Tom Lane wrote:
> It's really all kind of messy ... we need to trade off simplicity of
> definition, ease of use, backwards compatibility, and standards
> compliance (though the standard has only 1-D arrays so it's of just
> limited help here).
AFAICT, the standard would certainly allow a type specification of
basetype ARRAY ARRAY ARRAY ...
which gives you a multidimensional array (in the same sense that C has
multidimensional arrays). If my reading of everything is right, the
subscript order that this would require is the opposite of what
PostgreSQL uses for multidimensional arrays. Talk about messy ...