Could this be too general a problem to "solve" without narrowing down major uses-cases and industries, and likely first-adopters? Will there need to be differen

Could this be too general a problem to "solve" without narrowing down major uses-cases and industries, and likely first-adopters? Will there need to be different evolutionary/adoption paths? Will Postal's rule need to play a role? A priori, the design space I'd want to look at is something between protobuf and csv. Perhaps optionally at the end of each line [somewhere non-breaking] you add metadata for that line, including specifying a parser implementation you know can perform perfectly. The OP suggestion might address this if metadata(s) could apply to particular ranges of lines. It would be more fragile to the extent of being in different file.