(Agreed with Matthias Damm that this is genuine and should be made into a Mantis issue for him to resolve. May invite Etienne Rossignon/Sterfive SAS as he was the one who triggered some of the questions here).

As a note, DataSetMetaDataType is used:

• in the PublishedDataSetDataType
• in DataSetReader parameters (DataSetReaderDataType, DataSetReaderType)
• in StandaloneSubscribedDataSetDataType
• as DataSetClasses field in PubSubConfiguration2DataType
• in UADP DataSetMetaData announcement message
• in JSON DataSetMetaData discovery message
• in DataSetClasses Folder of PublishSubscribeType
• in PublishedDataSetType
• in AddPublishedDataItemsTemplate and AddPublishedEventsTemplate methods

It derives from DataTypeSchemaHeader (part 5), where for its fields like structureDataTypes, enumDataTypes, simpleDataTypes, it uses formulation like "Description of Structure and Union DataTypes used in the data that follows the DataTypeSchemaHeader. This includes nested Structures".

An interoperability issues arose around this. In my case they had to do with DataTypeSchemaHeader which was in the UABinaryFileDataType from the publisher. Question are e.g.:

1) When the spec says that structureDataTypes etc. should contain "Description of Structure and Union DataTypes used in the data that follows the DataTypeSchemaHeader", does it have to include the description for datatypes that are in defined in the spec (namespace 0)? [most likely not]

2) Take this example: The publisher will publish messages which contain a custom structure. This structure, however, is not quite "used in the data that follows DataTypeSchemaHeader", i.e. in the metadata. The metadata only contains the node Id of the datatype; the datatype is only really used "on the wire", in the PubSub messages. The spec should make it clear what constitutes a "use" (Part 5 wording), because in this case, the intent of the spec was that description of such structure type shall be included in the DataTypeSchemaHeader.

3) An IOP WS participant argued that his publisher might not know the datatypes upfront (when generating this structure), OR, in polymorphic case, that it might know the base type but not all the derived types. He had not put them into the DataTypeSchemaHeader, but my subscriber, obviously was then not able to consume the PubSub messages.

4) Need to touch on the fact that in some forms of published messages, a generic Variant may be used, and it contain an extension object with any data type in it. Clarify whether such datatypes go into the DataTypeSchemaHeader.

5) If the datatype descriptions can be put into multiple places ("levels"), such as in the UABinaryFileDataTupe of the PubSub configuration, and in the dataset metadata, where should the generating code place them?

6) If the datatype descriptions are found in multiple places ("levels"), such as in the UABinaryFileDataTupe of the PubSub configuration, and in the dataset metadata, how should the consuming code interpret or combine them?

7) The answers about what goes in and what not can depend on where the DataTypeSchemaHeader is used, see the list of usages at the beginning.