"ALL SEMANTICS OF A DATASET NEED TO BE CONTAINED WITHIN THE DATASET"
That means that no significant logic may reside in the filename, or
placement of data in the file, or in the program that has generated
The reason for that is that different programs may interpret the data
differently, and unless clear instructions are contained within the
file itself, problems will arise (or we'll need a tome of extra
We will have to agree on a minimal ontology and vocabulary of file
format, but the rest needs to be fully expressable inside the file.