Government acknowledgement is not named-operator awareness.
28
Event lineage and decision traces
v0.4C reconstructs lineage from existing events only. EVENT_PRECEDES_EVENT is temporal order, not causation. Transmission is not receipt. Receipt is not necessarily awareness. Action is not proof of authorization. Restoration is not verified recovery. Missing links are research gaps, not institutional failure. Partial public traces do not promote case maturity. Traceability is not execution authority.
29
Institutional accountability evidence
v0.5 asks what must be established before responsibility can be evaluated. REVIEWABLE, PARTIALLY_REVIEWABLE, EVIDENCE_GAPS, NOT_ESTABLISHED, and OUT_OF_SCOPE describe evidence sufficiency for a defined research review. They are not liability, negligence, fault, or wrongdoing findings. Traceability is not accountability. An available patch is not necessarily an available institutional action. Later knowledge must not be projected backward. Responsibility cannot be inferred from chronology alone. Research reviewability is not a finding of legal liability.
30
Duty, authority, and alternatives
Statutory and regulatory duties require jurisdiction, effective-date, entity-scope evidence, and human legal review before applicability is treated as an established legal conclusion. Internal policy is not law. Job title does not prove legal obligation. Formal authority, technical capability, operational capacity, resources, and dependency constraints remain independent. Hypothetical alternatives stay hypothetical. Public cases treat only evidence-backed options as historically available.
31
Ex-ante versus ex-post evidence
Publication time is not recipient awareness. A report published after a decision cannot automatically explain that earlier decision. Date-only overlapping intervals cannot establish exact availability order. Later evidence may be retained for retrospective analysis and must remain visibly separated. Hindsight is not proof of negligence.
32
Historical decision points
A PUBLIC decision point requires documented institution-specific evidence. If the time or institution is unknown, the record remains a research question with NOT_ESTABLISHED status. A decision-maker identity is never invented. Date-only precision is preserved. Conflicting timestamps remain conflicting. PUBLIC and SYNTHETIC decision points stay separate.
33
Possibility, availability, feasibility, and outcome
TECHNICAL_POSSIBILITY, HISTORICAL_AVAILABILITY, INSTITUTIONAL_FEASIBILITY, and OUTCOME_UNCERTAINTY remain independent. Possible is not available. Available is not feasible. Feasible is not preferable. Preferable in hindsight is not proven preferable at the time. An unobserved outcome cannot be directly verified. These dimensions are not collapsed into a score.
34
Institutional feasibility as research classification
EVIDENCE_SUPPORTED, CONDITIONALLY_FEASIBLE, HYPOTHETICAL, NOT_ESTABLISHED, and CONTRADICTED_BY_EVIDENCE describe documented prerequisites. They are not ALLOW, HOLD, or DENY. EVIDENCE_SUPPORTED means documented prerequisites were satisfied in a narrow scope. It does not mean the action was performed, optimal, or that a better outcome would follow. Technical capability is not formal authority. A patch or tool is not operational readiness.
35
Counterfactual scenarios and tradeoffs
A counterfactual is not an observed historical event. UNKNOWN_BASELINE is used when historical action records are missing. Simulated outcomes stay hypothetical. No pseudo-precise probabilities, avoided-damage estimates, or certain prevention claims are produced. Documented tradeoffs clarify constraints. They do not recommend an optimal historical choice. Counterfactual analysis is not a negligence finding and does not rewrite accountability readiness.
36
Independent annotation and inter-rater reliability
v0.7 freezes public metadata and bounded claims into a hashed snapshot. Independent reviewers classify one task at a time under vrb-annotation-protocol/0.7. Agreement is not truth. Consensus is not independent verification. Unknown is a valid research outcome. Disagreement is interpretive variation, not reviewer failure. Reliability statistics are computed only on matched units, by task type, after protocol and snapshot compatibility checks. Synthetic fixtures and AI-assisted labels are not independent human judgments. Local blinding is a workflow feature, not an adversarial security boundary. Agreement never maps to ALLOW or DENY.