Well, when you got 4 megs of RAM you gotta cut corners. If the "chunks of data" will be small, say, 1(horizontal row)x10(columns)x10(layers) positions like I suggested, the "loading time" would be insignificant and rather "quick".
But at the end of the day, it's Smealum's engine. He knows best what to do with the source.
It's not like it's loading "graphics", it only loads the dimensions in which the engine has to spawn pre-prepared squares. Cloning them will be much faster than rendering them in real-time.
EDIT: I just found a flaw in my idea. The binary would have to be stored in FAT, Nitro doesn't allow writing, lol. Silly me.
But at the end of the day, it's Smealum's engine. He knows best what to do with the source.
It's not like it's loading "graphics", it only loads the dimensions in which the engine has to spawn pre-prepared squares. Cloning them will be much faster than rendering them in real-time.
EDIT: I just found a flaw in my idea. The binary would have to be stored in FAT, Nitro doesn't allow writing, lol. Silly me.