After reading through this post, both sides of the coin have some valid points on what should, could, and in the future might possibly be.
That said, it's always easiest to say what one would do "if"...... (insert phrase)
Kind of hard to see the big or overall picture when one does not have all of the pieces, or at least have some notion of what is and is not possible. If anything, time has shown, the human brain (intelligence) can quite possibly figure out any solution to a problem. That said, and not trying to get too philosophical, it's only once you understand the rules can you truly break them.
One of the main bullet points that have been brought up is that if a new feature is used in the program, what will it be in the past version. The solution was thrown out; just turn it into a "dumb" feature. But think about that for a second..... about how long your Feature Manager Tree (FMT) can get, how many Parent child relationships have been established, how this is just not a straight line anymore. Parametric programs are now akin to spider webs.
And I know that we all know and understand that these "rules" are in place. But take a step back for a second and really start to examine what is being asked.
Take my FMT and examine the 2,3,400 features that I have, Track and understand what was not in the previous version of the program Change that specific feature into a "dumb" solid AND, make sure that ALL relationships don't blow up, so that when I make a change everything holds true.
Not saying impossible, just a mighty big undertaking. Granted, there is tons of new user's everyday, but by now most of you seasoned veterans know about the sting, so there isn't any real reason to get caught by the upgrade bug.
I dunno, just 2 cents......