Multiple DSDs?

[albertmeronyo]

Reading the QB specification I thought of a rather ambiguous point: are all components described in the DSD supposed to apply to all the observations of the dataset?
I think this is an important question, since, if true, we would need to essentially generate different DSDs, one for each of all our current slices (since the components might change on a census type + year basis).

[cgueret]

The DSD apply to all the observations attached to it. So if we have to change de components we would indeed need different DSD, but I don't think its the case.