Hi guys, could help but jump in here.
I'd actually like to see surface modeling eliminated. Yes, you heard me correctly... What I mean is that if we had enough options with SW, I'd like to stay with solid modeling, yet be able to do local modification to face/s of the model without having to go back and forth between hybrid surface and solid modeling. So much of what we SWX surface modeler are doing is "getting around the limitations of the history tree". You need to keep in mind that surface modelers ultimately want a solid model.
I think that this is the direction to go because solid models are more intuitive than surfaces (to the visual mind). Surfacing can be so meticulous and it doesn't need to be. Unfortuneately, SWX has borrowed a lot of the surfacing methodoloy from the way surfacing has been done over the last 30 years (ship building = loft).
Instead, think of a future SWX modeler that started with a solid box and apply radius to all 8 edges of it. At this point you break from traditional modeling and go in and start to do local modification to the connection of the faces of the solid model. Say, you had a modification feature that you could pick the radius and change them or both the radius and the adjacent model faces (flat sides) to a G2 connecton with a single button. Then, while still in the feature, you had drag handles with a number box, to tweak the continuity at each of the boundaries. would this be much more intuitive than the multitude of curve network features that drive the surface?
Its time for a new paradigm, and althou its been tried and had many false starts, "object oriented" modeling could be implemented in the area of surface modeling. This would, for example, be the ability to drag and drop complex, pre-parametrized features on to complex face/s and then the feature is modified thru values. Think of a end cap feature that can be dropped onto the end face of a shaft, and that feature would match the adjacent faces of the "complex - non-analytical" shaft with G2 continuity. In this way we get away from certain features that we "re-create" over and over using sketch curves. So many of the features presently are so dependent on 2D planes and 2D sketches. These are great for a lot of prismatic - analytical modelling but doesn't cut it for non-analytical (surface) modelling. Why do we still try to use so many 2D methologies for surface modeling? The second area that I think SWX need to really consider is consolidation of features or what I call "Super-features". This is a trend that is taking hold with SolidEdge and Wildfire. Think of the flexibility and freedom to redefine the design intent if you had the ability to redefine a loft into a sweep or a sweep into a fill. So a general surface feature that could be fill, loft, sweep, dome, etc. This should also apply for the analytical modeling features. The Extrude command should encapsulate: Extrude cut, Extrude Boss, Extrude cut surface, Extrude cut Boss, Thin feature. Revolve should do the same.
SWX should also be smarter about where to give up tolerance for the sake of completing the feature successfully. This would be especially true for shell, but also for many of the surface modeling command that rely on "manifold" rules. We all know that there are many cases that don't solve just because the "math" tolerance, which is extremely small in deviation, isn't met, but the visual result is completely exceptable.
SWX developers, mathematicians, and coders, need to break away from their notion that surfaces are mathmatical; as far as the end user thinks, they are not, they are visual, I repeat that again: THEY ARE VISUAL, and this is the most important aspect of them, not that they are mathmatically correctly. Because they are visual, more drag handle, spin boxes and other dynamic type controls need to be introduced into the interface. I know that this is hard to hear for a ME, but is is absolutely true and it is why we struggle so hard to do surfacing. It is so hard because we are trying to apply analytical, parametric concepts to a non-analytical, non-parametric feature (surface) in principle. Now, there are cases where parametrics work quite nicely for surfaces, and I happen to believe that the parametric surface model is the way of the future, but only if we allow ourselves not to force surface modeling features into the prismatic modeling box, metaphorically speaking.