[just6979]

Pointers are still very useful for many paradigms. Think about something like Blender or a game level editor: there can be _a lot_ of controls visible at once, trying to navigate them all with the keyboard is just unfeasible. And doing a fully context sensitive setup to limit visible controls, like the MS Office Ribbon, is also infeasible because the changes would be happening almost continually as different objects are selected and modes are chosen.

Your bad UI example of resizing windows is way less about the round corners or lack of obvious grab area (handle). It's more that the handle is way too small. It's a couple pixels (maybe just one?) wide/tall on screens that are thousands of pixels wide! It's just too easy to overshoot. I'd say it comes from the obsession with minimalism and flat design such that there is almost no visible seperate border to act as a target. Combined with trying to remove ambiguity as to which window the click should go to (if you click two pixels "outside" a window, should the click go to the window beneath or be interpreted as trying to grab the border?), the grab handles are tiny, almost matching the actual (lack of) pixels of the border, instead of being a usable target to click on.

To me it points to a lack of usability testing, or at least lack of generalized usability testing, ie: they tested their own workflows, which seem to be just always leaving windows as the OS creates them initially, or maximizing everything, not much resizing at all. Similarly, generally testing a [mostly] keyboard interface is tough to do thoroughly without providing a thorough cheat sheet. You know the commands because you made them, easy to test how you work, but others need to learn them first.