I guess we have to agree to disagree on the utility/necessity of 3D constraints then.

3D constraints as I mean them are different from classical 2D sketch constraints. Not only in the obvious aspect of being defined on 3d geometry, not sketch level but those are more meant as an expression of design intent, telling the direct editing software which aspects are important (so what not to "accidentally" change when pulling somewhere) and what to change from how it is to how it is supposed to be (by adding a new constraint and changing it's value). This requires a lot more implicit "understanding" from the system as it is more about finding the solution for an underspecified system (full explicit specification would require a fully constrained body), filling in the gaps with most likely intention of the designer: "Make it like now, "just" that section 10 cm instead of 7 cm". A human would usually immediately know how the result should look like and if 999 out of 1000 would come to the same conclusion then this is most likely a well defined transformation. One based on implicit knowledge. Certainly not easy. But there are direct modelers that can do this pretty convincingly (NX, SE, Ansys etc.) so it seems to be doable.

Sorry for starting to derail "Michael's Plasticity rants" :)