It all hangs on the fact that a linear workflow produces the most believable lighting and there's no achieving this unless gamma is addressed. In a nutshell, gamma correction boosts the overall values of pixels to compensate for the fact that the voltage that monitors use to power said pixels just doesn't allow for the colors to look quite how we'd expect. This is why RAW images look dark and saturated and it's why common images like jpegs for example, have the compensation embedded in order to look normal. And it's why lighting without a linear workflow would involve a boat load of lights with all kinds of intensities, decay rates and shader tricks to balance out the illumination.

The color management's input color space allows you to counteract the embedded gamma correction (sRGB suits the typical 8 bit image and of course Maya's color picker now has color space options). The now proper colors are passed into the rendering space for processing (you can change this if the output is for projectors or to be comped with film footage etc) then there is the view transform which is the color space used for simply previewing output in a way that looks right to eyes of us mere mortals (both Maya and Arnold's render views allow you to save the image with the view transform baked in).

I think it causes confusion because of how messy linear workflow was in older versions and then the 2015 color management was half baked as it only corrected textures and not swatches so a bunch of gamma nodes were still necessary. And the only reason I don't use 8bit images for displacement and other data maps is because of banding and artifacts otherwise 8bit is fine (or else I'd have to ditch my Arroway CDs lol).