As has long been recognized, the isomorphism between word edges and word-internal syllable edges is far from perfect. This paper examines the fit between an Optimality-Theoretic account of edge/interior asymmetries, using position-specific faithfulness constraints to protect edges or interiors of morphological constituents, and the actual typology of attested edge/interior asymmetries. A detailed analysis of the Indonesian language Balantak, in which the first member of a CC cluster is severely restricted unless that first C is root-final, is compatible with the positional faithfulness account, but is problematic for accounts that explain greater freedom at edges solely in terms of licensing by higher prosodic structure or by phonetic context. I argue that a theory of edge/interior asymmetries must incorporate two possible functions of phonotactic restrictions: to facilitate recovery of segmental contrasts, and to facilitate the parsing of strings into morphemes.