The question of what the “right” models of logic programs are has been discussed intensively. Following Clark’s completion semantics in 1978, various other semantics have been suggested, along with proposals for syntactic criteria for deciding whether a program has an acceptable model. In this paper a class of programs called “effectively stratifiable” programs is defined, and for each program in this class a unique model (its “default model”) is constructed. Some programs outside the class also have unique default models.

The paper is written clearly, and it gives a thorough comparison of stratifiable and locally stratifiable programs, and perfect, weakly perfect, stable, and well-founded models. The authors show that default models are equivalent to  stable  models (and note that default model semantics were defined prior to stable model semantics). On the other hand, neither weakly perfect model semantics nor default model semantics is more general than the other.

The claim that default semantics is more right or natural than other semantics is doubtful. The intended model is supplied by the programmer, not derived from the semantics. The paper hardly mentions any practical proof procedure that might be sound or complete with respect to default models. Without this, a programmer cannot reasonably ask whether a program has the behavior she or he intended.