ONTOLOGY PROTOCOLS
08. Ontology Development and Verification Protocol
Revision 1.3 · synchronized version
1. Purpose
The Ontology Development and Verification Protocol establishes the rules by which the system may change its own semantic content, structure of relations, and architecture of representation while preserving the ability to establish: what exactly changed; why the change was made; whether semantic continuity was preserved; which elements of the system were affected; on what grounds the change was accepted; what checks it passed; and what the previous state of the system was. The Protocol does not determine what the content of future knowledge should be. It determines the conditions under which a change in knowledge may be represented as controlled development rather than hidden substitution of previous content. Change ≠ development. In this Protocol, verification means checking integrity, traceability, semantic continuity, transmissibility and, where applicable, justification and correspondence with external data. Verification is not a guarantee of truth.
2. Boundary of the Protocol
Documents 01–07 successively establish the architecture of meaning, reasoning, reference, ontological analysis, graph, concept, and machine representation. Document 08 considers not a separate element of this architecture but change of the system itself over time. Its subject is the transition: Ontology in state n → Ontology in state n+1. This notation denotes a sequence of system states but says nothing by itself about the quality of the change. A transition may be a correction, refinement, expansion, narrowing, redefinition, splitting, merging, abandonment of a previous construction, architectural transformation, or metamorphosis. Every change must therefore first be distinguished and only then assessed as a particular type of change.
3. State of the Ontology
At every considered moment, at minimum, the current canonical state, the history of previous states, and the space of proposed changes must be distinguished. A proposed change is not part of the canonical Ontology merely because it has been formulated. The canonical state is likewise not true merely because it is canonical. Proposed ≠ Canonical ≠ True. These are three different characteristics.
4. Change Event
A change event is a recorded transition in which at least one significant element of the system’s representation or architecture changes. It must be established what exactly changed. A change may affect a machine record, wording, content, boundaries of meaning, relation, reference, assertion, truth status, epistemic status, classification, architectural role, linguistic representation, type of semantic unit, representation rules, or the architecture of the Ontology itself. Change at one level must not automatically be interpreted as change at another.
5. Classes of Change
For analysis of development, at least the following working classes must be distinguished. Technical correction — a change of record without a presumed change of represented content. Refinement — increased definiteness of representation while preserving the principal semantic distinction. Expansion — inclusion of previously unrepresented aspects in the content. Narrowing — restriction of previously broader content. Redefinition — change of definition capable of affecting the boundaries or structure of a concept. Splitting — emergence of several independent semantic units from a previously unified semantic line. Merging — establishment of sufficient grounds for reducing previously distinguished units to one semantic line or creating a new unifying construction. Deprecation — termination of canonical use of a record or construction without destruction of its history. Architectural change — change in the very method of organizing content. This classification is working and is not declared finally closed. A technical correction is considered non-semantic only provisionally. If verification shows that it changed represented content, boundaries of meaning, reference, status, or essential relations, the change must be reclassified as substantive.
6. Change of Record and Change of Meaning
Protocol 07 establishes the distinction: change of record ≠ change of content ≠ change of semantic identity. Protocol 08 makes this distinction the basis of the development procedure. A new record revision may leave content unchanged. A change of content may preserve semantic continuity. But a change may reach such depth that the assertion of preservation of previous semantic identity requires separate justification or becomes untenable. A version number therefore cannot serve as a criterion of semantic identity.
7. Semantic Continuity
Semantic continuity means that there are sufficient grounds to consider the changed representation a continuing representation of the same distinguishable semantic unit. Verification must take into account not only the name or wording of the definition, but also canonical content, boundaries of meaning, essential semantic distinctions, architectural role, essential relations, referential characteristics where applicable, and the history of formation of the content. None of these features alone is established as a universally sufficient criterion. Within this Protocol, semantic continuity is considered preserved when the change does not destroy the essential distinction by virtue of which the unit was distinguished as an independent semantic unit and sufficient continuity of its canonical content, boundaries of meaning, and essential relations is preserved. If this cannot be established on available grounds, the regime of unestablished identity defined in §9 applies. Semantic continuity is a subject of verification, not a consequence of a technical decision.
8. Semantic Unit ID and Continuity
Semantic Unit ID addresses presumed semantic identity but does not create it. Preservation of a Semantic Unit ID is therefore admissible only when the system has sufficient grounds to continue treating the new state as a representation of the same semantic unit. Preservation of the ID itself is not proof of such grounds. Justified semantic continuity may be grounds for preserving Semantic Unit ID; preservation of Semantic Unit ID is not grounds for semantic continuity. If semantic continuity has not yet been established, the previous Semantic Unit ID must not automatically be transferred to new content as confirmation of identity. At the same time, the system is not obliged to prematurely create a new Semantic Unit ID. An explicit link recording the open identity question must be preserved between the representations until grounds permit a decision. An assertion of preservation or violation of semantic continuity is itself a separately addressable assertion. Where applicable, applicability of truth evaluation, truth status, grounds, and epistemic status are recorded separately for it.
8.1. Continuity in the Five CCO Identity Spaces
The rules of semantic continuity, splitting, merging, and preservation of identity are applied separately to the five independent but interconnected spaces of persistent CCO identity: Object ID ↔ Class ID ↔ Domain ID ↔ Semantic Unit ID ↔ Semantic Structure ID. A change of identity in one space is not automatically transferred to another. A change of Semantic Structure does not mean a change of the Semantic Units included in it. A change of Class or Domain does not mean a change of a Concept or another Semantic Unit. A change of an Object that is the carrier of a source does not automatically change the content of the associated Semantic Unit. For each space, preservation of identity, split, and merge must be established on its own grounds. Concept ID remains a specialized Semantic Unit ID.
9. Unestablished Identity
If available grounds are insufficient to decide whether semantic identity is preserved, the system must be able to leave the question open. It is not required to force a choice between “the same unit” and “a new unit.” The state “semantic continuity not established” is admissible. Until the question is resolved, the system must preserve the necessary distinctions and the history of both representations. Indeterminacy of identity is not grounds for hidden merging.
10. Splitting a Semantic Unit
Splitting becomes necessary when a previously unified representation reveals several distinguishable contents whose continued joint storage leads to substantial semantic loss or conflation. Before splitting, it is necessary to establish what content existed before the split; what distinctions were discovered; why they cannot be adequately represented within the previous unified unit; what new units arise; and what continuity relations exist between them and the previous unit. The history of the original unit is preserved. Splitting must not rewrite the past as though the new distinctions had always been represented by the system in their present form.
11. Merging Semantic Units
Merging is admissible only when there are sufficient grounds to consider that the previous distinction no longer corresponds to the established semantic structure or that several units should be represented by a new common construction. Coincidence of names, individual definitions, referents, classification position, or machine characteristics is not sufficient grounds for merging. Content, boundaries, relations, reference, and history of distinction must be investigated. If identity is not established, the distinction is preserved.
12. Merging and Creating the New
Two cases must be distinguished: A and B are recognized as two representations of one semantic unit; A and B are used to create a new semantic unit C. In the second case, the emergence of C does not automatically destroy A and B. Merging records ≠ establishing semantic identity ≠ creating new synthetic content. Each operation must be represented separately.
13. Metamorphosis
Metamorphosis differs from ordinary change in that not only a state within the previous architecture of distinctions is transformed, but the architecture of the space of possible states itself. Ordinary development occurs within the existing space of semantic possibilities. Metamorphosis changes the very space in which further distinctions become possible. Working form: Ω_SEM⁽ⁿ⁾ → Ω_SEM⁽ⁿ⁺¹⁾. Here Ω_SEM denotes the space of semantic distinctions of the Ontology. This notation is not an assertion about metamorphosis of Reality itself. The formula does not define metamorphosis by itself; it compactly records the distinction already expressed in words. A necessary sign of metamorphosis within this Protocol is the impossibility of adequately representing new content in the previous space of distinctions without changing the architecture of possible distinctions itself. This sign is not declared a final exhaustive criterion and is subject to testing as the Ontology develops in practice. Qualification of a particular change as metamorphosis is itself an assertion and requires grounds. Metamorphosis may be local: change in the space of distinctions of one domain does not automatically mean metamorphosis of the entire Ontology. In metamorphosis, the Ontology architecture version must change; Semantic Unit IDs change only where semantic identity is not preserved.
14. Metamorphosis and Ordinary Version
Metamorphosis must not be hidden under an ordinary increase of a version number. If the previous architecture can no longer represent new content without substantial loss, the system must record a change at the architectural level. At the same time, not every architectural change is metamorphosis. For example, technically adding a field to a machine schema may merely implement a distinction that already existed in the semantic architecture. It must therefore be established whether the technical means of representation changed or the space of semantic distinctions itself.
15. Architectural Change
An architectural change affects the general rules of organization of the Ontology. It may include introduction of a new type of semantic unit, a new relation type, a change in the status system, reference rules, identity principles, graph architecture, machine representation, or the development protocol itself. An architectural change requires broader verification than a change to an individual concept because it potentially affects many existing elements.
15.1. Change of Schema and Change of CCO Architecture
It is necessary to distinguish a change of CCO content or schema from a change of the architecture of CCO itself. Adding a new concept, statement, relation, record field, or technical means of representation is not automatically an architectural change if it is implemented within the already established architecture of distinctions. An architectural change of CCO is a change of its fundamental invariants: introduction or elimination of a basic identity space; change of the relations between Object, Class, Domain, Semantic Unit, and Semantic Structure; change of the principle of independence of these spaces; change of the basic distinction Relation ≠ Statement; or another transformation that changes the very means by which established semantic distinctions are preserved. Such a change requires a separate Ontology Architecture Version. A change of Content Version, Record Revision, or Relation History does not by itself create a new Architecture Version.
15.2. Spatial Reprojection and Semantic Identity
A spatial model projection, defined by Protocols 07 and 09, is a means of representing an already distinguished semantic or structural unit within an explicitly specified version of a geometric model. It does not enter into the fundamental semantic identity of a Semantic Unit and does not itself form a new space of persistent CCO identity. Change of Spatial Projection ≠ change of Semantic Unit. A change of coordinates, Spatial Object type, geometric model, model version, or other parameters of a spatial projection is not by itself grounds for changing a Semantic Unit ID, splitting, merging, or recognizing a new semantic unit. The question of semantic identity is reconsidered only when the change of projection is accompanied by an independent change of essential semantic content, boundaries of meaning, architectural role, or other characteristics that, under this Protocol, can actually affect semantic continuity. If a previous spatial projection is recognized as erroneous, incomplete, or less adequate to a new geometric model, the system must record the change of projection and its grounds while preserving the history of the previous state. Correction of a model projection must not be retrospectively represented as a change of the semantic unit itself. Consequently: Spatial Reprojection is a change of model representation. Semantic Identity Review is required not by the fact of reprojection, but only when there are independent grounds to consider that an essential semantic distinction has changed.
15.3. Boundary of Revision 1.3
This clarification does not change the five-space CCO architecture, the criteria of semantic continuity, split/merge, or metamorphosis. It only explicitly extends the already operative principle “change of representation ≠ change of content ≠ change of semantic identity” to spatial model projections.
16. Change Proposal
Any substantive change first exists as a change proposal. The proposal must be separated from the current canonical structure. At minimum it must make it possible to establish what is proposed to change; what current state is affected; what new state is proposed; why the change is necessary; what distinctions it creates or removes; and what dependencies it may affect. The existence of a proposal does not itself change the canonical Ontology. A change proposal must be an independently addressable object with provenance and management status. Management status of a proposal — for example proposed, under review, accepted, rejected — is not a truth, epistemic, or ontological status of its content.
17. Grounds for Change
A substantive change must have representable grounds. Depending on the type of change, grounds may include a discovered contradiction; new observation; new data; an explanation with a broader confirmed scope of applicability, greater explanatory sufficiency, or better justification under explicitly established criteria; discovered semantic ambiguity; inadequacy of an existing classification; inability to represent new content without loss; discovery of hidden conflation of levels; an established error; need to refine reference; result of a semantic stress test; or another explicitly represented reason. The presence of grounds does not yet mean that the proposed change is correct. It means that the change has a verifiable reason for consideration.
18. Source of Change and Grounds for Change
The source of a proposal is not its grounds. An author, model, researcher, external document, or automated system may be the source of a proposal. Authorship does not establish correctness of the change. Source of proposal ≠ grounds for change ≠ truth of result. This distinction must also be preserved in the machine record.
19. Impact Analysis
Before canonization of a substantive change, its possible impact on related elements must be established. Dependent concepts, definitions, assertions, relations, classifications, referential links, epistemic statuses, causal projections, linguistic representations, machine records, external mappings, and other protocols may require verification. A change in one element must not covertly change the meaning of other elements. Absence of a registered dependency link is not proof of absence of dependency. Impact analysis must take into account possible incompleteness of the dependency graph itself.
20. Change Cascade
One substantive change may require several related changes. Such a process must be represented as an explicit cascade rather than an unnoticed automatic rewriting of dependent elements. For every element of the cascade it is necessary to distinguish: a directly accepted change; a logically or architecturally required change; a merely presumed change; and a change left open. Cascade dependency does not imply automatic truth of every subsequent change.
21. Contradiction After Change
If new content contradicts existing content, the system must not automatically delete one side. It must first establish whether the contradiction is genuine; whether the assertions concern the same context; whether the level of description is the same; whether the meanings of terms coincide; whether the same truth criteria apply; and whether the grounds of one assertion have changed. Until analysis is complete, conflicting assertions may coexist with explicitly designated statuses. Discovery of contradiction ≠ automatic resolution of contradiction.
22. Versioning
It is necessary to distinguish: machine-record revision; version of represented content; change of semantic identity; version of the Ontology architecture. These types of versioning may be related but must not merge. A new record revision does not necessarily mean a new content version. A new content version does not necessarily mean a new semantic unit. An architectural change does not necessarily mean a change in all existing semantic units. Content version denotes the version of the canonical semantic profile and must be distinguished from Record Revision — revision of a specific machine record.
23. Reversibility
The system must preserve the ability to reconstruct a previous canonical state of the Ontology. Reversibility does not mean an obligation to return to it, but the ability to establish what existed; how it was represented; why it changed; what change occurred; and on what grounds the new state was accepted. History is part of the system’s verifiability. Returning to previous content does not erase intermediate revisions or turn history into the appearance of a continuously unchanged state.
24. Abandonment of Previous Content
Even if a previous proposition is recognized as erroneous, obsolete, or unsuitable, it must not disappear from the system’s history as though it never existed. “No longer used as canonical” and “never existed in history” must be distinguished. Correcting an error does not require destruction of the fact of the error. Preserving history makes it possible to investigate the mechanism by which the error arose and to prevent repetition.
25. Internal Verification
Before canonization, a change must pass checks appropriate to its nature. Internal verification may include: semantic verification — whether necessary semantic distinctions are preserved; logical verification — whether rules of reasoning are violated; referential verification — whether what the assertion concerns is correctly established; epistemic verification — whether status corresponds to strength of grounds; control of the ontological status of the assertion — whether the structure of representation has been turned without separate grounds into an assertion about the structure of Reality itself and whether the assertion has been assigned a stronger ontological status than its grounds permit; architectural verification — whether the change is consistent with the current protocol system; machine verification — whether the change is technically represented correctly. Successful completion of these checks is not independent proof of the truth of an assertion.
26. Cross-Protocol Verification
A fundamental-level change must be checked against the system’s current protocols. For the present architecture, this means checking against Documents 01–07. The purpose is to establish whether the change causes a hidden violation of distinctions introduced earlier. Particular attention must be paid to conflation of meaning and referent; assertion and state of affairs; truth and epistemic statuses; representation and Reality; concept and definition; Semantic Unit ID and semantic identity; record and content; canonicity and truth. Current protocols are not immutable dogma. If the change under review reveals a justified defect in an earlier protocol, the protocol itself becomes an object of change and the consequences of that revision must be explicitly propagated through dependencies.
27. Semantic Stress Test
After internal verification, substantial new content may undergo an independent semantic stress test. An external system is given the published content without hidden authorial explanation and asked to reconstruct the meanings of key concepts, boundaries of their meaning, relations among them, statuses of assertions, open questions, and distinctions among established, hypothetical, and unknown. The reconstruction is compared with the canonical meaning. The purpose is to test transmissibility of meaning, not truth of the theory under investigation. The result of an individual stress test also depends on the capabilities of the particular interpreter. A single incorrect reconstruction does not by itself prove ambiguity of the source text. A semantic problem is considered established when the divergence is reproducible, can be argued from the published content, or reveals objectively admissible competing readings. For reproducibility, the tested corpus, its revision, test conditions, interpreter used, and result are recorded. The test result must be addressable and have provenance. In case of divergence, at minimum the interpreter’s limitations, ambiguity of the corpus, and an unresolved problem must be distinguished.
28. Reference Semantic Profile
A reference semantic profile of the tested content must exist for a correct semantic stress test. It records what the authorial system actually asserts and what it does not assert. The reference profile must not be created retroactively merely to refute an inconvenient external interpretation. If external analysis reveals ambiguity genuinely present in the published text, the problem belongs to the quality of the representation itself. In that case the text or architecture is corrected rather than declaring the external interpreter wrong merely because its conclusion differs from the intended meaning. At the same time, assessment of divergence must take the interpreter’s own limitations into account and rely on reproducible grounds rather than a single result.
29. Classes of Semantic Divergence
In a stress test, the character of interpretation error must be distinguished. Working classes include: loss of distinction — two different contents are read as one; false distinction — one content is erroneously split into several; expansion of meaning — more is attributed to a concept than established by the canon; narrowing of meaning — part of canonical content is lost; substitution of level — for example, a semantic proposition is taken as ontological; substitution of status — a hypothesis is read as established knowledge; false reference — content is correlated with the wrong presumed referent; filling a gap — absent content is filled by assumption; loss of context — a limited assertion is read as universal. The classification remains open to refinement. A single divergence does not automatically determine whether the source of error lies in the text, architecture, or interpreter; this is established by separate analysis. For each divergence, where possible, the presumed source is recorded: corpus, representation architecture, test conditions, interpreter, or unresolved cause. Class of divergence and source of divergence must not be conflated.
30. Machine Verification and Truth
Machine validity ≠ truth status of an assertion ≠ epistemic establishment ≠ ontological actuality. Machine verification can establish conformity of representation to formal rules but does not by itself establish the truth of represented content or the ontological status of its referent or state of affairs.
30.1. Mandatory Verification of Architectural Changes
For local changes, the semantic stress test is applied according to the nature of the change. For a change of CCO architecture, two independent checks are mandatory. Semantic Stress Test: an external machine interpretation without hidden additional context is compared with the canonical meaning and checked for loss of distinctions, substitution of levels, statuses, reference, and context. Round-trip Test: CCO → export representation → CCO′. After reverse transformation, the established identities, types of semantic units, relation types, statements, statuses, contexts, dependencies, and other essential distinctions must be preserved. JSON-LD, RDF, and other export formats must not impose their own ontological architecture on CCO. A successful round-trip confirms preservation of representation, but not the truth of content.
31. External Criticism
The Ontology must permit substantive external criticism. The origin of criticism must not determine its status in advance. A critical remark is evaluated by precision in identifying the problem, correctness of reference, logic of argument, grounds, reproducibility, and ability to demonstrate a real violation of distinctions or mismatch with data. External criticism may become grounds for a change proposal but does not automatically change the canon.
32. Self-Correction
If an error is established, the system must be able to correct it without having to defend the previous formulation merely to preserve an appearance of consistency. The history of the error is preserved. The correction must indicate what was recognized as erroneous; on what grounds; what new content was accepted; which related elements were revised; and what questions remained open. The ability to correct its own error is a property of a developing knowledge system.
33. Protection Against Self-Confirmation
The canonical Ontology cannot use its own existence as sufficient proof of its own truth. The scheme “This is true because it is in the canonical Ontology; the canonical Ontology is correct because it contains true propositions” is inadmissible. Canonicity determines the normative state of representation within the system. It does not replace the grounds of assertions.
34. Independence of Verification
Where possible and substantively justified, significant propositions should permit verification by a method not reducible to repeating the same reasoning by the same means. Independence of verification may be logical, empirical, referential, model-based, semantic, machine-based, or external expert verification. Not every type of content permits every kind of verification. Absence of an inapplicable type of verification is not in itself a deficiency.
35. Verification of Competing Changes
If several variants of a change exist, they must be compared by the same criteria. Requirements must not be raised for an alternative variant while being lowered for the preferred one. For each variant, its content, grounds, premises, consequences, unresolved questions, scope of applicability, and impact on the existing architecture must be available. Preference for one variant must not be concealed in the form of an allegedly absent alternative.
36. Canonization
Canonization is a separate event. It means that after the necessary checks, a particular state is accepted as the normative current representation of the system. Canonization does not mean absolute truth, immutability, ontological fundamentality, or impossibility of further criticism. It means: “In the current state of the system, this representation is accepted as canonical on the available grounds.” Canonization is an Ontology-management event. Canonization status and other management statuses must be stored separately from truth, epistemic, and ontological statuses of content.
37. Criterion of Readiness for Canonization
A change may be considered ready for canonization if, to the degree required for its class: what is changing is clearly defined; grounds are specified; the affected level is established; semantic continuity has been checked or its indeterminacy explicitly recorded; dependencies have been analyzed; necessary distinctions have been preserved; applicable internal checks have been performed; discovered hidden substitutions of levels have been eliminated; history has been preserved; and remaining open questions have been defined. In addition, consequences for dependent elements must be assessed and, where applicable, rules for their migration or compatibility preservation must be defined. A semantic stress test is applied where external interpretability is an essential characteristic of the change.
38. Development Without Forced Completeness
The Ontology must not seek formal completeness at the cost of filling the unknown. If a question remains open, that state must be preserved. If a concept does not yet have a sufficient definition, absence of a definition must be represented as absence of a definition. If a relation has not been established, it must not be created merely for graph connectivity. Incompleteness explicitly represented as incompleteness is preferable to false completeness.
39. Development and Accumulation
Development of the Ontology is understood as a controlled sequence of changes directed toward increasing adequacy and precision of representation, discriminating capacity, coverage, verifiability, justification, referential definiteness, interpretability, ability to detect its own limitations, and ability to represent previously unrepresentable content while preserving history and controlling semantic losses. Acceptance of a change as a stage of development is not by itself proof of objective improvement. Accumulation ≠ development.
40. Architectural Openness
Protocol 08 does not declare even its own system of rules final. If development of the Ontology reveals a type of change or verification that cannot be correctly represented by the existing architecture, the architecture of development itself must become an object of analysis. However, changing the rules cannot be used as an arbitrary means of bypassing an inconvenient verification result. Revision of the Protocol requires the same principles of transparency, grounds, history, and verification that the Protocol requires of the rest of the system.
41. Fundamental Limitations
Change does not imply development. A new version does not imply new content. New content does not imply a new semantic unit. Preservation of Semantic Unit ID does not imply preservation of semantic identity. Change of Semantic Unit ID does not automatically imply change of referent. Coincidence of names does not imply identity of concepts. Difference of names does not imply difference of concepts. Merging records does not imply semantic identity. Splitting a record does not imply existence of two different referents. Discovery of a contradiction does not imply automatic falsity of one preselected side. Elimination of a formal contradiction does not imply truth of the result. Successful machine validation does not imply truth. A successful semantic stress test does not imply truth. Canonization does not imply truth. External criticism does not imply an error in the system. Internal agreement of the system does not imply its truth. Absence of an answer does not imply a right to complete it. Completeness of a machine record does not imply completeness of knowledge. Change in representation architecture does not imply change in Reality itself. Management status does not imply truth. An unsuccessful semantic stress test does not automatically imply a defect in the source meaning. Successful cross-verification does not imply truth. Absence of a registered dependency does not imply absence of dependency.
42. Canonical Development Cycle
Development of the Ontology must be representable as a sequence of distinguishable stages: Change proposal ↓ Identification of affected level ↓ Classification of change type ↓ Analysis of semantic continuity ↓ Analysis of dependencies and possible cascade ↓ Representation of grounds ↓ Internal verification ↓ Cross-protocol verification ↓ Semantic stress test — when applicable ↓ Canonization decision ↓ New canonical state ↓ Preservation of the complete transition history. This sequence is an architecture for controlling change, not an assertion that every real research process always proceeds linearly. Inquiry may return to previous stages. A decision to canonize or reject canonization must be recorded as a separate management event together with the grounds for the decision.
42.1. Extended Cycle of Architectural Change
If a change affects the CCO architecture, the canonical cycle of §42 is supplemented by the mandatory sequence: Cross-protocol verification → verification of Ontology Architecture Version → Semantic Stress Test → Round-trip Test → migration and backward-compatibility verification → canonization decision. Migration verification must establish whether existing identities, records, relations, statements, statuses, and history can be transferred into the new architecture without hidden semantic substitution. If full backward compatibility is impossible, the boundary of incompatibility and transition rules must be explicitly recorded before canonization.
43. Place of Error
Error is not treated as an element that must be hidden to preserve the authority of the system. If an error is recognized and its history preserved, it becomes a source of information about the boundary of previous understanding. A mature system must therefore distinguish the presence of an error from inability to correct an error. The former is possible in any developing knowledge system. The latter creates a mechanism for reproducing error within the architecture of knowledge itself.
44. Canonical Principle
A developing Ontology must preserve not the immutability of its formulations but continuity in distinguishing what exactly it asserts, on what grounds, in what status, and why this content changes. Change is admissible not because the new is automatically better than the old, but because the system can explicitly show the grounds for transition and preserve the history of that transition. Semantic continuity is not created by an identifier. Canonicity does not create truth. Internal consistency does not replace verification. Machine validity does not replace knowledge. External interpretation does not replace canonical meaning but may reveal insufficient definiteness of its expression. The unknown must not be filled for the sake of system completeness. Error must not be hidden to preserve its authority. The architecture must be capable of change, but change of the architecture itself must also remain subject to grounds and verification. The maturity of the Ontology is therefore determined not by its immobility or the amount of accumulated content, but by its ability to develop without loss of distinctions, correct itself without rewriting its own history, and expand the space of the representable without turning assumption into established knowledge. Internal consistency does not replace external, referential, or empirical verification where such verification is applicable.