NWB and BIDS

[vigji]

Moving a conversation started here.

I have seen how DANDI distinguishes either a full-BIDS approach for neuroimaging data and a BIDS-like folder structure + nwb files for animal data.
I have also seen that BIDS-animal-ephys supports nwb data; but as @bendichter remarked in the conversation, there is a tension there in terms of duplicated metadata and what should go in the NWB file (potentially everything; or just the ephys, as far as I understand from the animal-ephys specifications). I would like to now thoughts - approaches that people are taking (do not know if someone from animal-ephys BIDS is here to discuss). Is there any consensus emerging around how to combine the two? What are the irreducible inconsistencies that prevent the mapping/automatic conversion of a nwb file into a BIDS-structured folder? I am still getting to learn details of both approaches, I would love to hear any thought about this.