WG211/M13Scholz

Tinkering is Dead --- Long Live Tinkering

Code bases that evolve over the years often suffer from a tinkering effect: The increasing complexity of code leads to modifications that are made without complete awareness of the context of the modified bits. Switching between versions is tedious and typically requires going the full cycle from checking out an older version of the system, compiling it and then deploying the binary.

In this talk we suggest a more dynamic approach towards tinkering: we allow the programmer to keep several versions of parts of his code and provide him with a tool to dynamically choose which one he wants to explore. We show how this can be beneficial when targeting a range of different hardware platforms.