In order to convey the information that one dataset has used the content of another progenitor data set , we need a link to identify the relationship like the '#derived' term in the datalink vocabulary .
This is a way to trace a kind of provenance for the datasets .
If I remember correctly, in one early implementation of ObsCore at CADC , the ivoa.obscore table was joined to another obscore.part table to exactly represent datasets as single images and as the result of coadded images as well, each being identified in the obscore table.
Pat Dowler may say more details about this way of exposing data sets and their progenitors.