I've been playing around a bit with bagginbill's dome and using like 8000 pixel photographs as the image map for backgrounds. This is nice, but is there a way to control the exact location where my characters will appear in the photograph? The only thing i can do is increase or decrease the focal length of my main camera, but that generally only moves them a little bit forward or backward in the same spot. Which means i can only use one location in that photograph.

Is this a limitation of using that method, or is there a way i can control where the characters will appear in the image?

one way is to y-rotate the envsphere, as the character and dome are nearly concentric.

Yes, the y-rotation as mentioned will bring the other image elements into view.  However, there is a drawback to this method.  At some point, if the panorama is not exact, you will run into an image seam caused be the extreme ends where they meet. 

You can move your models anywhere within the stage, there is no restriction as long as they are not placed outside the dome.  But that is not possible with BB's dome, it covers a great expanse.

I did try moving the characters, but for some reason, no matter where i moved them, it still rendered the exact same spot from the image. I'm wonder what i'm doing wrong.

You did nothing wrong, that is the nature of the skydome, any skydome.  The background perspective remains the same.  Now, by varying the focal length of your camera (i.e., choosing 24mm as opposed to 75mm) you will include more sky in your scene.  But this wider angle make your models appear further away, so to compensate you must move your camera closer.  You sky background will remain the same unless you rotate the camera.  The disadvantage of using wide angle values is lens distortion as your values get smaller similar to a true camera.  There is a trade-off at some point.

If you want some form of perspective efect, add some foreground or 'facades' in the form of smaller spheres textured with a transparancy map. You may then get the effect of say trees halfway the background and the figure move with respect to the 'real' background.  Make sure to switch off shadows for those elements.    

Ahh ok. I guess every lighting setup has their ups and downs. I still find using diffuse IBL's with HDR files or high quality LDR files to be really good, though slightly less quality than that of IDL. I'll just go with the flow and use whatever lighting the situation calls for i guess.

My only petpeeve with IBL is it's tendency to fill shadowed areas unrealistically.  Using it at low intensity values will compensate but it takes some tweaking.  Try placing a structure with openings in an outdoor environment and render the scene with IBL only then IDL only.  You will find that the IBL will light the interior evenly while IDL will only take light sets into consideration.

IBL gives the possibility to 'fake' shadows and highlights, and the effect works only if you use the appropriate light probe. Typically if you take a photo somewere to use as a background, and when taking the picture you put a probe sphere in the intended position of the figure, the probe will give the light information needed to light the figure correctly for that scene. So, if you take a picture of a large solitary tree with the intention to use it as a background for a scene in which the model leans to the trunk of the tree, you place the sphere where the figure will be. The side where the trunk is will be fairly dark, the opposite side light, the top shaded (by the leaves), and the color tone will match the ambient light.
In the actual application the figure will normally cover the probe.  

There are several issues with this. IBL works best is the shadows are there but not form a definite source, so something flurry from tree leaves in a forest scene or color lights from advertisement boards on a city square.  

The advantage, you need to render only the figure, using ambient occlusion as a tool to generate the local shadows.

IDL requires that you provide actual props to give the shadows and refractions and gives best results in populated scenes with, for example indeed, a sky sphere.