When investigating Jython performance Steve made a simple change of the hash map used in String interning from a WeakHashMap to a VM_HashMap. This gave mixed benchmark improvements from a 1% slowdown (noise?) to 3 and 4% speedups. http://www.nabble.com/Re%3A-Jython-performance-p12219797.html Given this is a simple and isolated change we should investigate switching how interning is implemented as the potential 4% win on a few benchmarks is worth having.