I've been developing a system that can tell exactly where on a prim someone touched, but I'm running into issues with a drifting of the camera while in mouselook that's throwing off the accuracy by slight margin that is noticable when trying to define small areas of interaction on the prim face.
The drift occurs when passing through one of the cardinal directions relative to the sit rotation. It's a very very slight left or right movement (to the left when rotating left, to the right when rotating right), and from my calculations it is not a set value but is anywhere from .005 to .009. I'm running an older client (not-up dated) and am wondering if it's an issue with the client. I'm going to make the assumption that it's not something that's been bug reported and "Fixed" but it's making for fine-tuning clickable areas very frustrating.
Eventually I'm hoping to have this integrated with a streaming media texture that would display a webpage of some sort and more or less fake html browsing on a prim.