Originally, we started with attribute at the level of observation. These attributes are used to provide information about the quality or the way the value for a particular observation has been computed
SDMX uses also an attribute at the level of a dataset and IMHO is equivalent to a dimension that would be identical for all series in the dataset. It can be units, valuation, etc ...
We have also considered attribute to the level of the seriesm but I don't see the difference with a dimension
It would be useful to introduce a dataset attributes as a new field at the dataset level.
- It would be a list of pair of pair: [((name code, attribute name), (value code, attribute value)), ...]
- It would store any dimension that takes only one value in the entire dataset, a characteristic that is common to the dataset and not present in the dataset name, an SDMX dataset attribute
- Dataset attributes would be displayed at the dataset level in the UI
- Dataset attributes would be retured with other dataset metadata by the API (even if a single series is requested)
An important advantage would be to remove non informational dimension values when we build a series name from its dimensions. It would also suppress facets in the UI with only one possible value