My take is that unless you really do your homework, smoother is better. There has been evidence to suggest that some intentional roughness can offer reduced skin friction (try Googling: friction riblets), but in order to see the small benefit, the orientation and geometry of the roughness needs to be precise. My intuition is that for most Reynolds Numbers in play for RC, if your finger can't tell the difference in roughness, the air probably won't either (I suspect there will be disagreement on this). I agree with rmh that it's tough to isolate the effects of roughness for different construction techniques because other factors (like stiffness) will come into play.