I've been working with the first example some...

It seemed to work well enough in RenderMonkey. And there were alot of issues trying to convert it into 3DGS. To date, it has not been a raging success and makes horrible assumptions about the way the model is skinned, IMHO. If you skin the model wrong, it will not give correct results. It also has bad issues of "swim" at higher bight coefficients. The overall effect (while fast) seems to be best suited to increase the effect of normal/bump mapping on shallow surfaces (like tile and tree bark).

For the time being, I've abandoned that perticular parallax design hoping to find a beter solution.

[edit]
Your second example is interesting, though... I'm reading that now... first glance is that they don't give alot of detail about what the predefined variables are coming from the vertex shader (anyone know what tsL and _tsE represent? I'm thinking transformmed light and view vectors.) It also uses ps3_0, meaning it would not function on many of today's common 3D cards (like my Radeon 9600).



Just thinking out loud, Specterdragon