Well the results gained with world shaders on fixed objects, is considerably better than with tangent. Depth of detail can really bring the normal mapping to life. In fact I have seen little difference between the tangent normal mapping and standard bumpmapping... of course this may be bcz they are indeed the same thing

If you apply a tangent normal map to an object, illuminate one side and then take a look at the normalmap on the opposite side to the light. It is the opposite effect. I produced a simple object with a tangent based map and rotated it in GS. Close inspection i see the normal map on the reverse remains reversed and incorect as the object rotates.

Maybe you could point me at something that demonstrates this not to be the case... possibly the shader code I was using in my earlier GS experiments was incorrect or flawed itself, I was using one found somewhere on this site I believe...

I think that world based normals have their place, even if they are unsuited to dynamic objects, as the detail is considerably better and more realistic than any results I have had with a tangent based map.