Deployment
Context an agent needs to record, inspect and reason about a deployment: the identified act and standing fact of a release installed into a runtime environment target, with declared intent, execution state, history and evidence references.
Bundle → Layer → Finding → Questions Filled
6 bundles · 12 layers · 24 findings · 92 questions
Deployment identity and subject binding What counts as one deployment, how it is identified and classified, and how it binds a referenced release to a referenced environment target and to the resulting running instances.
Deployment definition, identity and classification
The definitional core: the unit of deployment, its identifier rules and the kind of deployment being recorded.
Unit of deployment and record granularity
Defines what is counted as one deployment (one release, one target scope, one intent) and distinguishes a deployment occurrence from the deployable object, the pipeline run and the running service instance.
- What exactly constitutes one deployment record, and what is the smallest change that starts a new one rather than updating an existing one? definition
- When a single release is rolled out to many targets or many components at once, is that one deployment or many, and how is the grouping represented? composition
- How is a deployment record distinguished from the pipeline run that produced it and from the service instance that results from it? relationship
- Does a configuration-only or parameter-only change with no release change produce a deployment record? decision
Deployment identifier and correlation keys
How a deployment is identified across systems, including the priority of authoritative identifiers over minted ones and the natural keys used to reconcile duplicate reports of the same deployment.
- Which system is the authoritative master system for this deployment identifier, and what identifier does it assign? identity
- When no authoritative identifier exists, what governed identifier or minted surrogate is used and who minted it? authority
- What natural key set is used to detect that two reports describe the same deployment occurrence? validation
- Is the deployment identifier stable across retries, resumed rollouts and orchestrator restarts? constraint
Deployment kind and intent classification
The controlled vocabulary distinguishing initial install, upgrade, rollback, configuration-only change, redeploy and removal, aligned to the CDEvents service predicates.
- Which deployment kind applies, and which controlled vocabulary term is used for it? classification
- Is this a planned deployment, an emergency or hotfix deployment, or unplanned rework triggered by an incident? classification
- Is the deployment declared reversible, and what is the declared reversal path? constraint
Referenced subject bindings
The pinned references to the release and to the environment target, and the correlation to the instances that result from the deployment.
Release reference and binding integrity
How the deployed release is referenced and pinned, carrying the reference, the artifact coordinate and the digest used to prove the binding, without reproducing release lifecycle.
- Which release is bound to this deployment, and by what pinned reference and artifact coordinate? relationship
- Which content digest and algorithm were recorded at binding time to prove that the intended release was the one deployed? evidence
- What happens to this deployment record if the referenced release is later revised, withdrawn or re-tagged upstream? exception
- Can one deployment bind more than one release or component, and how is the set bounded? composition
Environment reference and target placement
The pinned environment reference plus the deployment-specific target selector describing which cells, regions, tenants or device groups received the release.
- Which runtime environment is targeted, and by what pinned reference and environment tier name? relationship
- Within that environment, which subset of targets was selected for this deployment? composition
- In which geographic or jurisdictional locations did the deployed instances actually run as a result of this deployment? spatial
- Which tenants, customers or device cohorts were in scope, and were any explicitly excluded? constraint
Deployed instance and installed-software correlation
How the deployment record correlates to the concrete running or installed instances it produced, so that observed installed software can be reconciled to a deployment.
- Which service instance identifiers, container identities or device identifiers are attributable to this deployment? identity
- What installed-software evidence on the target corroborates that this release is present? evidence
- How is an observed running instance attributed to exactly one deployment when several deployments overlap in time? validation
Declared intent and deployment configuration The declarative desired state for the deployment, the deploy-time parameters bound to it, and the planned rollout strategy and window.
Declared desired state and parameters
What was declared to be true after the deployment, and which subject-specific parameter values were bound at deploy time.
Desired state declaration and its version
The declarative statement of the intended post-deployment state, its immutable version reference and its source of truth, kept distinct from the reconciliation that applies it.
- Where is the declarative desired state held, and which immutable version of it does this deployment assert? provenance
- What does the declaration assert about the target after deployment, in format-neutral terms? definition
- How is divergence between the declared desired state and the reported actual state recorded on the deployment? state
- Who is authorised to change the desired state declaration for this target, and is the change recorded as a new deployment? authority
Deploy-time parameter and secret-reference binding
The subject-specific configuration values, feature toggles and opaque secret references bound at deployment time, together with the secure-default baseline they deviate from.
- Which parameter keys and values were bound for this deployment, and which deviate from the documented secure default configuration? constraint
- How are credentials and key material referenced without being carried in the deployment record? security
- Which bound parameters are sensitive or personal-data-relevant and therefore require redaction on read? privacy
- Which parameter values were actually effective at completion, as opposed to merely requested? validation
Rollout strategy and deployment plan
The declared rollout method and the planned timing, sequencing and constraints for carrying it out.
Declared rollout strategy
The strategy declared for this deployment, such as recreate, rolling update, blue-green or canary with staged weights and pauses, recorded as intent rather than as executable mechanics.
- Which rollout strategy was declared, and which strategy-specific limits such as surge, unavailability or step weights were set? classification
- Which pauses in the declared strategy require an explicit promotion decision rather than elapsing automatically? decision
- Under what declared conditions is the rollout to be aborted, and what does abort mean for the recorded state? exception
- Does the declared strategy accept service interruption, and was that interruption expected and communicated? requirement
Deployment plan, sequencing and permitted window
The planned schedule, ordering dependencies among deployments, and the freeze or maintenance windows constraining when the deployment may proceed.
- Within which permitted window was this deployment scheduled, and in which local offset is that window expressed? temporal
- Which other deployments must complete before or after this one, and what happens if a predecessor fails? process
- What plan-level mapping of components to target nodes was declared before execution began? composition
- Did execution deviate from the plan, and what deviation was recorded? exception
Execution state, history and time The recorded state of the deployment over its life, the revision history that determines which deployment is current for a target, rollback linkage, and the time semantics that make all of it comparable.
Deployment state model
The state vocabulary, permitted transitions and the criteria by which a deployment is judged complete or failed.
Deployment state vocabulary and transitions
The controlled set of deployment states, which are terminal, and which transitions are permitted, recorded as an append-only transition history.
- Which states can a deployment occupy, and which of them are terminal? state
- Which state transitions are permitted, and which actor or source may assert each of them? lifecycle
- How is a paused or awaiting-promotion deployment represented, and does pause time count toward deployment duration? state
- When the state is reported by an external orchestrator, how is the reported value mapped into this model's vocabulary? interoperability
Progress reporting and completion criteria
The declared criteria by which a deployment is judged to have progressed, completed or failed, and how externally observed progress reports are recorded against them.
- What criteria must be satisfied for this deployment to be declared complete, and who declared them? requirement
- What deadline applies before a non-progressing deployment is treated as failed, and what happened when it elapsed? constraint
- Which observed counts or signals were reported against the declared criteria, and by which reporting system? measurement
- When a deployment failed, what reason code was recorded and was that reason produced locally or received from the orchestrator? provenance
History, supersession and rollback linkage
How deployment records accumulate into a history per target, which one is current, and how rollbacks and recoveries are linked.
Revision history and current-deployment resolution
The ordered history of deployments for a target, the rule that resolves which deployment is currently in effect, and the bounded retention of superseded revisions.
- Which deployment is currently in effect for a given environment target, and by what deterministic rule is that resolved? state
- How are deployments to the same target ordered when their reported timestamps are equal, missing or out of order? temporal
- How many superseded revisions are retained in usable form, and what is retained once a revision falls outside that bound? retention
- How is a gap in the history detected when deployments occurred without being recorded? quality
Rollback, roll-forward and recovery linkage
The typed links between a remediating deployment and the deployment it reverts or repairs, together with the recovery outcome as reported.
- Which prior deployment does this rollback revert, and which release state is restored as a result? relationship
- What triggered the remediation, and was it an automated abort or a human decision? event
- How long did recovery take from the failed deployment to a restored serving state, and against which two recorded instants is that measured? measurement
- Was the rollback complete, partial or unsuccessful, and what residual state was left behind? exception
Deployment time semantics
The distinct instants that a deployment record carries and the rules that keep them comparable across systems and jurisdictions.
Event time versus observation and ingestion time
Separates when the deployment was requested, started, completed or superseded from when those facts were observed and ingested, and fixes the representation rules for all of them.
- Which distinct instants are recorded for this deployment, and which of them are mandatory? temporal
- When was each reported fact observed by the reporting system and ingested into this model, and how is that kept separate from the event instant? provenance
- In which offset was each instant originally expressed, and is that original offset preserved alongside any derived normalised value? constraint
- Which upstream instant anchors change lead time for this deployment, and is that anchor owned by this model or referenced? measurement
Authorisation and assurance evidence Who authorised the deployment, which gate verdicts were relied upon, and which integrity and post-deployment verification evidence is referenced.
Authorisation to deploy
The recorded authority for the deployment and the references to gate decisions relied upon at the deployment boundary.
Deployment authorisation and accountable actors
The recorded authorisation for this deployment, the change or approval reference relied upon, the actors involved, and the separation-of-duties assertion, all carried as references to externally governed decisions.
- Which change or approval record authorised this deployment, and what was its recorded verdict at the time of binding? authority
- Which human or automated actors requested, approved and initiated the deployment, and in which roles? ownership
- Was separation of duties satisfied between the actor who approved and the actor who initiated, and how is a violation recorded? constraint
- If normal authorisation was bypassed under an emergency procedure, what retrospective authorisation is required and by when? exception
Deployment gate decision references
References to verdicts issued by external verifiers or policy engines that were relied upon at the deployment boundary, recorded without any local evaluation, re-evaluation or enforcement.
- Which gate verdicts were relied upon for this deployment, and which verifier issued each one? evidence
- Against which policy identifier and version was each relied-upon verdict issued, and where does that policy live? provenance
- How recent must a gate verdict be to remain valid for this deployment, and what happens when it is stale? constraint
- Was any failing or missing gate verdict overridden, and what override record justifies proceeding? exception
Integrity and verification evidence
Evidence referenced before deployment to prove the intended release was deployed, and the outcome recorded after deployment.
Pre-deployment integrity evidence
The evidence relied upon before installation that the artifact bound to the deployment is authentic and unmodified, referenced from the archived release and signature or attestation sources.
- What integrity check was performed before installation, and was its result recorded as passed, failed or not performed? validation
- Which signature, attestation or archived-release reference supports the authenticity of the deployed artifact? evidence
- What is recorded when the digest presented at install time differs from the digest bound at planning time? exception
Post-deployment verification outcome and acceptance
The recorded outcome of post-deployment checks and the acceptance decision closing the deployment, captured as received verdicts rather than as executed tests or live health monitoring.
- What post-deployment verification verdict was received, from which checking system, and at what instant? evidence
- Who accepted the deployment as successful, and on what basis was that acceptance recorded? decision
- Which defects, degradations or incidents were attributed to this deployment after completion, and how are they referenced? quality
- Is there a declared observation period after completion before the deployment is considered stable, and when does it end? temporal
Delivery measurement and record governance Metrics derived from deployment records, and the provenance, access, sensitivity and retention rules governing those records.
Delivery performance measurement
Metrics computed over this model's own deployment records, with their windows, denominators and validity conditions.
Delivery metrics and measurement validity
The metric definitions computed from deployment records, the population and window they apply to, and the record-completeness conditions under which the results are valid.
- Which delivery metrics are computed from these records, and by which exact definition and denominator? measurement
- Over which time window and target population is each metric computed, and in which offset is the window expressed? temporal
- Which record-completeness conditions must hold before a computed metric may be published, and what is reported when they fail? quality
- Which deployments are excluded from metric populations, and on what documented basis? constraint
Record provenance, access and retention
How deployment records came to exist, who may read them and at what sensitivity, and how they are disposed of.
Deployment record provenance and reconciliation
The origin, capture path and trust level of each deployment record, including whether it was asserted by a pipeline, reconciled from observed state, or entered manually.
- Which system asserted this deployment record, and through which capture path did it reach the model? provenance
- What confidence level is attached to the record, and which fields are inferred rather than asserted? quality
- When observed installed state contradicts the recorded deployment, which source wins and what correction entry is written? validation
- Which parts of a terminal deployment record are immutable, and how are later corrections expressed? constraint
Record access, sensitivity and redaction
The sensitivity classification of deployment record content and the rules governing who may read which parts, including production target detail and parameter bindings.
- What sensitivity classification applies to each part of the deployment record, and who set it? privacy
- Which roles may read production target placement, parameter bindings and evidence references, and at what scope? access
- Under what emergency conditions may normal read restrictions be bypassed, and what must be recorded when they are? exception
Retention, disposition and tombstones
How long deployment records are kept, what reduced form survives disposition, and which external policy owns the execution of deletion.
- What retention period applies to a deployment record, and from which instant is it measured? retention
- What minimal tombstone survives disposition, and which fields are purged? retention
- What suspends disposition, such as legal hold or an open investigation, and who may set and release it? authority
- When a referenced release or environment record is deleted upstream, what happens to this deployment record? exception
Interoperability and external alignment How deployment records are projected onto, and reconciled with, external deployment vocabularies without asserting conformance.
Standards alignment and identifier exchange
The recorded mappings to external standards, the identifier schemes exchanged, and the known points where mappings are lossy.
External standard alignments and identifier exchange
The declared alignments to external deployment vocabularies, the identifier and digest schemes used to exchange deployment facts, and the explicit statement that alignment is not conformance.
- Which external vocabulary terms does each local element map to, and at which version of that vocabulary? interoperability
- Where is a mapping lossy, approximate or unavailable, and what is recorded in place of a value? quality
- Which identifier and digest schemes are exchanged with external systems, and how are they normalised on ingest? identity
- Is conformance to any external specification asserted, and if so on what cited evidence? evidence
Classifiers Filled
- Family
- World Models
- Category
- Information and virtual systems
- Entry kind
- aggregate
- Navigation path
- NAV.INF.SFT.DEP
- Domain
- INF.SFT.DEP
- Industry
- Cross-industry
- Tags
- deploymentinf.sft.dep
What it is Filled
A Deployment reifies the binding between a release (WM-SFT-008) and a runtime environment target (WM-SFT-010). It is not a plain edge: the same release may be deployed to the same target repeatedly, and each occurrence carries its own identity, desired state, state history and outcome. The model owns the deployment record, its identity, classification, binding keys, subject-specific deploy-time parameters, rollout intent, the deployment state machine, revision history, current-deployment resolution, rollback linkage, deployment time semantics, and references to authorization, gate and verification evidence produced elsewhere. It describes and records deployments; it does not orchestrate, execute, evaluate or enforce them, and it is not the system of record for release content, environment definition, policy decisions, telemetry or organizational audit trails.
In scope
- Deployment identity, classification and the release-to-target binding keys
- Declared desired state, deploy-time parameter binding and rollout strategy intent
- Deployment state machine, completion criteria, revision history, supersession and current-deployment resolution per target
- Rollback and recovery linkage between deployment records
- Deployment time semantics separating requested, started and completed times from observation and ingestion times
- References to authorization, gate decisions and verification evidence produced by other systems
- Delivery measurement over deployment records, plus record provenance, access, retention and interoperability projections
Out of scope
- Release identity, versioning, artifact content, SBOM and build provenance (owned by WM-SFT-008)
- Runtime environment definition, topology, capacity, infrastructure inventory and environment lifecycle (owned by WM-SFT-010)
- Execution of rollouts: orchestrators, reconciliation agents and traffic shifting mechanics
- Policy evaluation, gate decision logic, enforcement engines and their runtime decisions
- Runtime observability, alerting, incident and problem management, and service level objectives
- Organizational audit-trail systems and evidentiary record keeping beyond this model's own records
- CI/CD pipeline definition and pipeline run orchestration semantics
- ITSM change record lifecycle, approval workflow engines and human identity management
- Secret values, credentials and key material (only opaque references are carried)
Why it exists Filled
Context an agent needs to record, inspect and reason about a deployment: the identified act and standing fact of a release installed into a runtime environment target, with declared intent, execution state, history and evidence references.
Distinguishing features Filled
- Records each occurrence of putting a release onto a target as its own object with identity, desired state and outcome.
- Differs from a release, which can be deployed many times, and from the runtime environment it targets.
- Holds state history and rollback links, unlike an orchestrator that executes the change.
- Supports delivery metrics such as deployment frequency and change failure rate.
What robots and AI may and may not do Filled
Must not
- Start a production deployment without the required change approval.
- Bypass failed policy gates or verification checks.
- Record a deployment as successful before the observed state confirms it.
- Deploy an unsigned or unverified release.
- Delete deployment history needed for audit or incident analysis.
Only with a human decision
- Approving deployment to production or safety-critical environments.
- Overriding a failed gate in an emergency.
- Deciding on rollback during an incident with customer impact.
May
- Register a deployment record and its state transitions.
- Ingest observed state reports from the target environment.
- Link a rollback or recovery deployment to the one it replaces.
- Compute delivery metrics from recorded deployments.
Moral aspects Filled
- Failed deployments can interrupt services people rely on, including health and safety services.
- Delivery metrics should judge the process, not single out individual engineers.
- Emergency overrides need clear accountability.
Who is affected
- Users of the deployed service
- Operators and on-call engineers
- Owners of dependent systems
Owners Filled
Steward
The adopting Dimension must name a single system of record for deployment identifiers per environment class and register it as the authoritative master system.
Roles
- Model steward
- Own the scope statement, boundary notes and state vocabulary; Approve breaking changes and version transitions; Adjudicate ambiguous current-deployment resolutions
- Deployment record custodian
- Register and maintain deployment records and their history ledgers; Apply deduplication, quarantine unverified artifacts and write correction entries; Maintain confidence levels and gap markers
- Boundary liaison for referenced models
- Maintain the pinned reference contracts to WM-SFT-008 and WM-SFT-010; Detect and escalate restatement of target-owned content; Handle upstream withdrawal and dangling-reference notifications
- Interoperability maintainer
- Maintain the alignment mapping set and external version and stability tracking; Publish loss reports for lossy projections; Prevent unstable external terms being treated as normative
- Retention and disposition officer
- Set retention periods and hold status; Authorise tombstone reduction and emit disposition markers; Confirm that execution of deletion is performed by the storage policy owner
Links to other meta-models Filled
references
- WM-SFT-008 Release - A deployment installs a release. This model carries the pinned release reference, artifact coordinate and binding digest only; release identity, versioning, artifact content, SBOM, build provenance and release lifecycle remain owned by the target model.
- WM-SFT-010 Runtime environment - A deployment targets a runtime environment. This model carries the pinned environment reference, the deployment-specific target selector and reported placement; environment definition, topology, capacity and environment lifecycle remain owned by the target model.
- Pipeline or automation run model (not yet registered) - Carry an opaque reference to the run that requested the deployment so that change lead time can be anchored upstream; pipeline definition, orchestration and run lifecycle are not modelled here. No sibling model is registered, so this link is a declared structural gap.
- Change authorisation or configuration change control record (adopting-Dimension owned) - Carry the authorisation reference and its received verdict at the deployment boundary; change workflow, impact analysis and approval lifecycle remain with the configuration change control regime.
- Verifier-issued verification summary attestation - Carry gate verdict references with issuer, policy identifier and verification instant; evaluation, policy authorship and enforcement are owned by the verifier and are explicitly not acquired by referencing them.
child
- WM-SFT-002 software product or service - This model is a child of the parent software product or service model, which owns product identity, ownership and commercial context; this model adds only the installed-into-environment occurrence.
aligned
- CDEvents continuous deployment event vocabulary - Align deployment kinds and the service and environment subject bindings to the CDEvents predicates and required subject fields, recorded as an alignment with mapping evidence, not as a conformance claim.
- OpenTelemetry deployment and service resource attributes - Align environment tier naming and deployed-instance correlation to the stable deployment environment attribute and service instance identity, while flagging that deployment identifier, name and status attributes are not yet stable.
- SPDX 3.0.1 Core LifecycleScopeType - Align this model's runtime-phase assertions with the runtime lifecycle scope used for SPDX relationships, noting that SPDX defines a relationship scope rather than a deployment entity.
- ISO/IEC 19770-2 software identification tags via NISTIR 8060 - Align deployed-instance correlation with endpoint installed-software evidence so that observed installations can be reconciled to deployment records; tag production and asset inventory remain external.
extends
- Fleet, device and over-the-air update deployment specialisation - Permit a specialisation that adds device-cohort staging, update slot and offline-target semantics for embedded and device fleets, specialising only the deployment subject; generic identity, authority, lifecycle and conflict handling stay in this model.
neighbor
- WM-SFT-008 Release - The release owns what is being installed: version, artifact digests, SBOM, build provenance and release approval. This model carries only a pinned release reference plus the digest used for binding integrity, and never reproduces release lifecycle or release-level functions.
- WM-SFT-010 Runtime environment - The environment owns where software runs: environment identity, tier, topology, capacity and its own lifecycle. This model carries a pinned environment reference plus the deployment-specific target selector (cell, region, tenant slice) and never models environment creation, modification or deletion.
- WM-SFT-002 parent software product or service model - Product and service identity, ownership and commercial context are inherited from the parent. This model adds only the installed-into-environment occurrence.
- Deployment orchestrator or reconciliation agent - Kubernetes Deployment objects, GitOps reconcilers and progressive-delivery controllers execute and observe rollouts. This model records declared intent and reported outcome as data; it grants no ownership of execution, reconciliation or traffic-shifting semantics.
- Policy verifier and gate evaluator - SLSA verification summary attestations and equivalent gate verdicts are produced by external verifiers against external policies. This model records only the verdict reference, issuing verifier and verification time; it does not define evaluation, re-evaluate, or enforce.
- Change management and configuration control process - Change authorization, impact analysis and access restrictions for change are governed by the adopting Dimension's configuration change control regime. This model carries the authorization reference and its verdict, not the change workflow.
- Software asset inventory and installed-software tagging - Endpoint software identification tagging asserts what is installed on an asset. This model may reference such install evidence to corroborate a deployment, but tag production and asset inventory remain external.
parent
- WM-SFT-002
What else AI and robots need to interact with it Filled
Identity and identifiers required Filled
- Authoritative master-system identifier: the deployment identifier assigned by the registered system of record for deployments in the adopting Dimension, stored together with a reference to that issuing system.
- Governed global identifier or IRI: a resolvable identifier minted under a governed namespace when no master-system identifier exists.
- UUID or ULID assigned by the adopting Dimension, recorded as locally minted with its minting authority, used only when neither higher-priority identifier is available.
- A date, a version string, a release tag, an environment name, a commit hash or an artifact digest is never used as the identifier of a deployment; digests are binding evidence, not identity.
Direct properties not applicable Not applicable
Not applicable
Institutional or informational subject: no invented physical properties.
Recognition optional Filled
- A deployment links one release to one target with a desired state, actual state and timestamps.
- Often confused with a release, a pipeline run and the running instance itself.
Capabilities and actions required Filled
- Register deployment record: Create a deployment record binding a pinned release reference to an environment target with a declared intent, without initiating any execution.
- Record deployment state transition: Append a state transition asserted by an external orchestrator or actor, mapping the reported status into the local vocabulary and retaining the raw value.
- Resolve current deployment for a target: Determine deterministically which deployment record is currently in effect for a given environment target as of a stated instant.
- Ingest observed state report: Ingest an externally produced report of observed installed or running state and record its agreement or divergence with the declared desired state.
- Link rollback or recovery deployment: Establish the typed link between a remediating deployment and the deployment it reverts or repairs, and record the recovery interval.
- Attach verification evidence reference: Attach a reference to an externally issued integrity, gate or post-deployment verdict, recording issuer, policy identifier and verification instant without evaluating or recomputing it.
- Compute delivery metrics: Compute delivery metrics over this model's deployment records for a stated window and population, qualified by record coverage.
- Project deployment to an external vocabulary: Produce a projection of a deployment record into an external deployment vocabulary using the maintained mapping set, marking lossy and unmapped elements.
- Apply retention disposition marker: Evaluate a deployment record against the adopting Dimension's retention policy and, where eligible, reduce it to a tombstone and emit a disposition marker for the storage owner to execute.
Hazards and failure modes required Filled
- Outages from a faulty deployment.
- Configuration drift between recorded and actual state.
- Unauthorized changes reaching production.
Standards and interfaces required Filled
- NIST SP 800-218 Secure Software Development Framework.
- NIST SP 800-53 Rev. 5 configuration management controls.
- ISO/IEC/IEEE 12207 software life cycle processes.
- CDEvents specification.
- OpenTelemetry for observed state signals.
Context of use required Filled
- Data residency and sovereignty constraints on where a release may lawfully run vary by jurisdiction; the model records resulting placement regions as facts but does not encode jurisdictional rules.
- Regulated sectors such as life sciences, financial services and public safety may mandate additional pre-deployment approval and validation artifacts that this model can only reference.
- Retention periods, legal hold authority and permissible purge scope for deployment records are set by the adopting Dimension and differ by jurisdiction.
- Maintenance and freeze windows are locally defined, so explicit offsets are mandatory; assuming UTC would silently misplace windows across daylight-saving boundaries.
- Personal-data exposure through actor identifiers in deployment records is treated as jurisdiction-dependent and handled by sensitivity classification rather than a fixed rule.
Sources Filled
- CDEvents Continuous Deployment Events - Continuous Delivery Foundation (CDF)
- OpenTelemetry Semantic Conventions - Deployment attribute registry - OpenTelemetry (CNCF)
- Kubernetes Documentation - Deployments - The Kubernetes Authors (CNCF)
- OpenGitOps Principles v1.0.0 - OpenGitOps (CNCF App Delivery TAG)
- SLSA Verification Summary Attestation (VSA) - OpenSSF / SLSA project (Linux Foundation)
- NIST SP 800-218, Secure Software Development Framework (SSDF) Version 1.1 - National Institute of Standards and Technology (NIST)
- NIST SP 800-53 Rev. 5, Security and Privacy Controls for Information Systems and Organizations - National Institute of Standards and Technology (NIST)
- NISTIR 8060, Guidelines for the Creation of Interoperable Software Identification (SWID) Tags - National Institute of Standards and Technology (NIST)
- Deployment and Configuration of Component-based Distributed Applications (DEPL), Version 4.0 - Object Management Group (OMG)
- ISO/IEC/IEEE 12207:2026, Systems and software engineering - Software life cycle processes - International Organization for Standardization (ISO)
- SPDX Specification 3.0.1 - Core LifecycleScopeType vocabulary - SPDX project (Linux Foundation)
- DORA software delivery metrics: the four keys - DORA (Google Cloud)
- Argo Rollouts - Canary Deployment Strategy - Argo Project (CNCF)
Open questions
- Obtain ISO/IEC 20000-1:2018 clause 8.5 and the ITIL 4 deployment management practice in full text and re-audit the plan, record and reporting attributes; both were skipped as paywalled and may impose mandatory attributes this model omits.
- Obtain clause-level access to ISO/IEC/IEEE 12207:2026 and ISO/IEC 19770-2, then re-ground SRC-010 and SRC-008 dependent findings; current support is a catalogue abstract and NISTIR 8060 respectively, so clause alignment is asserted nowhere.
- Resolve the DORA metric-set question against the current published guidance: the cited page is the four-keys guide while the conflicts list asserts five metrics, so fnd-delivery-performance-metrics must name the edition it pins and state the change-fail-rate definition it uses.
- Model or explicitly delegate database schema migration and data backfill coupled to a deployment, which is the one self-declared gap in the checklist and currently reachable only through the reversibility flag; search for a primary source that structures migration-coupled irreversibility.
- Decide whether sibling models are required for pipeline or automation runs and for ITSM change records, since q-unit-vs-neighbours and fnd-deployment-authorization both depend on referents that have no registered model and no relation row.
- Specify the currency termination rule for a deployment whose referenced environment is withdrawn upstream, and the treatment of append-only history ledger entries that reference a record already reduced to a tombstone.
- Deepen firmware and over-the-air fleet update coverage (staged cohorts, A/B update slots, offline and intermittently connected targets), which is currently accommodated only through an EXTEND link that does not exist in the frozen relation ledger.
- Author the five missing functions identified in this audit (deduplicate or merge, supersede terminal record, redacted read, quarantine unverified artifact, set and release legal hold) in a follow-up pass, since add_functions is contractually empty under the single-provider waiver.
- ISO/IEC 20000-1:2018 clause 8.5 release and deployment management and the ITIL 4 deployment management practice were not consulted in full text because both are paywalled; they may impose additional plan, record or reporting attributes.
- ISO/IEC/IEEE 12207:2026 and ISO/IEC 19770-2 full texts are paywalled; the catalogue abstract and NISTIR 8060 respectively were used instead, so clause-level alignment is not asserted.
- No sibling model is registered for pipeline or automation runs, or for ITSM change records; those composition links are declared not-required structural gaps rather than canonical structure.
- Firmware and over-the-air fleet updates (staged cohorts, A/B update slots, offline targets) are only accommodated through a permitted EXTEND link and are not modelled in depth.
- Database schema migration and data backfill coupled to a deployment, and their effect on reversibility, are acknowledged but not structurally modelled.
- Per-tenant entitlement, licensing and commercial gating of what may be deployed to whom are excluded.
- Cost, capacity and resource-consumption consequences of a deployment are excluded as environment-owned.
Machine files
Provenance
world-models research · reviewable-draft
Built from: models/wm-sft-009-deployment/spec.yaml, ver-cy/world-models/card-supplements/wm-sft-009-deployment.json