Yoann Rodière, ah yes, that's a good observation, constraint definition contributors already provide that capability. Indeed I'm wondering though why we made that a separate "top-level construct" rather than going through ConstraintMapping as you suggest. I'd like that better, too, as it resembles the XML structure. Hardy Ferentschik, do you remember why we did it that way? It's marked as experimental, so I'm inclined to alter it as described above. WDYT?