this keeps PLIST_DB consistent, requested by sturm@ solution discussed with espie@
a bit complex, but doesn't work well with SUBST_VARS for now