Python concurrency research asks tasks to declare the data they need

A Python summit proposal explores scheduling tasks around the data they need, instead of asking programmers to manage threads and locks directly. Behaviour-oriented concurrency associates protected data with owners and declares dependencies before tasks run. The bocpy prototype demonstrates the interface with subinterpreters, but enforcing the full ownership model would require changes inside Python’s runtime.

Python Insider · Sep 30, 2026

Why it matters

Python developers can examine safer concurrency interfaces beyond simply removing the interpreter lock.

More context

Try the public bocpy examples as a proof of concept. The current package implements isolation; free-threading support is planned. The report discusses alternatives rather than announcing a standard-library choice.

Keep in mind

The prototype lacks full ownership enforcement, and communication-heavy tasks pay subinterpreter costs. The proposed safety properties must not be confused with guarantees of the current package; 1min has not tested it.

Original read