Comparison of OPC UA complex data models

From OPC Labs Knowledge Base

OPC UA offers two complex data models. They do not have "official" names, but the historically first one is based on the DataTypeDictionary type, and the second one is based on the DataTypeDefinition type. We will therefore usually refer to them as "DataTypeDictionary model" (or "data type dictionary model"), and "DataTypeDefinition model" (or "data type definition model").

The table below summarizes the main difference between the two models.

DataTypeDictionary Model DataTypeDefinition Model
OPC UA version 1.00 and later (deprecated in 1.04 and later, but still available) 1.04 and later
Described in OPC UA Specification Part 5 (Information Model) OPC UA Specification Part 3 (Address Space Model)
Compatible with OPC UA Client-Server yes yes
Compatible with OPC UA PubSub no yes
Complexity of implementation, and number of OPC UA operations needed to work with the model high somewhat lower
Extensibility Two "data type systems" (how are data types described, and their resulting encoding) are defined by OPC Foundation (OPC Binary data type system, and XML Schema data type system). In theory, anybody can add new data type systems, but we do not know of any being already defined, and it is very unlikely that they ever will. Not extensible (besides the fact that any OPC UA encoding defined becomes supported under this model).
OPC Binary Data Type System XML Schema Data Type System
How are data types described using a XML document conforming to OPC Binary XML schema defined by OPC Foundation using XML Schema (commonly referred to as XSD), defined by W3C using OPC UA structures
Resulting encoding Binary XML any encoding defined in OPC UA (Binary, XML, JSON)
Suitable for small devices yes usually no (XML schema is too big) yes
Can refer to data types in other servers yes yes (details not specified) no
Data type inheritance no no yes
Selectable byte order yes no no
Simple data types yes as per XML Schema capabilities yes
Ability to specify maximum string length no yes
Opaque data types with bit-sized granularity yes no
Enumerated data types yes yes
Structured data types yes yes
Structured data types with optional fields based on simple boolean presence flag yes (using switch fields) yes (directly)
Structured data types with optional fields based on more complex conditions yes no
Sequences and one-dimensional arrays with given length yes yes
Sequences with specified termination pattern yes no
Multi-dimensional arrays need to be modeled as sequences of sequences directly supported
Supported in QuickOPC Version 2018.1 and later no Client-Server, binary encoding: Version 2020.2 and later