Not sure this suggestion is exactly in the spirit of Postel's law. Citing my own law, I will say that I much prefer the return value for "This is invalid/malformed JSON" to be distinct from "The path you specified does not exist in this valid well-formed JSON". I won't advocate for it in this case because there are other functions that can check the JSON validity, including in CHECK constraints, and in my projects (as I would advise to all) the SCHEMA would be tasked with ensuring JSON validity prior to trying any function upon said JSON values. This means that by the time a value is in the DB, it is already confirmed to be valid, and so any null returns can only mean that either the specified path is wrong, or it doesn't exist in the JSON. Ditto when acting upon the output of a prior JSON function. I think that is just fine in this case.