Henry B Hotz <hotz at jpl.nasa.gov> writes: > This avoids inventing a whole new subroutine just to allow you to return > a single string from another one. It sounds like a great idea to me. What am I missing? It means I can take the error string out of the API entirely. -- Russ Allbery (rra at stanford.edu) <http://www.eyrie.org/~eagle/>