Distinguish dimensions from attributes
Following a mail exchange with a DBnomics user:
You mentioned the provider ILO but not the dataset code: let's take for example "EMP_2EMP_SEX_AGE_NB" (https://db.nomics.world/ILO/EMP_2EMP_SEX_AGE_NB).
Indeed it seems that the dimensions "ref_area" and "source" are fully correlated.
Until now we consume ILO data by downloading their CSV files (cf https://www.ilo.org/ilostat-files/WEB_bulk_download/html/bulk_indicator.html), which do not distinguish the dimensions (constituting the series code) from the attributes (purely informative, don't appear in the series code).
It seems clearly that the "source" should not be a dimension but an attribute.
After looking twice at ILO website I just noticed that they offer a SDMX API here: https://ilostat.ilo.org/resources/sdmx-tools/
When requesting the same dataset, it appears that SDMX data distinguishes well the dimensions from the attributes:
<generic:Series>
<generic:SeriesKey>
<generic:Value id="REF_AREA" value="AFG"/>
<generic:Value id="FREQ" value="A"/>
<generic:Value id="MEASURE" value="EMP_2EMP_NB"/>
<generic:Value id="SEX" value="SEX_T"/>
<generic:Value id="AGE" value="AGE_YTHADULT_YGE15"/>
</generic:SeriesKey>
<generic:Obs>
<generic:ObsDimension id="TIME_PERIOD" value="2013"/>
<generic:ObsValue value="7255.05"/>
<generic:Attributes>
<generic:Value id="DECIMALS" value="1"/>
<generic:Value id="UNIT_MEASURE" value="PS"/>
<generic:Value id="UNIT_MULT" value="3"/>
<generic:Value id="UNIT_MEASURE_TYPE" value="NB"/>
<generic:Value id="SOURCE" value="ILO - Modelled Estimates"/>
</generic:Attributes>
</generic:Obs>
</generic:Series>
So as long as we use CSV files, this will stay as-is. But thanks to your question, maybe we'll switch to SDMX; but in general consuming data over HTTP APIs is way slower than downloading bulk files.