[weimedog]

I also would be interested in success storys. At one time I was tasked with writing a specification for a front end to a CAM package to
allow that concept of associativity to function. I used a solid modeling concept where
"attributes" (basically raw data to be later analysed) migrated thru the process either
thru boolean subtractions or thru the front end software...I assumed that there would
be situations where the process evaluation would fail and there would need to be manual
intervention of the re-evaluation process. Used software flags that would be
set to true or false based on the state of the evaluation of the process on those
specific "objects". Each object pointed to the prior one in a process model and to the
next one in the process model. (yea it was old C++ & object oriented programming
concepts). A set of "objects" would be the "process model"..each object in addition
to pointing to the next and prior would point to a geometry set and process definition
on how to machine that geometry. (Could point to the results of a boolean subtraction
of a toolvolume from a stock model defined by a PRIOR process.) The machining
process definition could be a series of "APT" like commands or a more automated
toolpath generation style function...like "solid areaclear" which would bounce a tool
around the inside of a solid model of stock to be removed. The system architechure was
designed to allowed a developer to blend any type of User interface they desired
to either the entire process modeler or each object as the process model was built.
This was to allow flexibilty on how to handle special cases where as a result of up
stream changes, down stream objects were required to re-evaluate or be removed and
replaced with different ones.

Basically it worked pretty well simply because I didn't attempt to automate the
entire process...gave a developer plenty of options on how to handle a stituation
where because of upstream changes, down stream portions of the process model
where no longer possible or even relevent. The system was flexable enough to
allow a completely manual toolpath with each process obect containing an old
command style toolpath pointing to a list of geometry features (surfaces, faces,
lines, arcs, splines, or "profiles") ....implemented in this way it would look completely
as a conventional command based system would look. Or it could handle a more
automated approach where areaclear, solid areaclear, profile, solid profile, or
even more automated surfacing routines could be bundled together and modeled
to define a machining process. The best part was the capability of blending BOTH
styles and still allowing the output of one process to be the input of the next.

Anyway I'm rambling and the concept IS possible as LONG as the developer doesn't
try an ALL automation style to the package.

Hard part was to market the concept. Most people in the industry were not interested
in pushing that edge until someone else did it first..an endless loop.

Also the NIH concept was tough as everyone (including myself) has a pet theory on
how CAD/CAM systems are supposed to look along with an architechure legacy with
current products. I was an "outcast" in my company as I didn't beleive everything
could be automated thru some crazy "knowledge based" scheme!. I believed that a truely
useful system had to blend automation and manual NC programming technics.
Bottom line was many companies did proof of concepts and a few actually implemented this
concept in bits and pieces.(I will never tell which ones were based on where I came from
as I have been out of that business for quite a while) ...and now many years later I would
really be curious who actually figured it out and built a functioning system flexible
enough to be practical and smart enough to automate where it makes SENSE!

Actually the part that really intrigued me was that the places where the
MOST progress was being made in pushing these concepts were overseas.
Pacific Rim in particular. Watching how my customers overseas were
relative to those here...it was only a matter of time (ten years ago) where
the most inovative SW product development will be done somewhere other
than the USA. But I guess thats where the Manufacturing is....

Anyone want to chat about SW theory on how to develop a system with
Associativity?? That might be interesting.