Tony,
I'm with you for the most part, not to say that I'm against what John or anybody else has said here. All good advice!
Primitives are parametric in their overall dimensions, and the can be moved using the move feature function. It has to be said that when positioning holes and pockets relative to faces then maintaining those relationships using by moving separate primitives about is vastly inferior.
What I will say is the UG has lots of really powerful tools that each have their uses, and I do believe in using them. I advise people to do things a certain way for a whole bunch of reasons, but I have to make room for people to be as productive as they can in their own way. At the end of the day getting the best "Bang for your buck" out of the CAD system relates in part to how the CAD system enables people to work productively. I think UG is one of the better CAD systems for doing that because it is flexible.
Examples of what we do ask users to do are things like separating booleans from feature creation, (i.e. extrude can be united or subtracted at the time of creation, but we don't prefer that). This is simply to make the structure of the model a little more visible on the feature list and easier to maintain.
I'll go one step further that talking just about thinking like a manufacturing process. We often work up studies ad preliminary models where we design very conceptually, which is to say we don't know how the model is going to come out until we're done hacking away at it. Sometimes it has to be built like a tool, other times you're packaging for space by chipping away at the target solid to come up with a usable result. Under these circumstances we often don't know whether the concept will develop into anything worthwhile so we cannot afford to invest a lot of time into it. Often I'll use your technique of slicing away at primitives which I unite subtract intersect etc... I use far too many replace faces and offsets for comfort, but the job gets done possibilities are proved or disproved and we progress from there.
The temptation is to progress the best study into the production model. Our system is to distinguish between study models, that you probably wouldn't wish to maintain, from production designs where some thought is put into making the model maintainable for the next guy. Truth be told working from the basis of these fairly anarchic study files means that we have learned to avoid what may be described as overly complex associativity. This is as opposed to not wanting to work parametrically. What we want others to think about is how they can put together models that are easy to work on, (i.e. they can be changed easily).
The opposite to this is where somebody builds every feature in a model associated about a base feature which can for ever after never be changed without destroying the model. One such example was a small pocket with a hole in the middle. This was a notional interface point of in a large and complex panel. When the method of interface failed to work the pocket had to be removed which orphaned the simple hole and almost every other feature in the model. We were easily able to add a feature that removed the pocket just before the final hollow, thereby saving a lot of trouble, but I would have been better satisfied had I been able to remove the pocket discreetly earlier in the model without too many problems. The CAD system wasn't to blame, the person who made the model made poor choices in hindsight for what may have initially seemed good reasons. He failed to keep it simple stupid, (and never heard of Murphy's Law).
Hope this helps find the path of CAD enlightenment!
Regards
Hudson
