I often hear: "Sharpening really is the last thing to be done to any image, whether printing or uploading."
I lot of people still stick to this, but unfortunately it's far from "best practice" these days. Optimal sharpening usually requires at least 3 passes, at different parts of the workflow.
Capture sharpening is always done first to counter the effects of the camera's anti-aliasing filter and softness introduced during demosaicing, making the image much nicer to work on at high magnifications (Canon recommend 250 - 300%, 0.3 Radius, and varying thresholds depending on the ISO (thus noise) of the image (I normally use zero).
Content & creative sharpening are done mid-workflow - typically much lower amounts (40 - 120%, 2 to 5 pixel radius) to make things "pop".
Output sharpening depends on how big you're printing or displaying, thus it's pretty much the last thing done, but is best done on a copy of the image since it will need to be done if you're subsequently printing different sizes or displaying on a different medium.
The "industry standard" text on the topic was written by Bruce Fraser shortly before his death - "Real World Image Sharpening" http://www.amazon.com/World-Image-Sh...2402601&sr=8-2
Cheers,
Colin - pbase.com/cjsouthern