If this change is accepted, then also Part 5 should change: In ModelChangeStructureDataType Structure, field 'verb' (Table 156 in 1.04), there is currently Bit 4 for DataTypeChanged: "This verb may be used only for affected Nodes that
are Variables or VariableTypes. It indicates that the DataType Attribute has changed.". We should add bit 5 for ValueRankChanged.

I agree with the logic that ValueRank should be considered in the same way DataType is.

I agree - clients should be notified when any aspect of a Value changes - these can (and do) impact client behavior.

A change in any attribute that changes the "shape" of the data should trigger this event. This includes ValueRank, ArrayDimensions, and DataType; perhaps others, such as access rights, whether or not writes w/timestamps are supported, etc. Further, when any of these attributes change, all related attributes should be updated to the extent their new value can be ascertained.

For example,

• if the ValueRank changes to indicate the value is an array, the ArrayDimensions attribute should be updated to reflect the maximum length for all dimensions which are known.
• if the UA server "knows" that the max array length has changed, it should update the value in the ArrayDimensions and trigger this event.

This feature should be supported by a CU or facet that verifies this,

We agreed in the June 15 2021 weekly call that a change to ValueRank and ArrayDimensions should also be considered in the same way as a DataType change.

Need an errata for 1.04.

Added ValueRank and ArrayDimensions in addition to DataType in Table 13 'Variable NodeClass' and Table 14 'VariableType NodeClass'

Added Errata 1.04.11

Need to look at the bigger picture before committing to this solution.

I reviewed the recoding from 2021-10-05 to understand why this change was reverted. There is a lot of disagreement about what does today, or should in the future trigger a ModelChangeEvent, a StructureChanged bit or a SemanticsChanged bit. On one hand, some changes seem to be missing that should be included, on the other hand things are included that are technically unnecessary as the change is always evident in every value change received (e.g. the dimensions of the array is always included with the value).

For the particular case of ValueRank and ArrayDimensions -- see that no-fix was agreed to for Mantis 0007344

I added this to the Genoa F2F.