On Zero Duality Gap and the Farkas Lemma for Conic Programming

Recently S.A. Clark published an interesting duality result in linear conic programming dealing with a convex cone that is not closed in which the usual (algebraic) dual problem is replaced by a topological dual with the aim to have zero duality gap under certain usual hypotheses met in mathematical finance. We present some examples to show that an extra condition is needed for having the conclusion; this supplementary condition is also provided. We also give counterexamples for three results on hedging prices and simple proofs for two known solvability results (see Propositions 4.1 and 4.2).