An article has just been posted by Martyn Day , who appears to have quite an in-depth knowledge of the whole picture.
His point about subscription cycles allowing CAD companies to walk away from bugs and unfinished features has been traversed on this NG, although not always with the clarity he brings to the question.
But he has a particularly clear grasp, it seems to me, of the business imperatives which drive CAD company policy, and how they set up an essentially disfunctional dynamic.
One point he makes, at the end, I think misses an important qualification: he talks about the importance of users having a mixed CAD portfolio -- not putting all their eggs in one vendor's basket -- but he doesn't point out the cost penalties of doing so. I'm thinking of the problems which are compounded when you have to climb several learning curves at the same time.
It strikes me that the biggest problem of the learning curve with a subscription product is that it does not stay climbed. It's more like running up a 'down' escalator than scaling a cliff. The only way we can minimise the time the escalator is running, is by upgrading late, and seldom. Unfortunately, the need for interoperability with other businesses, and the continued absence of reverse compatibility, means many users don't have a lot of leeway in timing their upgrades.
The majority of the downtime and lost productivity is not associated with learning how features work, but learning how they *don't* work: ie limitations, bugs, misleading documentation.
As most actual users are only too well aware, the finite amount of horsepower ('smarts') they command has to be divided between understanding the tool and focussing on the task. Given the need within one office to operate on other people's models and drawings, it doesn't seem feasible to have different users unable to use each other's CAD package competently. If this is true, doubling up on tools doubles the amount of horsepower diverted away from the actual task at hand.
The article is at