tsalo
commented
2 years ago I agree with your conclusions. I think these rules are going to be difficult to enforce, especially when we want to consider derivative datasets independently of the associated "raw" datasets, but it seems like that's just the way things are (until 2.0, potentially).
One other conclusion I took from that section is that metadata fields in derivative datasets can't restricted like they are for (some types of) raw datasets, except in cases where specific fields are explicitly forbidden. I can't remember which data types restricted metadata fields (MRI?), but validation will be affected for at leave some.
christinerogers
effigies
While attempting to add validation rules for BIDS Derivatives, we've found there is room for differences of interpretation in https://bids-specification.readthedocs.io/en/latest/05-derivatives/01-introduction.html (source included below)
https://github.com/bids-standard/bids-specification/blob/73d9968b55ce9b87371522d7efa8cda57978e705/src/05-derivatives/01-introduction.md#L14-L71
To operationalize these, I see the following conclusions:
1) If the filename is a valid raw filename, then that file is raw and all raw BIDS rules apply to that file. 2) If the file has a valid raw suffix (but the filename is a valid derivative filename, e.g., with the
desc-entity), then the REQUIRED metadata for that suffix becomes RECOMMENDED, and the RECOMMENDED metadata becomes OPTIONAL.Whether or not this is the consensus position, I think it is worth writing the rules explicitly.