On disjoint access parallelism Alessia Milani, U Bordeaux Abstract It is desirable that implementations resulting from a universal construction or a transactional memory allow operations operating on disjoint parts of the data structure to progress concurrently. Disjoint-access-parallelism formalizes this by requiring that different operations cannot both access the same base object unless they operate on some common part of the data structure. This talk will provide a taxonomy of definitions for disjoint-access-parallelism including those in the literature. We show that all these definitions are provably incompatible with wait-freedom. Finally, we propose a new definition to overcome this difficulty.