Nowadays concurrency is everywhere: from big servers to small embedded devices. Most programs are now created with concurrency in mind.

But, although concurrency is as old as the discipline of Software Engineering, we still don't have a good grip on how to approach concurrency. We have deadlocks, livelocks, race conditions, and other safety problems; moreover, the performance of concurrent systems is sometimes worse than the performance of single-threaded programs. No wonder that "threading issues" often appears as the top annoyance for programmers.

This talk describes a change of perspective in approaching concurrency. Maybe, after more than 50 years of experience, we should abandon the model of using lock-based synchronization in favour of newer approaches. We will discuss a way of approaching concurrency without lock-based synchronization that promises safety and high performance.