Mantis issue 3750 (https://opcfoundation-onlineapplications.org/mantis/view.php?id=3750). The decicion to replace opc:String with opc:CharArray's definition in the typedictionary system does break all applications that are made with the 1.03 original (or before) spec.

Generally I do not see why types that do not have a NodeId were allowed in the first place (opc:String, opc:WideChar, opc:WideCharArray, these do not have any OPC UA Binary encoding nor DataType NodeIds and therefore do not exist) and having encoding for opc:CharArray that is the NodeId DataType String's equivalent, but anyway replacing a encoding with another as "Errata" is not a good solution.

This now means that if I encounter a dictionary that defines opc:String it is impossible to know which encoding it means. Because if it is made before this errata, then it should be a null terminated String, after errata it should be instead a prefixed one. It is not possible to determine from the Server which one it is.

Generally I would advice to fix the code generators that do not create correct code according to the specification and not to change the specification.