Since apparently I haven't followed my initial decision to not dwell much on this thread, I'll give you a few responses to your post above.

Olaf Finkbeiner
McNistor,
you cant discuss one without the other. When expanding the houdini toolset i think it is of upmost importance to not break the “houdini way”.

Nobody wants, to break the “Houdini way”. This detail has been discussed many time before. I don't get why some people see this as having something to do with improving modeling tools.

Olaf Finkbeiner
If you want to do speedy subD manual modelling there is plenty software out there to do just that. I dont see why you cant import the results into houdini and work from there. And there is special software like marvelous designer for clothing, i dont want all that stuff in houdini. Or rhino/grasshopper for nurbs…

I know you have good intentions, but that can be read as disrespectful. I'm not offended, I'm just telling you that's probably best to not argue in this manner. What if someone replied to how you'd like Houdini improved in say liquid simulations by telling you to use another tool, like Realflow or w/e else is out there, because after all we don't want Houdini to be a swiss army knife.

Olaf Finkbeiner
Houdini is all about procedural workflows.

Yeah, no. Houdini's all about creating compelling images. It just happens that proceduralism is the route Houdini took regarding how one goes about creating those images and it just happens that we like it this way.

Olaf Finkbeiner

I dont see why manual modelling tools need to be high up in the feature request list? Why is this so important to you?

Because I like modeling more than other things. I don't think I'm alone in preferring something over something else.

Olaf Finkbeiner
I see however a need for more procedural modelling tools also in connection to sub-d modelling.

Like I said before, a lot of stuff doesn't need procedural modeling - probably an important point of contention, especially when the confusion is not lifted by further clarification which I'm going to attempt below - and that “stuff” is featured on big budget movies, where Houdini has already a foot in the door.

OK, so what do I mean by “doesn't need procedural modeling”. It applies to situations where the final model doesn't need 1000 variations which would greatly benefit from a procedural approach by quickly dialing up and down or randomizing some params. It does not however mean that one shouldn't or couldn't use the same exact tools which comprise what we call “procedural workflow”.
It's not a zero sum game and it's by no means exclusionary in any way.

I think Zeus or Ares is watching us… PolySplitSOP is back! H16.0.618

PolyKnit is the sop in question which unfortunately is not back in 0.618
TBH, I don't think we'd want it to see it back as it was - it was almost unusable so I think it's best that we wait for a re-write. But we do want it back…

do you have a screen recording of the XSI tool working? Nuke's ModelBuilder tries to draw outside the geo but simple projects it way out into space or into the camera.

Not an animation, but there was an image posted on the transformations thread, but that got pushed down.



This image shows how the tool creates polys on “empty space”, in this particular case it was set to draw polygons in a construction plane.
When I get home I'll create a .gif more relevant to NNois' problem.

OK. It seems that Softimage does indeed create a flat polygon when you define the plane by connecting with three points/two edges first. The 4th point of the quad will be created in that plane.
When you connect with only two points results are bad, but in some cases it works a lot better, it's very unpredictable like that. Houdini could do better if the algorithm I described earlier would be implemented for two point/one edge connection. Or any other, it's not hard to come up with smth better than what XSI produces in this case.

I think the perfect tool would be just keeping the new topobuild

but simply with 2 varients

1. Using under the hood the new polyfill to help building the polys INSIDE the actual mesh

2. extending polys to help building polys OUTSIDE the actual mesh

Then, a simple HOTKEY modifier (not an option on the parameters pane please) to switch between the “auto guess” and using the construction plane (or the screenplane)

The topobuild node needs to work outside of the “topology building context” if that's the route to be taken with polyknit. I use this tool a lot (well, not this tool, but the correspondent “polyknit” in XSI) in SubD modeling, which has nothing to do with doing retopology on a model.

What I'm saying I guess, is that the topobuild has to use itself (the geometry I'm working on) as a template - automatically connect both inputs to the node above - via a hotkey toggle that switches between a mode with another file/geometry node as a template and one with both inputs pointing to “itself”.