Smolka made a presentation of the CLP programming language Oz and techniques for encoding unification grammars in it directly. Contrasting Oz with LIFE clarified the distinction between to different approaches to control localisation: controlled type unfolding effectively localises it within predicates, while Oz localises it in independent control constructs.
Podelski gave a talk entitled "An Order-Sorted Formal Theory of Unification", and a detailed technical discussion ensued about the various computational mechanisms and their mathematical models of sort hierachies and sort unfolding.
Feature structures may be modelled as either open or closed. Unification-based formalisms differ on this point, as do CLP systems; Oz, for instance, has closed FFS, while LIFE has open. Closed feature structures may be slightly more efficient (though only by a constant factor, since open FFS can be implemented with extensible hash tables, as in LIFE), but there is some question as to how efficient it is to simulate open FFS with them (in the absence of global analysis) if that is what the formalism being implemented requires. On the other hand, functional adjunction (which is perhaps independently motivated in semantics) is available under the closed interpretation, and it is less clear what that would mean with open feature structures.