There are differing schools of thought on this question.
For an XT8i, a common solution is to mount the fan using velcro pads directly to the back of the mirror cell (the metal back plate of the scope).
In my opinion, if the fan is behind the mirror it should exhaust air from the scope (pull it out of the scope past the mirror), to reduce tube currents. On the other hand, if the fan is in front of the mirror, it should be mounted at right angles to the plane of the mirror and move the air across the face of the mirror, to provide laminar flow, in which case it could either be an intake fan or an exhaust fan.
I have seen some very effective cooling fans that are used in pairs on opposite sides of the mirror: one exhaust/one intake.
However, the whole point with small scopes (mirrors below a certain mass) is to remove air trapped in front of the mirror. Cool-down times on small mirrors are generally acceptable, so it's the tube currents you want to deal with.
One way to tell if tube currents is the larger problem for your particular scope is to observe a bright star with the image defocused enough for you to spread the light across most of the field of view at high magnification. If you have tube currents, you will see what appear to be cloudy swirls in the image.
With larger mirrors (12 inches and up), there is sufficient thermal mass that the fans should be directed at the mirror from the cell side. Cool-down time is certainly an issue with higher masses. Even open-tube (truss type) large-aperture dobsonians often incorporate cooling fans behind the mirror cells for this purpose.
We call it cool-down, even though the proper term is equilibration: minimizing the time required for the mirror to reach thermal equalization (the air can be cooler, or warmer, than the mirror of course). Cool-down is much easier to say and type ...
The situation is different for different scope designs:
-
SCTs and Maksutovs trap air inside, so tube currents can be a worse problem -- particularly at high magnifications and large thermal differentials between mirror and air.
-
Refractors typically have a larger problem with OTA material equilibrium than with optical thermal equilibrium, especially since high-end refractors typically have temperature-compensating objective cells and/or focusers.
-
Open-tube designs don't suffer as much from tube currents.
-
Newtonian reflectors suffer from both tube currents and mirror instabilities.