Upper-cased attribute names are reserved by DBI and should be passed by the driver to DBI for storing and fetching.

Lower-cased attribute names should start with *a* recognized prefix, not necessarily the prefix of the driver in question. Lower-cased attribute names that are not a recognized prefix (e.g. mock_ or mysql_ or csv_ are recognized) should never be used.

And I agree with your previous post that DBD::Mock is a special case so you may want to allow unknown attribute names (though still they aboslutely should have a known prefix).