> Blender currently provides no convenient spline creation, modeling, or reshaping tool
Wow, I remember this recurring extra-modeling-schemes conversation about Blender, and other FOSS mesh modelers from at least 2003...? Lol.
(In 2D land, Inkscape actually retains a few of these various convenient add-on vector drawing methods to this day)
The 3D modeling pitches were also always supported by CAD folks, who were looking to sideload about 10 tons of CAD kernel onto the latest parametric undertaking. Like, give us infinite modeling accuracy and resolution or give us death!
One issue here is this scrappy, Blender-style "whatever works in the final cut" Hollywood mindset [1] vs. the "can I recommend this, as a platform, to all of my friends" mindset in which all contingency needs must be covered, and so Blender is never, ever good enough for you, maybe. Or perhaps the "model itself must be exported directly to high-accuracy industrial plans" mindset.
The former is easier on FOSS teams. The latter always invites this "fork it" mentality and you sometimes get these temporary Frankensteinian spinoffs called "AutoBlenderCAD 3D", etc.
Also reminds me of JPatch? And some other third-party FOSS projects, RIP...
And: Cool if it turns into something great, even if for a while!
> Blender currently provides no convenient spline creation, modeling, or reshaping tool
I suspect it is a bit of hyperbole on the part of the original poster. as Blender has had NURBS surfaces since.. well forever. However nurbs surfaces are very difficult to use in a reasonable manner. As a workable alternative Blender has subdivision surfaces. which as far as the end user is concerned are splines that actually integrate well with mesh geometry.
That specific quote initially led me to dismiss the thread as someone jumping in without understanding blender at all. But I read the whole thread. And my personal conclusion(which does not matter a snot with regards to blender development) is that the original poster wants some poorly defined property that existing spline implementations don't provide(I suspect that if I were more familiar with splines this property would be obvious) and most importantly is willing to provide the dev time required to bring it to blender. So as a occasional blender user I welcome improvements to splines.
In regards to modeling workflow. Realistically, something like blender is good enough for most projects. but my heart lies with constraint based pad and pocket parametric cad. it fits the way I think better.
It's that word "convenient" doing all the lifting, carrying years of depressing threads about how much more epic the NURBS implementation could be, if only, etc.
present but hard is a classic issue, long ago 3dsmax (v2) got nurbs modeling too, but limited and slow.. you'd assume that's the norm, until you hit alias waavefront software where it's fast and intuitive. game changer
Ah, I have very fond memories of modeling with nurbs back in… Maybe 2001? 3ds max v2 was such an incredible playground for a 14-15 year old. V3 was amazing at the time too.
I am, but only have time for reading and occasionally playing around. Every year or two I dig into some software or catch up with new Blender features. If days were 48 hours, I’d be all over it on a regular basis.
I fully intended to have a career in this stuff from age 13–20 or so. When it came down to it, the competition was insane and people were already willing to pay my good money to program, so… I took the path of last resistance :)
Blender users don’t just want to model a shape, though. They need to wrap a texture round it. Do things across the surface space. Polys lend themselves to an easy UV mapping paradigm. Parametric surfaces… less so.
I'll upvote a Maxis tribute comment as a fellow SimCity2000 fan, and also because the point at which the game voiced, "reticulating splines..." was the relevant point at which it made sure your map landform percentiles connected into a surface of either buildable or slopable terrains
OpenSubdiv supports conversion of an input control polyhedron into such patches that describe the limit surface (on CPU and GPU).
But AFAIK it will generate a variety of patch types (not just regular b-spline) to accurately represent extraordinary vertices and non-quad base faces.[1]
In general such representations are hard to manipulate. Stitching cubic patches ensuring C2 or C3 continuity is how we did organic modeling before the advent (or re-discovery, rather) of subdivision surfaces.
E.g. the dinosaurs in "Jurassic Park" were modeled from stitched NURBS patches.
