A middle way for robotics middleware

Robotics is changing. The amount of software available (and needed) is growing. For better or worse, the glue that holds that software together, the middleware, has a big impact on its viability. YARP is a middleware for robotics, with over a decade’s continuous use on various humanoid robots. YARP was designed to help code survive changes, to easily experiment with new code and integrate with other systems. In a world of constant transition, with a steady stream of hardware and software upgrades, YARP helps code last long enough to make a real impact, and avoid premature loss of good code through middleware muddles. We review the features of YARP that support this flexibility, describing those situations in which they have been practically useful. We argue that there are practices that any middleware author can adopt that benefit users of other middleware and raise costs to those users when neglected. Altruism is not required: we also argue a middleware’s own users will benefit from such practices right now (when collaborating) and in the future (when upgrading to a new version of the middleware). These concerns are not usually the first things on a user’s mind when choosing a middleware, but a responsible middleware author will foresee them and prepare a happy ending rather than a trap.