Keeping the backends themselves outside of the Django project is definitely the right move. This allows each of them to evolve at their own (probably faster) pa

Keeping the backends themselves outside of the Django project is definitely the right move. This allows each of them to evolve at their own (probably faster) pace independently of the Django project which keeps a strict schedule of backward compatibility. A similar approach is used by Andrew Godwin to keep the South DB migrations project moving forward quickly. It could be merged into Django itself, but it is better outside.