Any "raw" BIDS dataset is not given by god or a scanner, it is either result of manual labor, a script under code/ or use of helper tools such as heudiconv. The mere presence of sourcedata/ supports that claim. But nowhere in BIDS (dataset_description.json) we have formalized fields/form to inform provenance for the dataset. Currently it is only BEP003 (common derivatives) and future BEP on provenance #368 which touch upon that topic. As proper provenance tracking is a bigger job todo, we could at least implement in "raw" BIDS Pipelinedescription approach from BEP003.