The Problem with Prolog
                          by
                    David S. Warren
              Computer Science Department
               University at Stony Brook

Prolog and Logic Programming are based on SLD resolution, a
computational method that allows a set of logical formulas
(specifically Horn clauses) to be understood as a program in a
nondeterministic procedural programming language.  Thus a logic
program is both a set of logical formulas and a description of an
algorithm; it can be understood declaratively as specifying a logical
model or procedurally as a specifying a computation.  And the
foundational theorem of Logic Programming is that since SLD is a
complete proof strategy for Horn clauses, these ways of thinking are
equivalent.

But I claim that there is a serious fly in this ointment.  The
experience of Prolog programming, after the first two baby programs,
is not of declarative logical specification.  And Prolog has not
caught on the way one would expect of such an elegant and potentially
useful paradigm.  In this talk I will describe the current foundations
of Prolog in simple terms, analyze what the real problem is, and
propose a solution.