In fact, I choose to specify them in the file(s), but is there a reason for prefering one method to the other?

No, either is fine. I tend to prefer keeping the columns in a file because *) other programs like Lotus or MS-Access that read CSV files can use the field names *) if you forget what the names in the file are, you can either use SQL to query for them or just do a 'head' on the file.