-- LOD proposal --

	An LOD system would improve render performance and greatly
	increase the perceived draw distance, one of the worst issues in Minetest.

	* Absolutely worth it
		- Bump the draw distance from ~150ish to 2k and beyond
		- Minimal overhead (see below) and much, much bigger savings
		- Not as difficult as it sounds

	* Approximate mapblocks before fully generating them.

	* Store and send the "mipmaps" of a mapblock.
		- The costs increase only by about 0.142; the "mipmap principle" applies even harder in 3D

	* As the distance increases, information can be skipped.
		- Special drawtypes can be omitted
		- Textures can be approximated with color

	* Extremely far info need not be accurate
		- Can probably get away with just sampling mapgen data
		- At the same time, updating a 2x2 heightmap is dirt cheap

	* Approximate certain features in special ways
		- Most maps have a terrain with an ocean, handle these specially
		- Known light sources can be sprites

	* Player structures and decos can be approximated with imposter sprites
		- 3 flat cross sections will always be cheaper than the voxel equivalent

	* Retroactively regenerate the mips for blocks altered throughout the game
		- Lazy, low priority, separate thread, etc.

	* Cache and generate some of the data client side
		- Don't send me LODs of a mapblock I've just been to, I've already generated it locally
		- Hashes and timestamps to verify if the client already has the block

	* Play nice with Lua, and minimize the burden on mods
		- Let modders specify if a node is used for terrain, or if it's manmade, and if it should be a light source
		- ...but attempt to guess it in case it's not known! ( from biome and deco defs )
		- Also let node-modifying code specify if a change is important
			-- You can guess some of this, actually; VoxelManip changes and player digs are usually important, other stuff less so