[Cad] Article: The failed promise of parametric CAD

[David Kaufman]

Interesting article. It definitely addresses a lot of the problems I've
had with CAD, but I'm not entirely satisfied with the solutions. A
couple points of interest:

1. I like the quote by David Weisberg. Particularly of interest is the
last sentence: "Some people described designing with Pro/ENGINEER to be
more similar to programming than to conventional engineering design."
Sometime I find myself thinking of CAD like programming, but my solution
is typically to move _more_ toward programming (i.e. openSCAD), which
fixes some hidden dependency problems by making them much more explicit.
Does anyone know what he means by "conventional engineering design" and
how today's CAD differs from it?

2. (part 3 of the article) I'm not sure I understand what the author is
trying to describe by "direct modeling"; from what I can tell he's
saying "don't parameterize anything, just build it once", but that still
doesn't solve the problem of errors during editing. Minor point though,
since he doesn't seem to endorse it.

3. (part 4 of the article) I like the idea of "horizontal modeling". I
actually tried something similar once, although I thought of it more as
making an API for each of my components. For each part I first defined
the API (i.e. geometry for top/bottom/side planes) and then created the
part only referencing parts of the API. Then when I referenced a part in
the design on another part I tried to only reference API features. This
worked pretty well, in that the resulting model was editable, but it was
really tedious to do because Solidworks doesn't make it fast and easy to
create, position, and name half a dozen planes/lines per simple part,
and it _does_ make it easy to accidentally reference part geometry that
is coincident with API geometry. Tl;dr: I like the idea, but I don't
think current tooling is set up well for it.

4. (part 5) does he have to call it RMS modeling? We don't really need
that naming collision. I'm happy with just one RMS on my free CAD
mailing list.

5. But more seriously, here's a webinar on RMS (Resilient Modeling
Strategy) [1]. In general I like  the ideas in the presentation, but
they seem more like methods of mitigating flaws in modern CAD software
than proper solutions. Most of the ideas presented in RMS are centered
around only making parent-child relationships when you actually want to
and documenting features as you create them. I find both of these things
difficult to do in current CAD tools, but intuitively they shouldn't
have to be. For example tools like Antimony make the dependencies in CAD
very clear, making it easy for users to set up a reasonable dependency
chain and understand which changes will affect which features. Tools
like openSCAD make it easy to document parts of the CAD with variable
names, function names, or even actual comments. Why should users work
hard for these features when our tools can make it easy?

[1]
http://www.saratechinc.com/ondemand/introduction-to-the-resilient-modeling-strategy/

On 11/13/2015 09:53 AM, Matthew Keeter wrote:
> This is an interesting overview of parametric vs direct modeling:
> http://www.3dcadworld.com/the-failed-promise-of-parametric-cad/
> 
> -Matt
>