Good explanation, but it sounds like there is a second component to the issue that this doesn't address: the fact that user input can be converted to symbols, which will ultimately eat memory as new symbols are created.