To me the issue isn't how we should read in CSV (as you stated, we should use a different format that is less ambiguous) but rather getting vendors to provide the data in that new format. I've found that most vendors I work with are using a cobbled together solution built on top of a mainframe and getting them to add/remove a field from CSV is a monumental task, there is no way they will move to a different format like XLSX or JSON or XML. Until there is an open, industry-wide standard, I just don't see CSV going away