I've always considered single numerical data type as big benefit of Lua.
It does sound like, whatever Roberto's been playing with, it's intended to stay pretty close to the single type ideal. The idea is that the float/integer distinction would only matter in the one place where it really has to matter -- handling precision overflow.
Even so, I, too, am a bit suspicious of an integer "subtype". I'm quite happy with Lua's numbers as they are. I worry that integer multiplication might slow down some of my code, and there'd be a risk of unexpected precision overflows. Still, the _ENV semantics introduced in 5.2 were brilliant, so I'm trying to keep an open mind :)
-Sven
