There are different “open” issues with the “event” topic.
The definition in the UADP message type is probably the wrong place and too restricting.

The Variant representation is mainly needed for PublishedDataSet that are created based on real OPC UA events.
But the event messages can also be produced for other event like sources e.g. a list of variable values sent based on a trigger. We recently added a clarification that this depends not only on the type of PublishedDataSet but mainly on the KeyFrameCount == 0.

Input from Zbynek:

Table 24 – DataSetFieldContentMask Values
I think that the following is still not quite clear:
If one of the bits 0 to 4 is set, the fields are represented as DataValue.

The sentence should probably read “If the RawData flag is not set and one of the bits 0 to 4 is set, the fields are represented as DataValue.”, otherwise it is in conflict with the sentence that precedes it.

But my main concern is again with the Event messages. I understand that the event fields needs to be encoded as Variant. The spec should say, at this place, what happens if one of bits 0-4 is set but bit 5 is not set and we are dealing with an Event message. Which rule wins? Should the event fields then be encoded as DataValue, or as Variant? If as Variant, it really has to state it at this place, because currently the only interpretation it allows is that it will be a DataValue.

Should be discussed in F2F

Table 164 – JSON DataSetMessage definition
Payload
Added:
For MessageType “ua-event”, only Variant or RawData encoding shall be allowed. If bits for DataValue encoding are set, the Variant encoding shall be used.

Aggreed on change in meeting, needs errata for 1.04

Agreed to 1.04 Errata.