YANG is a protocol-independent language for describing structured configuration, operational state, notifications and remote procedure calls. Defined originally in RFC 6020 and updated in RFC 7950 (YANG 1.1), it lets a system describe data nodes, types, constraints and relationships in a machine-readable model. NETCONF and RESTCONF are common ways to access YANG-modeled data, and other interfaces can expose similar models. A YANG module is not itself a guarantee that a particular device implements every modeled feature, supports an edit operation, or offers an identical datastore.
Model-driven network management builds on, rather than simply replaces, CLI and SNMP. SNMP Management Information Bases are machine-readable and remain useful for monitoring, while device CLIs often expose vendor-specific command behavior. YANG adds typed, hierarchical descriptions of configuration and state, including constraints and reusable structures, for protocols such as NETCONF and RESTCONF. It does not make every platform interchangeable. Engineers gain a more explicit schema to test against the device capabilities actually advertised, not a promise that one request works identically on all network operating systems.
Origins of IETF Models
IETF publishes open standards and associated YANG modules through its working-group and review process. A working draft and a completed RFC carry different maturity and stability signals; not every proposed module reaches a published standard. RFC 8343, RFC 8344 and RFC 8349 are examples of published interface, IP and routing models, respectively. Their value is a documented set of names, data types and intended semantics, but implementation remains a platform question. Check the revision and supported features advertised by each server before relying on a model for configuration changes.
IETF-published YANG modules such as ietf-interfaces (RFC 8343), ietf-ip (RFC 8344) and ietf-routing (RFC 8349) give vendors and operators shared definitions for common subjects. Publication of a model does not require every network device to implement that model, every feature in it, or every optional deviation. The module’s actual presence, version, enabled features and deviations must be discovered on the target device. IETF models provide standards-based semantics where implemented, but operator automation cannot assume a uniform feature floor merely from an RFC title.
OpenConfig Initiative Background
OpenConfig is an operator-driven effort to provide consistent data models that serve common network management and telemetry needs across multiple vendors. Its YANG modules are openly maintained, with a public repository and documented schema conventions. The project does not claim to exhaustively model every platform feature or to standardize how every implementation performs an operation. Teams can use OpenConfig to reduce platform-specific code where a common supported subtree exists, while retaining vendor-specific handling for features outside that intersection.
OpenConfig represents operator-driven expectations for configuring and observing network systems. Its models commonly distinguish intended configuration from reported operational state using config and state containers, which can make the difference between a requested setting and an observed result explicit. The structure is a design convention across many modules, not a promise that every object on every vendor has identical paired nodes. Telemetry transport, update frequency, available paths and authorization still depend on the platform and management interface; a model alone does not create continuous streaming telemetry.
Cisco Native Model Philosophy
Cisco provides native YANG models for IOS XE, IOS XR and NX-OS feature families as well as support for selected open models. Native models often expose platform-specific capabilities that a generic IETF or OpenConfig definition omits. They are not a complete mechanical transcription of every CLI command or device-state variable. Cisco’s own Open IOS XE FAQ says native models cover the most frequently used features rather than 100% of its historical functionality. Model availability also differs by network OS, hardware, software release and feature configuration.
A Cisco-native model can be the right choice when an automation task needs a supported feature unavailable through the selected open model. That advantage is conditional on the exact device and revision: the developer should check its YANG library, advertised features, deviations and datastore behavior before writing an edit request. Scripts bound to vendor-native paths tend to be less portable to other platforms, and unsupported commands may still need a different configuration mechanism. The trade-off is access to implemented platform-specific capabilities, not guaranteed full CLI fidelity.
Structural Differences Between Models
The meaningful differences among IETF, OpenConfig and Cisco-native models are the modules’ intended scope, actual tree structure, status and implementation support. An IETF RFC may define common types and related modules, OpenConfig favors a consistent operational schema across frequently used network features, and a Cisco-native module may expose a product-specific configuration concept. None of those descriptions determines whether the data tree is flat, fully writable or complete on a particular release. Inspect the concrete model and advertised deviations; a broad label such as ‘standard’ or ‘native’ is not a substitute for testing its path and allowed operations.
OpenConfig models often organize common configuration and state under recognizable containers, while Cisco native and IETF modules reflect different ownership and design goals. The specific XML/JSON paths, list keys and feature boundaries must be read from the schema actually advertised by the device. Native YANG paths may resemble familiar CLI concepts, but they do not necessarily match the CLI hierarchy node for node. Operators should learn to navigate module namespaces and actual data trees rather than assuming a command can be pasted into a similarly named YANG path.
Namespace and Module Organization
A YANG module declares its XML namespace, and a client uses that namespace together with the module name and revision to interpret data nodes correctly. Common IETF modules use published ietf.org namespaces, OpenConfig modules use their own defined namespaces, and vendor-native modules use vendor-specific identifiers. The same standard module namespace supports consistent interpretation where implemented; it does not prove identical enabled features, schema revisions, datastore support or writable paths across vendors. Automation that combines model families must resolve names precisely and discover the device’s actual module set rather than guessing prefixes from CLI terminology.
OpenConfig modules use namespaces rooted at openconfig.net and prefix module names with “openconfig-” followed by the feature area, such as openconfig-interfaces, openconfig-bgp, or openconfig-platform. Cisco native modules use namespaces that reflect the specific operating system and feature area, with naming conventions like Cisco-IOS-XE-native for core IOS XE configuration or Cisco-IOS-XR-infra-intf-cfg for IOS XR interface configuration. When building automation tools that must consume multiple model families, the namespace differences require careful handling to avoid conflicts and ensure that XPath expressions and filter operations target the correct model hierarchy on each device.
NETCONF and RESTCONF Support
On devices that support and enable the relevant services, IETF, OpenConfig and vendor-native models can be consumed through NETCONF and, where implemented, RESTCONF. NETCONF (RFC 6241) uses XML-encoded RPCs and capability-dependent datastores; RESTCONF (RFC 8040) maps supported YANG resources to HTTP operations, commonly with JSON or XML encoding. A Cisco software release number by itself does not prove that a given hardware platform, model revision or RPC is enabled. Confirm management access, authentication, service configuration, datastore support and operational restrictions for the exact device.
During NETCONF session setup, the server advertises protocol capabilities in its hello message. Do not treat that message as a universal complete list of the YANG modules and revisions usable by every modern implementation. In particular, Cisco’s IOS XE guidance for YANG 1.1 directs clients to the ietf-yang-library data tree for module discovery instead of extracting the entire module catalog from hello. An automation client should collect the YANG library, feature and deviation data and compare those facts with the requested operation before constructing an RPC. RESTCONF implementations expose their own discovery paths under the supported YANG-library model.
BGP Configuration Comparison
BGP is a useful way to compare model families because a seemingly simple neighbor can involve address families, route-policy references, peer groups, timers and counters. An IETF module can define standard BGP-oriented structures, but draft and published module status should be distinguished and device support verified before planning operations. A model’s feature list does not establish whether a particular router accepts an edit for route-reflector settings or a specific policy construct. Start with an implemented, versioned module and a real read-only query instead of assuming that every item in a generic BGP schema is actionable.
OpenConfig BGP models commonly represent neighbors, peer groups, address families, policies and state paths in ways operators can reuse across platforms that implement the relevant portions. Cisco, Juniper and other vendors may implement different revisions, optional paths and deviations. A Cisco native BGP model may expose a platform-specific feature omitted from the chosen OpenConfig representation, but no published schema should be described as containing every configuration option in IOS XE or IOS XR. Test a read-only request, inspect actual modules and compare intended edits with the device’s supported capability before relying on the path in production.
Interface Management Across Models
Interface management provides a practical cross-platform comparison. RFC 8343 describes common interface data and RFC 8344 addresses IPv4/IPv6 interface configuration. Implementations may expose these modules, parts of them, vendor augmentations or different release-specific features. Their publication does not make support mandatory on every compliant router or switch. Successful automation discovers the device’s actual modules and deviations, then validates interface naming, edit support and resulting operational state on the target platform.
OpenConfig interface models commonly cover logical and physical interfaces, subinterfaces and operational counters. Cisco native interface models can expose IOS XE-specific settings, including some platform features outside common open models. Neither family should be assumed complete: supported nodes and writable behavior vary by software and hardware. In multi-vendor automation, a reasonable design is to use an open model where the required device implementations align and a vendor-native model for a verified gap, while documenting that exception and its test coverage.
Automation Tool Compatibility
Automation tools can consume YANG-modeled configuration and state only where both the device and the chosen client support the relevant protocol and schema. Ansible modules, libraries and collection versions vary in whether they expose NETCONF, RESTCONF, platform-specific APIs or CLI transports. A module designed to push CLI commands does not become a model-driven YANG client simply because the target device also supports NETCONF. For a portable workflow, define the desired operation, inspect the target’s advertised models and select the client implementation that can safely perform that operation on every supported platform.
Python clients such as ncclient can issue NETCONF RPCs against a server that permits the requested operation, but a successful connection does not prove that a particular YANG module is implemented. Netmiko primarily automates device CLI interactions over SSH and should be treated as a different interface, not a NETCONF fallback with equivalent transaction semantics. Differences in sessions, datastore locks, validation, rollback and error handling matter when selecting an automation path. Test read-only queries first, then use a narrowly scoped change with an explicit recovery strategy before expanding a write workflow.
One of the most important but often overlooked aspects of real-world YANG model deployment is the use of deviations and augmentations, which allow vendors to customize standard models to reflect the actual capabilities and limitations of their platforms. A deviation statement in YANG allows a device vendor to declare that a specific node from a standard model is not supported, has a different data type, or has different constraints than the standard specifies. Augmentation allows a vendor to add new nodes to a standard model without modifying the original, extending it with platform-specific capabilities while preserving compatibility with the base standard.
The support of augmentations, deviations, optional features and module revisions is specific to the server’s implemented schema. Cisco operating systems may advertise both standard and native modules, sometimes with feature limitations or deviations from a generic model. That is a reason to retrieve the YANG library and inspect the schema and enabled features, not to assume either complete OpenConfig support or deliberate divergence on every path. A successful read from one release does not prove an edit will work after a software upgrade or on a different hardware family.
Practical Model Selection Guide
Choosing a model family requires an inventory of devices, software releases, actual management services and the operations the automation must perform. OpenConfig can offer portability where the same modules and paths are implemented consistently, while IETF modules can provide common vocabulary for widely modeled functions. Neither is automatically the right first choice for every device or task. Compare read responses, candidate edit semantics, supported datastores, deviations and operational rollback before standardizing a codebase. A cross-vendor label is a starting hypothesis, not proof of identical behavior.
For a Cisco-focused estate, native models may be useful for supported platform-specific capabilities, while OpenConfig or IETF models can reduce vendor coupling for operations that both device and module actually support. Many teams will need a documented mixture rather than one universal schema. Maintain a per-feature matrix of model namespace, revision, required paths, supported platforms, permissions, expected state and verification tests. Recheck it when software or hardware changes; a successful example on one release does not establish compatibility with the next device. This is the practical way to make model-driven automation maintainable rather than treating model names as guarantees.