Two things come to mind: - Maximum stack depth greatly varies between targets. If you wanted to support a recursion depth of, say, 50, that may already be too much for some of the hardware this library needs to be able to run on, so the original problem still exists. - This would be add an arbitrary limit to the parsing capabilities of the library. Stack space is very limited to begin with compared to heap space, So any solution that remains recursive would probably require a limit way lower than a heap-based solution could offer.