_id == md5 hash of content?

[phiweger]

no:

• implementation tricky (Merkle tree?, how to sort? see deepdiff library for starters)
• every time a change is made to a document the _id would have to change - this seems like it creates a lot of bugs downstream (e.g. with two people collaborating on a dataset)

yes:

• easy deduplication

[phiweger]

no, keep separate