We show that the extrapolation to the case of global fields of characteristic p of a question posed by Stark in 1980, regarding abelian L-functions of order of vanishing 2 at s = 0, has a negative answer. We provide links between various versions of Stark's question and a natural refinement of Brumer's conjecture, in the general context of global fields of arbitrary characteristic. As a consequence, we show that the refinement of Brumer's conjecture is, in general, false for characteristic p global fields.
