My understanding was that the MEL script was a Maya specific automation script and has significance only to Maya. I believe that Maya can directly output Renderman files tho which I believe are entirely different and are the actual scene and shader descriptions. The PRMan (actual renderman app) takes the scene description files and then can make the visual image from them. I dont think there is any possibility of sending MEL scripts directly to the renderer.

So in a far removed sense, POVRay is like PRMan. My scant knowledge of Mental Rays is that its similar and has a data base where you can feed it similar files to set up the database. The files have an extension of .mi and are mental rays scene description language files. How close they are to renderman I dont know, but the same kind of thing happens where the app takes the database and renders the scene from it. Whether in PRMan you have the whole database as a set of renderman files or it too has a distinct database loaded from the files I have no idea.

So what Crayons is doing is like programming the scene directly in the scene description language and sending it to the renderer. It can be done but I think most people would use a GUI and let a 3D app do the hard work of all those numbers and treat the POVRay file as a mere intermediary. Of course knowing the internals of whats going on is a definite plus, and if you are going to script up an automated tool like the city builder then this is the right way to go. The script (Perl, Python, Java, etc) that is used to automate building the POVRay file is more like the MEL bit of Maya. Instead of using a Perl interpreter, Python interpreter or Java machine, you have Maya which also happens to be a script interpreter that has oodles of 3D capability and naturally much more capable. But the large increase in capabilities are the only difference between this and Hollywood - the fundamental process is the same.