> returns a null pointer. perhaps this should be trapped by the routine > and a more meaningful exception raised in its place outside the > routine? Right, this change has been suggested on this list some time ago and made in CVS. Arno