Attribute relationship vs attachment level

[hoehrmann]

https://github.com/sdmx-twg/sdmx-json/blob/71fe5eaa9fcd29e3c15f2f0216a19b9b650b1dbd/data-message/docs/1-sdmx-json-field-guide.md#component says »Depending on the message context (especially the data query) an attribute value can however be attached physically in the message at a different level.«

But https://github.com/sdmx-twg/sdmx-json/blob/71fe5eaa9fcd29e3c15f2f0216a19b9b650b1dbd/structure-message/docs/1-sdmx-json-field-guide.md#attributeRelationship says »An attribute with this relationship will its values always have presented at observation level.«

This seems contradictory, and the latter statement seems mistaken. Suppose the attribute is only there to be able to encode extremely unlikely cases should they ever occur. If there is no such case in the data, why shouldn't the value be attached to the data set? It also seems an unnecessary burden to generate an error or ignore the attribute if specified on the dataset level just because the DSD says different values per observation are possible.

(In doubt, please handle this as a public review comment on SDMX 3.1 once the comment period begins.)