Hi niko, also you wrote:
> Would the optimal way to create that surface be a
> network if I split the converging curves into 3 (for a
> total of 4 curves for the network)?
I think you're talking here about having a layout where the "rails" have shared tangents with the "profiles" ?
That tends to have another kind of semirelated problem  it's more like uneven "stretching" as things approach the shared tangency corner.
Also having a single NURBS surface that has tangent corners instead of distinct corners tends to also degrade the mathematics of the surface evaluation in that area as well  the definition of the surface normal is usually being the shared perpendicular (called the "cross product") between the u and v directions at any point on the surface. When the u and v directions are tangent with one another it is a degenerate situation where there are a whole lot of possible directions perpendicular to both instead of just one direction perpendicular to both.
So what you are talking about there can tend to have its own kinds of problems as well.
 Michael
