Information modelling

What you will be able to do

• You can design a type hierarchy instead of a folder of tags.
• You can author and compile a NodeSet2 file of your own.
• You can expose a custom structure that a generic client can decode.
• You can export a live server's address space and work against it offline.

Syllabus

1. 1
   Model types, not instances

   Why an ObjectType with three instances beats three folders of tags, and what a client can do with the type that it cannot do with the folder.

   Asset The same three pumps modelled both ways, browsed side by side

   8 min
2. 2
   Apply modelling rules

   Mandatory, Optional and the placeholder rules. What a client may assume, and the failure when it assumes an optional node exists.

   Asset A type with mandatory and optional members, and two instances that differ

   8 min
3. 3
   Use interfaces

   Multiple inheritance for capabilities. How the Machinery building blocks use interfaces, and why that is what makes them composable.

   Asset A node implementing two interfaces, with both sets of members browsed

   8 min
4. 4
   Define a custom structure

   DataTypeDefinition, field order, and the encoding a client needs to decode it.

   Asset A custom struct exposed by a server and decoded by a generic client, with the raw ExtensionObject

   9 min
5. 5
   Survive the legacy type dictionary

   Servers that offer only a binary type dictionary, clients that support only DataTypeDefinition, and the undecoded ExtensionObject that results.

   Asset The same struct read by two clients, one of which fails to decode it

   8 min
6. 6
   Handle unions and optional fields

   Where custom-struct decoding actually breaks, and how to model so it does not.

   Asset A union and an optional-field struct with the wire encoding of each written out

   7 min
7. 7
   Author a NodeSet2 file

   The XML that every companion specification ships as. Structure, namespaces, aliases and references.

   Asset A complete, valid NodeSet2 file for a small machine type, from scratch

   10 min
8. 8
   Compile a model

   UA-ModelCompiler as a dotnet global tool, its inputs and outputs, and the errors it gives you.

   Asset A model compiled with UA-ModelCompiler, including one deliberate error and its message

   8 min
9. 9
   Load a model at runtime

   Importing a nodeset into a running server, and what happens to namespace indexes when you do.

   Asset A nodeset imported at runtime, with NamespaceArray before and after

   8 min
10. 10
    Export a live address space

    NodeSet export from a running server so you can work offline and diff between firmware versions.

    Asset Two exports of the same machine after a firmware change, with the diff

    7 min
11. 11
    Publish to the UA Cloud Library

    What the library is for, its REST and Swagger surface, and when to publish rather than keep local.

    Asset A model uploaded and retrieved through the UA Cloud Library API

    7 min
12. 12
    Model for the consumer, not the PLC

    The habits that make a model useful downstream: stable identifiers, real units, honest ranges, and states as state machines rather than integers.

    Asset A review checklist applied to a real vendor model, with the eight problems it finds

    8 min

Lab

Last verified 7 September 2026 against OPC UA 1.05.07. OPC UA Connect is an educational resource and is not affiliated with the OPC Foundation.