I don't see why is this a good way to load assets, b/c for every asset you load you need to boot whole framework.

It allows you to run processors and minifies the assets and then will normally drop the files out into web-root (assuming you give it write permissions). This means that you don't need to boot up the framework to access any of the assets.

It also generates file names based on the contents, so as the assets change, so do the file names. This means that you won't need to add things like ?v=2.0 or whatever to your asset files locally in order to implement cache-busting.

It also caches local versions of your assets throughout the process, that way if you have 1 out of several files that are changed, it only actually has to touch and re-generate that one file instead of re-processing the files each read.

~awjudd