Regarding padding of String (and ByteString) specification is totally unclear.

Excerpts from specification:
1.
Table 7 – FieldMetaData structure
maxStringLength UInt32
If the dataType field is a String, LocalizedText or ByteString then this field
specifies the maximum supported length. If the maximum is unknown the value
shall be 0.
If the dataType field is not a String, LocalizedText or ByteString the value shall
be 0.
If the valueRank is greater than 0 this field applies to each element of the
array.

2.
7.2.2.5.9 RawData field encoding
The encoding of the DataSetMessage fields is handled like a Structure DataType where the
DataSet fields are handled like Structure fields and fields with Structure DataType are handled
like nested structures.
...
If the DataType of a DataSet field or a Structure field is String or ByteString and the
actual size is smaller than the maximum possible size indicated by the dimensions, the
field shall be padded with bytes with value zero.

Issue A)
It is allowed to specify maxStringLength for LocalizedText.
As far as I understand this member is used for RawData encoding only.
But LocalizedText is not allowed for RawData encoding.
Or is there any other place (e.g. Variant encoding) where the length of a localize text shall be validated too?
And if yes, what reaction is defined if length exceeds maxStringLength?

Issue B)
I just refer to String instead of "String and ByteString" at all the places.

According to 1. maxStringLength shall not be specified for structures.
According to 2. Strings shall be padded within structures "indicated by the dimension"

Nowhere is mentioned what that "indicated by dimensions" means.
Since maxStringLength cannot be specified, it could be the maxStringLength from "StructureField Structure" in part 3.
If it is so, please state is explicitely.
By the way - in part 3 nowhere is mentioned how encoding shall be affected if the string (ByteString or LocalizedText) exceeds this value...
For PubSub I assume that the DataSetPayload shall be marked as "Invalid".

De facto "encoding ... is handled like a Structure DataType" is not correct - a different behavior (algorithm) has to be implemented.
Since encoding in PubSub shall use padding of String values encoding of Structures shall not.
=> "encoding ... is handled like a Structure DataType modulo padding of String values (and arrays and matrices)" would be more correct.

I did not mention here arrayDimensions and row-wise padding in matrices (within structure fields); but the same issues are there as well.