Currently by looking at the spec it is not obvious how the output of a encoded JSON NetworkMessae/DataSetMessage should look like. There is risk many 'flavors' are invented to what should be a standard.

i) Specifically it is unclear how the 'FieldName' should be populated.
ii) Second the DataSetFieldContentMask is described in the following way:
• If the DataSetFieldContentMask results in a RawData representation, the field value is a Variant encoded using the non-reversible OPC UA JSON Data Encoding defined in OPC 10000-6.
• If the DataSetFieldContentMask results in a DataValue representation, the field value is a DataValue encoded using the non-reversible OPC UA JSON Data Encoding defined in OPC 10000-6.
• If the DataSetFieldContentMask results in a Variant representation, the field value is encoded as a Variant encoded using the reversible OPC UA JSON Data Encoding defined in OPC 10000-6.

Now adding to the confusing is some personal communication between Erich and Randy which revealed:
i) The Field Name should be the BrowseName. (This is a Qualified Name so it is unclear how to encode it as Key?)
ii) A DataValue should be encoded as reversible JSON. (This is exactly opposite to whats written in the spec above where a DataValue representation is encoded as a non-reversible)

Now looking at the MQTT prototype in the UA.-NET Standard repo there is also the DataValue encoded as reversible JSON.

Question: Is the spec correct ? Is the sample correct? Is the personal communication correct?

In fact, imho it would be preferred to follow the spec and publish the DataValue as non-reversible to reduce redundant information to be sent which is already in a meta data frame.
For the field name, it should be configurable to be very small, e.g. for each published node the key could be an alias to reduce data (NodeId 1 --> FieldName 'a', etc.)

Preferable would be a clarification or a sample in the spec.