What I normally do is this:
1. Have at least three layers in the following order: scanned-in lineart as the top layer, followed by a transparent layer, and finally a white background layer.
2. Add an alpha channel to the lineart layer and set the Layer Mode to Multiply.
3. Then switch to the transparent layer and paint the colors, trying to keep everything within the top layer's lines (the lineart layer should still be visible for this).
4. Shading is done by selecting the colors using the Select By Color tool and then airbrushing the color black (set at 40% opacity) with a Fuzzy Circle brush. Highlighting is done with the color white (again 40% opacity). During this stage the transparent layer's alpha channel should be locked so that uncolored areas don't get shaded.
5. Optionally I can select the background layer and fill it in with either one color or (for a really nice effect) a gradient.
This is what I've produced with my technique so far:
I realize there are problems with the anatomy, positions of the shading and highlights, and other fundamental aspects of the drawings besides the coloring, but let's not get into that here. What I want to ask is whether the technique I've learned is an ideal or at least acceptable coloring methodology. Have I been led astray all this time?