Modo, Lightwave, and Hexagon generates polygonal patches/surfaces out of curves (Hexagon supports Limited Bezier, Inerpolated, and Spline curves). I suspect all the other applications do to some degree (including Maya but I 'don’t claim to be a Maya expert). 3DSMax does support polymeshes with splines (please see attached image).
I don’t need super-accurate curves- I have Rhino3D for that- just something that’s useful. As I’ve stated, an implementation of a SubD line with the refinements I mentioned is fine. Lacking that, I’d settle for a well-crafted SDK that exposes as much of the internal mechanics as possible so I can roll my own Bezier plug-in, essentially adding a new type of geometry (I used to write device drivers for IBM AIX so getting my hands dirty isn’t new to me).
I’m primarily a NURBS modeler looking for a good polygonal tool. As such, I’m more accustomed to working with curves rather than edge and face pushing/pulling. My (and a lot of others out there) workflow is more akin to an illustrator’s rather than a sculptor’s (which is what Silo is advocating). Would you change your modeling paradigm for a program? Of course not. You first try to see if the developers can (or have the inclination) to implement features you need. If not, you try to see if other have done it. If not and you’re skilled enough (and have time to spare), you try to see if it’s even possible. If not, you move on to another package. Right now, I’m at the third stage waiting for 2.0.
The thing I like about working with curves is that it’s less algorithmic than SubD. Thus, if you’re planning to do a hi-res model, you can plan on the detail at the earlier stages of construction rather than subdividing and cleaning up later. I don’t use this for the entire model- just those parts that are noticable and need to be at a specific detail level because of, say a reuired shot I’m doing. But then, this is just my opinion.
I should state that I do primarily mechanical modeling (not much into characters) and I have tried to use Silo for that purpose. Their current toolset might lend itself well to character modelers but for hard-surface modeling, they are lacking some necessary tools (deformers and curves are just some of them). I do like Silo’s focused approach- I just think that as a hard-surface modeler (and I’m probably not the only one), the NC team needs to expand Silo’s toolset to gain more seats in that area.