Implemented feature inventory
The CLI exposes this same inventory through boxr features and boxr features --json. This document is generated from the feature registry and checked by the CLI test suite.
Inventory size: 900 capabilities. Each entry maps to implemented behaviour, an explicit sub-control, or a testable tooling target.
Implemented capabilities
- help and usage
- version metadata
- structured version fields
- JSON version output
- read-only host doctor
- JSON doctor output
- strict doctor warning mode
- run command with arguments
- unknown command diagnostics
- repeatable environment overrides
- environment key validation
- NUL-safe environment validation
- duplicate environment key collapse
- environment files with comments and export syntax
- single and double quoted env values
- hostname isolation
- working directory
- memory hard limit
- memory high threshold
- memory low protection
- memory min protection
- memory swap cap
- memory OOM group
- CPU quota
- CPU weight
- block I/O weight
- PID count limit
- cgroup v2 controller delegation checks
- cgroup rollback on failed setup
- cgroup cleanup after exit
- namespace startup barrier
- PID namespace init supervisor
- child process reaping
- signal forwarding
- workload exit-code propagation
- private mount propagation
- optional root filesystem
- chroot execution
- read-only root filesystem
- proc filesystem mount
- recursive bind mounts
- read-only bind mounts
- tmpfs mounts
- tmpfs size cap
- tmpfs mode
- private network namespace
- loopback interface activation
- host networking mode
- user namespace UID mapping
- user namespace GID mapping
- capability name parser
- drop selected Linux capabilities
- drop all Linux capabilities
- no-new-privileges
- rootfs archive path traversal rejection
- rootfs archive size limits
- rootfs archive entry limits
- rootfs safe symlink extraction
- rootfs absolute symlink normalization
- rootfs hard-link extraction
- rootfs special-file rejection
- rootfs atomic extraction
- rootfs pack
- rootfs inspect
- rootfs verify
- rootfs deterministic SHA-256 digest
- rootfs digest verification
- rootfs directory diff with added/modified/removed paths
- rootfs security audit without following symlinks
- rootfs audit detects escaping relative symlinks
- rootfs audit detects setuid files
- rootfs audit detects setgid files
- rootfs audit detects world-writable paths
- rootfs audit treats sticky world-writable directories as informational
- rootfs audit detects special files
- rootfs audit JSON report
- rootfs audit strict failure mode
- image audit by local tag
- image audit reuses safe rootfs extraction
- image audit JSON report
- image audit strict failure mode
- rootfs audit configurable entry limit
- rootfs audit rejects non-positive entry limits
- rootfs audit fails on limit rather than returning partial results
- image audit configurable entry limit
- versioned JSON run config
- strict JSON config field validation
- config size limit
- config example generator
- config name inheritance
- Go source formatting target
- race-enabled test target
- go vet target
- cross-platform CLI builds
- Linux amd64 build
- Linux arm64 build
- Windows CLI build
- macOS Intel CLI build
- macOS Apple Silicon CLI build
- named containers
- collision-resistant 12-character container IDs
- duplicate container-name protection
- persistent JSON container records
- private 0600 container state files
- environment-secret redaction from state
- container labels
- created lifecycle state
- running lifecycle state
- exit status persistence
- container create/start/exit/remove event journal
- bounded event-log reader
- recent event tail limit
- active container listing
- completed container history listing
- container filter by ID prefix
- container filter by name
- container filter by status
- container filter by label
- container inspect JSON
- completed-container removal
- running-container removal protection
- remove-all-completed workflow
- automatic record removal after exit
- age-based state pruning
- live cgroup memory stats
- live cgroup CPU stats
- live cgroup PID stats
- live cgroup I/O stats
- container stats JSON output
- host and runtime info report
- host info JSON output
- non-Linux stats limitation reporting
- combined stdout and stderr log capture
- private 0600 container log files
- bounded container log reads
- log tail by line count
- live log following
- container stop with graceful timeout
- container kill with named signals
- container wait for completion
- signal-aware exit code persistence
- log cleanup on container removal
- named persistent volumes
- volume create with metadata
- volume labels
- volume inspect JSON
- volume list table and JSON
- volume duplicate-name protection
- volume path traversal rejection
- volume directory permission hardening
- volume metadata permission hardening
- volume symlink path rejection
- volume remove with data cleanup
- volume prune with explicit force
- active volume usage detection
- run with named volume bind mounts
- read-only named volume mount
- named volume persistence across runs
- volume list name and label filtering
- volume prune label filtering
- volume export as safe gzip tar archive
- OverlayFS writable upper layer
- persistent overlay upper and work directories
- overlay path traversal and nesting validation
- overlay read-only remount compatibility
- overlay mount path option validation
- separate writable layer per caller-selected directory
- OverlayFS upper/work concurrency lock
- canonical OverlayFS storage lock path
- symlink-resistant OverlayFS lock file
- private 0600 OverlayFS lock permissions
- OverlayFS lock alias contention detection
- empty OverlayFS workdir and mountpoint validation
- local OCI image layout store
- OCI image index schema validation
- OCI image manifest descriptors
- OCI image config descriptors
- content-addressed SHA-256 blob storage
- custom documented Boxr rootfs layer media type
- rootfs import as image reference
- strict image reference validation
- image tag replacement without duplicate index entries
- image listing in table format
- image listing in JSON format
- quiet image reference listing
- image inspection with platform and layer metadata
- image manifest blob integrity verification
- image config blob integrity verification
- image layer blob integrity verification
- safe image unpack to new rootfs destination
- image tag removal without deleting shared blobs
- portable OCI layout archive save
- portable OCI layout archive load
- archive descriptor validation before import
- image blob garbage collection dry run
- image blob garbage collection with explicit force
- private symlink-resistant image store
- atomic OCI index updates
- local image retagging without copying blobs
- image alias removal preserves shared content
- built-in Linux seccomp baseline profile
- seccomp architecture guard in BPF filter
- seccomp denylist for namespace and kernel administration syscalls
- seccomp denylist for kernel module and BPF operations
- seccomp denylist for process tracing and cross-process memory access
- seccomp denylist for perf and userfaultfd
- seccomp automatically enables no-new-privileges
- seccomp profile name validation
- seccomp filter generation tests
- seccomp filter subprocess integration test
- OCI uncompressed layer diff_id verification
- image prune rejects unexpected blob filenames
- image store root symlink rejection
- streaming SHA-256 verification for image layers
- bounded OCI blob sizes
- streaming portable image save and load
- full image graph validation before garbage collection
- redacted container execution dry-run plan
- dry-run validates complete runtime options
- dry-run reports environment variable names only
- dry-run redacts environment and label values
- dry-run avoids creating container lifecycle records
- explicit local image integrity verification command
- image store unique blob count and byte accounting
- image store reachable content byte accounting
- image store reclaimable content byte accounting
- full image store blob digest audit
- OCI Distribution registry reference parser
- Docker Hub default registry resolution
- Docker Hub implicit library namespace resolution
- custom registry host and port references
- registry pull by tag
- registry pull by immutable SHA-256 manifest digest
- registry manifest-list platform selection
- registry platform defaulting to Linux host architecture
- OCI Distribution Bearer challenge handling
- scoped registry access-token requests
- registry username and password environment credentials
- HTTPS-only token realms for HTTPS registries
- registry manifest digest verification
- registry image config blob digest verification
- streaming registry rootfs layer digest verification
- registry descriptor size verification
- registry image config platform verification
- registry rootfs diff_id verification
- bounded registry manifest and config downloads
- bounded streaming registry layer downloads
- explicit unsupported registry layer media type errors
- temporary registry layer cleanup
- registry image import into local OCI layout
- digest-pinned pull mapped to stable local tag
- OCI Distribution image push by tag
- registry image push rejects immutable digest targets
- registry push requests pull and push token scope
- registry Bearer token reuse during upload
- registry blob mount before uploading content
- OCI monolithic registry blob upload
- registry relative upload Location resolution
- registry upload Location same-origin enforcement
- streamed large rootfs layer uploads from disk
- registry upload request replay after authentication challenge
- registry config and layer descriptor verification before upload
- registry rootfs diff_id verification before upload
- registry manifest publication after blobs
- registry published manifest digest verification
- verified local OCI bundle export for registry push
- private stable temporary layer copy for registry export
- registry image push CLI supports local and remote references
- registry credentials restricted to trusted token hosts
- registry host port validation
- registry manifest response content type validation
- registry image config media type validation
- manifest-list descriptor media type and size validation
- registry Content-Length validation for rootfs layer downloads
- registry image push supports localhost test registries
- registry cross-platform pulls preserve OS architecture and variant in the local OCI image config
- registry auth file explicit path support
- Docker config.json inline auth support
- Podman auth.json inline auth support
- base64 username-password registry auth decoding
- Docker Hub legacy auth host alias matching
- registry auth file symlink rejection
- registry auth file regular-file validation
- registry auth file size limit
- registry auth file JSON validation
- environment registry credentials precedence over auth files
- REGISTRY_AUTH_FILE credential discovery
- DOCKER_CONFIG credential discovery
- default Docker and Podman credential file discovery
- credential helper execution is explicitly disabled
- registry push authfile flag
- registry pull authfile flag
- event query by action type
- event query by exact container name
- event query by container ID prefix
- event query by relative time window
- event query by RFC3339 timestamp
- event filtering before result limiting
- safe volume restore from gzip-tar archive
- volume import uses rootfs archive path validation
- volume import rejects escaping symlinks and duplicate paths
- volume import stages extraction before publication
- volume import atomically publishes completed data
- volume import supports metadata labels
- volume import cleans up failed extraction
- portable volume export and import round trip
- volume logical disk usage report
- volume usage counts regular files and directories
- volume usage counts symlink target lengths
- volume usage never follows symlinks
- volume usage reports special entries separately
- volume usage exposes JSON and snapshot semantics
- volume clone by portable archive snapshot
- volume clone rejects a source in active use
- volume clone restores into a distinct target volume
- volume clone supports target metadata labels
- volume clone cleans temporary archive after completion
- volume clone uses hardened archive extraction
- ps quiet ID-only output
- ps newest-N selection across completed and active records
- ps non-truncated command display
- image list reference-prefix filtering
- image list operating-system filtering
- image list architecture filtering
- image list AND-combined filters
- image list rejects unsupported filters
- state directory symlink rejection
- state record symlink and file-type validation
- event journal symlink rejection
- container log symlink rejection on read and write
- image audit verifies manifest descriptor graph
- image audit verifies config descriptor digest
- image audit verifies compressed layer digest
- image audit verifies uncompressed rootfs diff ID
- image audit snapshots verified layer under store lock
- image audit rechecks snapshot digest before extraction
- image audit reports image platform metadata
- image audit reports manifest and config digests
- image audit reports layer media type and digest
- image audit combines integrity and filesystem findings in JSON
- run directly from a verified local image tag
- run image extraction is isolated in a private temporary directory
- run image temporary rootfs is removed after command completion
- run image dry-run verifies metadata without extracting the image
- run image and explicit rootfs options are mutually exclusive
- run image provenance is recorded as a container label
- safe non-executing JSON image build recipe
- image build strict required tag validation
- image build configurable recipe path
- image build context directory selection
- image build explicit COPY files and directories
- image build runtime environment metadata
- image build entrypoint and command metadata
- image build working-directory creation
- image build labels and user metadata
- image build JSON metadata output
- image build quiet reference output
- image build rejects unknown recipe fields and executable RUN steps
- image build bounded .boxrignore file
- image build .boxrignore comment and blank-line parsing
- image build .boxrignore basename glob matching
- image build .boxrignore nested path glob matching
- image build .boxrignore directory subtree exclusion
- image build refuses negated ignore patterns
- image build rejects ignored explicit COPY sources
- image build rejects symlinked ignore files
- image build rejects oversized ignore files
- image build creates parent directories for file COPY destinations
- image build explicit Linux target platform selection
- image build supports amd64 and arm64 target images
- image build supports ARM v5 v6 and v7 variants
- image build supports ARM64 v8 variant metadata
- image build supports 386 ppc64le riscv64 and s390x metadata
- image build rejects non-Linux and malformed platform declarations
- image diff compares two verified local image rootfs trees
- image diff reports added paths
- image diff reports removed paths
- image diff reports modified file contents and metadata
- image diff does not follow rootfs symlinks
- image diff supports JSON output
- image diff stages image extraction in a private temporary directory
- image diff removes staging files after comparison
- image lock pins a tag to its immutable manifest digest
- image lock records OS architecture and variant
- image lock uses a versioned strict JSON schema
- image lock file is written atomically
- image lock file is private mode 0600
- image lock rejects symlinked destinations and parents
- image lock rejects non-regular and oversized files
- image lock verification detects mutable tag drift
- image lock rejects unknown JSON fields and trailing data
- image lock supports JSON CLI output
- image SBOM inventories the verified rootfs filesystem
- image SBOM records the manifest digest and target platform
- image SBOM records regular file sizes and permission modes
- image SBOM generates per-file SHA-256 checksums
- image SBOM records symlink targets without following links
- image SBOM reports directories and symlinks separately
- image SBOM deterministic path ordering
- image SBOM configurable entry limit
- image SBOM rejects non-positive entry limits
- image SBOM fails rather than return a partial inventory
- image SBOM private temporary extraction cleanup
- rootfs archive normalizes entry modification timestamps
- rootfs archive normalizes access and change timestamps
- rootfs archive normalizes UID and GID metadata
- rootfs archive omits host account names
- rootfs archives are byte-identical for equivalent trees with different mtimes
- named runtime security profiles
- default security profile preserves explicit runtime settings
- baseline security profile enforces no-new-privileges
- baseline security profile enables the built-in seccomp filter
- baseline security profile drops NET_RAW
- restricted security profile disables container networking
- restricted security profile enables a private user namespace
- restricted security profile enforces no-new-privileges and seccomp
- restricted security profile drops high-risk Linux capabilities
- untrusted security profile requires an explicit root filesystem
- untrusted security profile enforces read-only rootfs
- security profiles apply after user flags so weaker flags cannot relax mandatory controls
- security profile is recorded in container metadata and dry-run plans
- strict JSON Schema for versioned run configuration
- config schema command emits a JSON Schema document
- config schema rejects unknown top-level properties
- config schema describes security profile options
- system storage usage summary
- system storage usage includes verified image-store bytes
- system storage usage includes named volume logical bytes
- system storage usage includes container metadata and logs
- system storage usage JSON output
- system storage scanner does not follow symlink entries
- system storage scanner rejects symlinked roots
- system storage scanner treats absent directories as empty
- named resource sizing profiles
- micro resource profile defaults
- small resource profile defaults
- medium resource profile defaults
- large resource profile defaults
- resource profiles fill only unset memory limits
- resource profiles fill only unset CPU quotas
- resource profiles fill only unset PID limits
- resource profiles fill CPU and I/O weights
- resource profile preserves explicit config limits
- resource profile preserves explicit command-line overrides
- resource profile name is validated
- resource profile appears in dry-run plans
- image list filter by ARM variant
- image list filter by digest prefix
- image list filter by metadata label presence
- image list filter by exact metadata label value
- image list filters combine with AND semantics across platform and metadata
- image list rejects empty metadata label filter keys
- privacy-preserving support bundle generation
- support bundle includes Boxr build and Go runtime metadata
- support bundle includes diagnostic names and statuses without detail strings
- support bundle excludes environment values and command arguments
- support bundle excludes container logs, state files, configuration and filesystem paths
- support bundle rejects symlink destinations and requires explicit overwrite confirmation
- support bundle archives and reports a SHA-256 digest
- support bundle archive and entries use private file permissions
- support bundle includes host kernel release metadata
- support bundle reports cgroup v2 availability
- live container statistics watch mode
- configurable container statistics sampling interval
- bounded container statistics sampling count
- timestamped JSON-lines container statistics output
- statistics watch exits cleanly when its container stops
- statistics watch rejects invalid intervals and counts
- statistics watch preserves the existing one-shot JSON interface
- statistics watch reports output and cgroup read errors
- live container event streaming
- event follow supports action and container filters
- event follow supports configurable polling intervals
- event follow supports bounded capture counts
- event follow streams compact JSON Lines or table rows
- event follow supports since-based historical catch-up
- event follow handles event log truncation by resuming
- event follow reports event source and output errors
- wait for container startup readiness
- wait for container process completion
- configurable lifecycle wait polling interval
- lifecycle wait enforces timeout for either condition
- lifecycle wait returns structured JSON results
- lifecycle wait reports containers that exit before becoming ready
- lifecycle wait supports multiple container IDs
- lifecycle wait preserves the default exit-code output
- container prune supports non-mutating dry runs
- prune dry run lists candidate container IDs and states
- prune dry run respects age thresholds
- prune dry run excludes running and never-started containers
- prune supports machine-readable JSON planning output
- prune dry run leaves container records and logs untouched
- named volume prune supports non-mutating dry runs
- volume prune previews unused volume names and labels
- volume prune reports volumes protected by active containers
- volume prune dry run respects name and label filters
- volume prune supports machine-readable JSON plans
- volume prune requires force for destructive cleanup
- container inspect optionally reports state and log disk usage
- inspect size report includes separate state and log byte counts
- inspect size report includes total bytes
- inspect size scanner treats an absent log as zero bytes
- inspect size scanner rejects symlink and special-file targets
- container stats supports human-readable table output
- container stats watch supports live table rows
- container stats table includes timestamp memory CPU and PID metrics
- container stats preserves JSON as the default output format
- volume du inventories every named volume
- volume du aggregates file and directory counts
- volume du aggregates logical bytes across volumes
- volume du all mode supports name and label filters
- volume du all mode supports JSON summary output
- image du inventories every tagged image
- image du reports per-tag rootfs layer bytes and platform
- image du all mode supports existing image metadata filters
- image du distinguishes referenced layer bytes from unique store bytes
- image du all mode supports JSON output
- image du preserves the existing unique-blob usage report
- image prune supports structured per-blob JSON reports
- image prune reports the size of each unreferenced blob
- image prune aggregates reclaimable bytes before deletion
- image prune dry-run preserves every unreferenced blob
- image prune retains explicit force confirmation for deletion
- standard Docker save archive import through image load
- plain tar and gzip-compressed Docker archive support
- bounded Docker archive extraction and strict path validation
- Docker archive manifest and config size validation
- Docker archive repository tag and platform validation
- multi-layer Docker root filesystem application in layer order
- Docker layer uncompressed SHA-256 diff_id verification
- Docker layer whiteout deletion semantics
- Docker opaque-directory whiteout semantics
- Docker image environment entrypoint command working-directory user and label preservation
- Docker archive multi-tag import with complete preflight validation
- Docker layer safe symlink and hard-link handling with special-file rejection
- Docker-save-compatible image archive export
- plain uncompressed tar output compatible with Docker load
- Docker image config filenames derived from SHA-256 config identities
- Docker archive repositories tag map for legacy-compatible consumers
- Docker export preserves rootfs diff IDs platform and runtime config
- Docker image layer is rehashed while expanding it for export
- format-selectable image save with the existing OCI format as default
- multi-image Docker save export from a single command
- combine tags with identical Docker image configurations into one manifest entry
- deduplicate shared rootfs layers across Docker image exports
- deterministic ordering of Docker manifest config and layer entries
- enforce Docker export image-count and duplicate-reference limits
- standard OCI gzip layer media type support
- standard OCI uncompressed tar layer media type support
- Docker distribution gzip layer media type support
- Docker distribution uncompressed tar layer media type support
- bounded multi-layer registry manifest validation
- per-layer compressed registry blob digest verification
- per-layer uncompressed rootfs diff_id verification
- ordered multi-layer registry filesystem application
- registry layer whiteout and opaque-directory semantics
- versioned multi-service deployment manifests
- strict bounded deployment manifest loading
- deployment service and stack name validation
- per-service command and runtime option validation
- deployment dependency reference validation
- duplicate and self-dependency rejection
- deployment dependency cycle detection
- deterministic topological deployment ordering
- deployment plan service resource summaries
- secret-redacted JSON deployment plans
- deployment plan example and validation commands
- local OCI image references in deployment services
- stable stack-scoped service container names
- reserved deployment identity labels
- attached multi-service deployment supervisor
- dependency-ordered service process startup
- service running-state startup gates
- fail-fast service exit handling
- reverse-order graceful supervisor shutdown
- stack-scoped deployment process status
- reverse-dependency-order deployment shutdown
- deployment stop timeout escalation
- safe optional deployment record and log removal
- deployment lifecycle dry-run planning
- parallel startup of dependency-independent deployment services
- deployment-wide unique registry image prefetch
- deployment pull platform selection
- deployment pull registry auth-file support
- canonical local image reference resolution before service start
- deployment pull validates cached image architecture and variant
- image-backed service startup enforces the host Linux platform
- deployment TCP readiness probes for loopback services
- deployment HTTP readiness probes for loopback services
- readiness targets reject non-loopback destinations and URL credentials
- readiness probe intervals timeouts attempts and status codes are bounded
- readiness retries honor supervisor cancellation and startup deadlines
- secret-redacted readiness metadata in deployment plans
- readiness validation accounts for mandatory network-isolating security profiles
- dependent service startup waits for configured readiness probes
- parallel service waves select readiness probes by service identity
- deployment-wide wait for all services running
- deployment-wide wait for all services stopped
- bounded stack-state polling with configurable deadlines
- secret-redacted JSON deployment wait results
- non-executing image recipes support scratch and local OCI image bases
- local image bases are integrity-verified before reuse
- image base platforms must match requested Linux build targets
- base filesystems are safely extracted before context COPY operations
- base environment variables inherit with deterministic key-based overrides
- base entrypoint command working directory user and labels inherit with build overrides
- stack-wide service-prefixed deployment logs
- deployment logs filter by service name
- follow mode streams and drains all service logs
- image build writes resolve in-root symlinks and reject host-root escapes
- container-absolute COPY symlinks are normalized inside the staged rootfs
- versioned deployment named persistent volume mounts
- named volume names and mount targets validated before startup
- deployment preflights every named volume before creating services
- named volumes reuse Boxr’s secure persistent volume store
- volume archive exports never overwrite an existing destination
- volume archive exports publish atomically from a same-directory staging file
- volume export JSON includes SHA-256 integrity and filesystem manifest metadata
- versioned deployment service profiles
- validated lowercase service profile identifiers and duplicate rejection
- unprofiled baseline services remain enabled by default
- repeatable command-line service profile selection
- additive activation of services matching any selected profile
- profile-disabled services are omitted from deployment plans
- full manifest validation occurs before profile filtering
- unknown deployment profiles are rejected
- profile-filtered dependency graphs reject inactive dependencies
- deployment image prefetch and startup are limited to selected profiles
- deterministic service profile inventory
- profile inventory groups services by opt-in profile
- always-enabled services appear separately from opt-in profiles
- services assigned to multiple profiles appear in every matching group
- secret-free JSON and text profile inventory output
- per-service ordered environment file loading in deployments
- deployment environment files override service-inherited values in declaration order
- explicit service environment entries override environment file values
- environment file path traversal and duplicate path rejection
- deployment environment file count and path length bounds
- secret-free environment file metadata in deployment plans
- per-service startup timeout overrides the deployment-wide default
- zero per-service startup timeout inherits the deployment-wide default
- service startup timeout bounds are validated between 10 milliseconds and 10 minutes
- secret-free deployment plans expose per-service startup timeout metadata
- per-service startup timeouts govern process and readiness startup gates
- per-service shutdown grace overrides the stack-wide deploy down timeout
- zero per-service shutdown grace inherits the stack-wide command timeout
- service shutdown timeout bounds are validated between 10 milliseconds and 10 minutes
- per-service shutdown grace is persisted as reserved non-secret runtime metadata
- attached deployment rollback and interrupt shutdown honor service-specific grace periods
- deploy down applies each service-specific shutdown deadline independently
- only services exceeding their own shutdown grace are escalated to SIGKILL
- recurring per-service TCP or HTTP healthchecks run after startup gates pass
- healthcheck intervals are bounded from 10 milliseconds to 60 seconds
- healthcheck retry budgets are bounded and count consecutive failures
- successful healthchecks reset the consecutive failure counter
- healthcheck startup grace periods are bounded to 10 minutes
- unhealthy services fail the attached deployment supervisor and trigger stack shutdown
- healthcheck targets remain loopback-only and absent from deployment plans
- managed bridge network creation with an owner-private persistent metadata store
- versioned network store with bounded strict JSON decoding
- network metadata writes are atomically published with private file permissions
- IPv4 bridge subnets are canonical and bounded between /16 and /29
- bridge network gateways are validated as usable in-subnet addresses
- bridge networks reject overlapping IPv4 subnets
- bridge interface names are validated against Linux interface length limits
- bridge provisioning creates and raises a host Linux bridge interface
- bridge networks optionally enable scoped IPv4 forwarding and NAT
- network-specific firewall chains isolate Boxr forwarding and masquerade rules
- network deletion removes only its own firewall chains and bridge interface
- per-network IPv4 address allocation excludes network broadcast and gateway addresses
- container network attachments reserve unique addresses by container ID
- bridge attachments create host-side veth pairs and join container net namespaces
- container bridge interfaces receive an eth0 address and default gateway
- failed bridge attachment rolls back its veth and IPv4 allocation
- container exit removes its bridge veth and releases its address
- bridge network removal is blocked while container attachments remain
- network list and inspect provide sorted secret-free JSON metadata
- bridge mode requires an explicit managed network name
- bridge mode creates a private network namespace before host-side attachment
- container run joins and leaves bridge networks through lifecycle hooks
- network metadata rejects duplicate IDs names bridges and invalid allocations
- host TCP port mappings support IPv4 host bind addresses and container destination ports
- host UDP port mappings support IPv4 host bind addresses and container destination ports
- published port protocol defaults to TCP and accepts explicit TCP or UDP
- published port bindings validate host and container port ranges from 1 to 65535
- published port bindings normalize wildcard IPv4 host address and protocol
- wildcard host port mappings conflict with specific address mappings on same port
- host port reservations are checked across all managed bridge networks
- bridge port mappings install destination NAT rules for inbound traffic
- bridge port mappings install host OUTPUT NAT rules for host-originated connections
- published port forwarding allows new traffic only to mapped container addresses and ports
- published port firewall rules are removed when bridge container exits
- partial port mapping firewall setup is rolled back with veth and IPAM state
- published port mappings are persisted with network leases in private network metadata
- network store validates canonical published port mapping and cross-network reservation uniqueness
- run CLI supports repeatable –publish and -p mappings with IPv4 host bind and protocol syntax
- run config JSON schema describes strict published port mapping objects
- dry-run plans show normalized published port mapping values without starting container
- published ports are accepted only with a named managed bridge network
- deployment manifests validate service published port mappings through container option rules
- deployment service plans expose normalized TCP and UDP published port bindings
- deployment subprocess runs apply and release published port rules through the container lifecycle
- Linux network store transactions use an advisory cross-process lock for IPAM and port reservations
- Linux network store lock files are opened without following symlinks and restricted to owner read/write
- non-Linux CLI builds retain process-local network store locking without claiming runtime support
- published port setup checks exact PREROUTING OUTPUT and FORWARD rules before insertion
- published port rollback removes only firewall rules inserted by the current attachment attempt
- failed firewall cleanup retains the IPAM lease and host-port reservation until cleanup can be retried
- live network connect attaches a running Boxr container with a private network namespace to a managed bridge
- live network disconnect removes a running container’s managed bridge interface and address lease
- live bridge attachment requires the container to have been started with network mode none
- live bridge attachment rejects host-network containers before attempting namespace entry
- live bridge attachment rejects user-namespace containers when host-side namespace control is unavailable
- live network connect verifies the target PID executable and Boxr namespace-helper command line
- live network connect supports repeatable TCP and UDP published host ports
- container exit detaches dynamically connected managed networks even when they were not present at run startup
- container exit releases dynamically connected published port rules veth links and IPAM reservations
- network detach is idempotent when a container has no remaining managed network lease
- managed bridge IPAM prevents a single container from acquiring conflicting eth0 attachments on multiple networks
- managed veth interfaces carry a network and container-specific Boxr ownership alias
- bridge disconnect verifies the Boxr ownership alias before deleting a host veth
- stale network attachment discovery filters leases against verified live Boxr helper PIDs
- network prune defaults to a read-only stale attachment report
- network prune requires an explicit force flag before changing host interfaces or firewall rules
- stale attachment pruning removes only Boxr-owned veths and matching published port rules
- stale attachment pruning releases the corresponding IPv4 IPAM lease after successful cleanup
- network prune retains reservations when a veth has a foreign or missing ownership alias
- network prune reports removed attachment IPv4 and published port counts
- runtime process liveness checks validate the Boxr helper command line and network namespace
- bounded parallel OCI and Docker registry layer downloads
- registry layer download worker count capped at four
- parallel registry layer verification preserves manifest order
- registry pull cancels outstanding layer downloads after the first failure
- parallel registry pull removes every partial temporary layer on failure
- concurrent registry requests share a synchronized Bearer token refresh
- OCI registry pulls accept Zstandard-compressed tar layers
- Zstandard registry layer digest verification streams decompressed bytes
- Zstandard decompression caps decoder memory and window at 256 MiB
- rootfs layer application auto-detects Zstandard-compressed tar archives
- pulled Zstandard layers are verified against OCI uncompressed diff IDs before application
- Linux detached run launches a dedicated Boxr supervisor in a new session
- detached run waits for a post-cgroup startup-barrier handshake before returning the container ID
- detached supervisor captures launch diagnostics in an owner-only state-directory log
- detached run reports container ID host PID and launch log path after successful startup
- detached supervisor uses the standard persistent container lifecycle record and log stream
- detached mode rejects dry-run and preserves workload arguments through the standard flag parser
- registry pushes convert Boxr’s private rootfs layer media type to standard OCI gzip layer media type
- registry pushes preserve compressed layer bytes and SHA-256 digest during OCI media type normalization
- registry push verifies the published digest of the portable normalized OCI manifest
- detached container restart policies support no on-failure always and unless-stopped modes
- on-failure restart policy evaluates persisted non-zero container exit status
- on-failure restart policy supports a bounded retry count with zero meaning unlimited
- detached restart delay defaults to one second and is bounded from 100 milliseconds to five minutes
- restarted container attempts retain one persistent container ID and one log stream
- container state records persist restart attempt count and previous exit status
- manual stop intent is persisted before termination signals and suppresses restart policies
- boxr stop cancels a pending restart delay without signalling a non-existent process
- terminating boxr kill signals cancel a pending restart delay and suppress future attempts
- restarting containers remain visible in active container listings during backoff
- container removal is blocked while a restart attempt is pending
- restart policies require detached execution and reject automatic record removal
- Dockerfile parser enforces byte line and instruction-count limits
- Dockerfile builds support scratch and verified local single-stage FROM bases
- Dockerfile COPY supports local files directories and the build-context root
- Dockerfile WORKDIR resolves relative paths inside the container rootfs
- Dockerfile ENV and LABEL support quoted values and variable expansion
- Dockerfile USER and shell or exec CMD and ENTRYPOINT metadata are preserved
- Dockerfile builds apply bounded .dockerignore exclusions without mutating the context
- Dockerfile RUN ADD multi-stage builds and unsupported flags fail explicitly
- Dockerfile LABEL accepts dotted and namespaced metadata keys
- Dockerfile COPY supports multiple local sources with inferred destination directories
- Dockerfile RUN shell and exec instructions execute inside Boxr namespaces
- Dockerfile RUN commands operate on the staged image rootfs after preceding build steps
- Dockerfile RUN receives a clean host-independent environment with ARG and ENV values
- Dockerfile RUN defaults to an isolated network namespace
- Dockerfile RUN host networking requires an explicit –network host opt-in
- Dockerfile RUN enforces user namespaces no-new-privileges baseline seccomp and drops all capabilities
- Dockerfile RUN failures abort the build before the image tag is published
- Dockerfile RUN after a non-root USER is rejected rather than silently run as root
- exec runs an additional command in an existing container
- exec enters pinned mount UTS IPC network PID and optional user namespaces
- exec launcher joins the container resource cgroup before namespace entry
- exec inherits container environment and supports environment and env-file overrides
- exec reapplies configured capability no-new-privileges and seccomp restrictions
- exec forwards standard input output and error and propagates command exit status
- container records capture Linux process start-time ticks to detect PID reuse
- Dockerfile build accepts repeatable –build-arg KEY=VALUE overrides
- Dockerfile ARG defaults can be overridden by explicit build arguments
- pre-FROM ARG values parameterize FROM and require redeclaration inside the stage
- build argument values are available to RUN but are not persisted as image ENV
- invalid and unused Dockerfile build arguments are rejected
- Dockerfile COPY relative destinations resolve against the base image WORKDIR
- Dockerfile build API accepts explicit network and build-argument options
- Dockerfile ADD supports local context files and directories
- Dockerfile ADD automatically extracts plain tar archives
- Dockerfile ADD automatically extracts gzip-compressed tar archives
- Dockerfile ADD archive extraction applies rootfs path traversal link and special-file protections
- Dockerfile ADD archive extraction enforces bounded file counts and uncompressed byte limits
- Dockerfile ADD rejects remote URL and Git sources instead of silently fetching them
- deployment loader accepts a bounded single-document Docker Compose YAML file alongside strict JSON manifests
- Compose YAML rejects unknown fields duplicate mapping keys and multiple documents
- Compose services translate literal image references and safely tokenized command strings or arrays
- Compose environment maps and lists resolve pass-through keys and relative env_file paths
- Compose depends_on supports deterministic service_started dependency ordering
- Compose profiles labels working directories and hostnames map to Boxr deployment options
- Compose short-form IPv4 TCP and UDP port mappings map to Boxr published bridge ports
- Compose services attach only to explicit pre-created external Boxr bridge networks or explicit host/none modes
- unsupported Compose builds implicit volume creation implicit networks healthchecks TTY privileged mode and restart policies fail explicitly
- exec supports explicit -i stdin forwarding and closes stdin by default
- exec -t allocates a pseudo-terminal with a controlling session and combines terminal output
- exec TTY mode propagates SIGWINCH terminal size changes and restores host terminal settings on exit
- Compose external named volumes map to pre-created Boxr-managed volumes
- Compose relative host bind paths resolve from the Compose file directory
- Compose long-form bind volume and tmpfs mounts support validated existing sources external no-copy volumes and bounded tmpfs
- Compose bind sources reject missing paths symlink components and special files
- Compose bind and external volume mounts support explicit read-only and read-write modes
- image-backed deployments defer bind mounts until after image rootfs extraction
- Compose-created volumes volume drivers unsupported volume options copy-up subpaths host path creation bind relabeling and propagation other than rprivate are rejected
- Compose read_only maps image-backed services to a read-only root filesystem
- Compose cap_drop validates and applies supported Linux capability drops
- daemon serves the initial Docker Engine API through a local Unix domain socket only
- daemon refuses TCP exposure symlinked ancestors shared writable socket directories and non-socket path replacement
- daemon detects active and stale sockets uses owner-only 0600 permissions and removes only its own socket
- Engine API implements Docker ping and negotiated version information within API v1.24 to v1.45
- Engine API exposes runtime-backed info and filtered container list metadata
- Engine API exposes container inspect state mounts network port bindings and restart metadata
- Engine API streams bounded tail and follow container logs in Docker multiplexed stdout frames
- Engine API exposes cgroup v2 CPU memory and PID stats with snapshot and streaming responses
- Engine API streams persisted container lifecycle events and lists OCI image tags
- unsupported Docker Engine restart attach image volume network mutation and TCP listener operations are explicitly outside the API contract
- Engine API creates containers from verified local images using a strict supported Docker create subset
- Engine API container create supports validated structured bind volume and tmpfs mounts with explicit unsupported-option rejection
- Engine API starts created containers through the native Boxr detached supervisor
- Engine API gracefully stops containers with a bounded timeout and SIGKILL fallback
- Engine API sends validated Docker kill signals only to process-identity-verified running containers
- Engine API removes created and stopped containers through Docker DELETE with safe conflict handling
- Engine API inspects verified local OCI images using Docker-compatible image metadata and uncompressed layer DiffIDs
- Engine API removes unused local image tags and rejects removal while containers reference them
- Engine API pulls and verifies OCI/Docker registry images using Docker X-Registry-Auth and Linux platform selection
- Engine API pulls accept immutable SHA-256 digest references and map them to deterministic local image aliases
- Engine API pushes verified local single-layer OCI images using Docker X-Registry-Auth and registry upload verification
- Engine API creates Docker image tag aliases from verified OCI content and protects tags referenced by containers
- top-level pull command reuses verified OCI and Docker registry image downloads
- top-level push command reuses verified OCI and Docker registry image uploads
- top-level build command supports Docker-style -t image tagging and reuses the verified Dockerfile builder
- top-level images command lists the verified local image store using Docker-style output flags
- top-level tag command creates local image aliases without copying content-addressed blobs
- top-level rmi command removes unused local image tags while preserving shared content-addressed blobs
- container namespace exposes Docker-style lifecycle subcommands through the native verified Boxr operations
- native rm safely removes created containers without weakening running-container protection
- native rm –force stops running containers or cancels pending restarts before safe removal
- native run accepts positional verified local image references while preserving ordinary command execution
- state store renames containers with unique names and persists rename lifecycle events
- native rename and container rename commands preserve container IDs while changing unique names
- Engine API POST containers rename updates unique container names and returns Docker-compatible errors
- compose command aliases the supported Compose YAML deployment validation planning pull and lifecycle workflows
- Engine API POST containers wait blocks for not-running state and returns the persisted workload exit status
- container prune previews and filters stopped records and requires explicit force before removal
- Engine API DELETE force true delegates to native safe forced removal for running or restarting containers
- native create command records a verified image-backed container without starting it
- native start command launches a created image-backed container through the detached supervisor
- native start command restarts stopped image-backed containers through the detached supervisor
- native restart command stops then starts an image-backed container with a bounded graceful timeout
- Engine API start endpoint restarts stopped image-backed containers through the native detached supervisor
- Engine API restart endpoint performs a bounded graceful stop and launches the same saved container again
- Engine API lists local named volumes with name label driver scope and dangling filters
- Engine API creates local named volumes with validated labels and the local driver
- Engine API inspects named volume metadata and the private mountpoint
- Engine API removes named volumes only when no persisted container references them
- Engine API container create resolves pre-created Boxr named volumes from Docker HostConfig.Binds
- Engine API container inspect reports named volume type name driver and Docker-compatible Binds
- Engine API lists managed bridge networks with Docker name driver scope and label filters
- Engine API creates IPv4 bridge networks with validated IPAM options and managed NAT controls
- Engine API inspects managed bridge network subnet gateway bridge interface and labels
- Engine API deletes managed bridge networks only when no containers remain attached
- managed bridge networks persist validated Docker-compatible network labels
- Engine API creates anonymous local volumes when the create request omits Name
- Engine API prunes unused anonymous volumes by default with Docker-compatible label filters
- Engine API prunes unused named volumes when the all filter is true
- Engine API prunes unused managed bridge networks with label and until filters
- Engine API protects persisted container network references during network pruning
- Engine API exports one or more verified local image tags as Docker-save archives
- Engine API imports Docker-save archives after tag preflight and protects images referenced by containers
- Engine API builds images from bounded Docker tar contexts using the supported single-stage Dockerfile subset
- Engine API Docker builds support platform selection declared build arguments and multiple tags
- Engine API maps Docker default build networking to isolated none and requires explicit host network opt-in
- Engine API refuses to replace image tags referenced by persisted containers during builds
- Engine API prunes dangling images by default and unused tagged images when dangling is false
- Engine API applies Docker-compatible image prune until label and label-exclusion filters
- Engine API protects all aliases sharing image digests used by persisted containers during prune
- Engine API removes selected image tags atomically and reports reclaimed content-addressed blob bytes
- Dockerfile COPY supports safe octal –chmod modes for copied files
- Dockerfile multi-stage builds support named and numeric earlier-stage COPY –from references
- Dockerfile COPY –from supports verified local image references without implicit network pulls
- Docker Engine API creates starts inspects and streams non-TTY exec sessions through Boxr namespace-aware execution
- Engine API prunes stopped and created containers using Docker until and label filters and emits Docker prune events
- Engine API can expose remote TCP access only through mutually authenticated TLS with client certificates verified by a configured CA
- Engine API-created containers persist writable OverlayFS layers across stop and restart and remove owned storage with container deletion
- Engine API containers retain a private immutable image rootfs snapshot across stops, survive local image-tag deletion, and remove owned snapshots with container deletion
-
Engine API volume mounts populate empty pre-created local volumes from the immutable image rootfs by default and honor Docker VolumeOptions.NoCopy
- Linux cgroup v2 freezer transitions verify cgroup membership and kernel frozen state before reporting pause or unpause success
- Failed cgroup freezer transitions restore the previous freeze state and confirm rollback
- Container pause and unpause persist process-identity-checked lifecycle states and events with rollback on persistence failure
- Native CLI and Docker Engine API pause/unpause paths serialize lifecycle transitions and resume paused workloads before stop or kill
- Docker Engine inspect reports State.Paused and State.Running consistently with persisted paused status, while prune protects paused containers
- Docker Engine container changes reports bounded added changed and deleted paths from managed image and writable filesystem layers
- Docker exec supports validated USER[:GROUP] overrides and otherwise inherits the container’s configured identity
-
Container health monitors suspend probes while paused and discard probes overlapping pause/unpause so freezer delays do not create false unhealthy transitions
- Managed bridge endpoints persist a stable Docker endpoint ID and locally administered unicast MAC address, apply the MAC to the container interface, and remove both identities on disconnect
- Docker container inspect and network inspect expose the managed endpoint ID and configured MAC address from the same persisted IPAM endpoint
- Docker container list supports network-name/network-ID-prefix, health, exited-code, and negative-label filters against live managed endpoint metadata
- Docker exec supports upgraded stdin streams, non-TTY Docker stdout/stderr multiplex framing, Linux TTY/PTY raw streams, and bounded PTY resize requests
-
Engine API stdin socket placement falls back to a private per-user short path when a configured state directory exceeds Unix-domain socket path limits
- Docker network connect supports validated static IPv4 endpoint addresses and explicit unicast MAC addresses, persists them in IPAM state, and configures the container interface
-
Docker EndpointConfig rejects unsupported aliases links IPv6 link-local addresses and driver options instead of silently dropping them
-
Docker bridge IPAM persists, validates, inspects and allocates only within configured IPRange prefixes, including explicit static endpoint address checks
-
Docker bridge IPAM auxiliary IPv4 reservations are persisted, returned by inspect, excluded from dynamic allocation and rejected for static endpoint assignment
-
Docker container listing evaluates ancestor image, volume name/source/destination, published host/container port and exposed port/range filters
-
Docker container list before and since filters resolve existing container IDs/names once per request and compare persisted creation times; invalid and unknown filter inputs return errors
-
Docker container list and inspect size=1 return SizeRw and SizeRootFs logical-byte counters using bounded, symlink-safe base-rootfs and OverlayFS-upper walks
- Docker container create
Tty launches the workload under a controlling PTY, merges terminal output, forwards signals to the PTY process group and preserves workload exit status