On Thu, May 10, 2007 at 09:49:30AM -0500, Spencer Janssen wrote:

The only concern I have is that focus changing in tiled mode becomes less obvious. In the original model, focus would always move to the next window in layout order -- which is visually clear in tiled mode. In full mode, however, focus changing is often unintuitive (especially when popup-like windows are involved).

In the new model, focus follows some invisible ordering -- an ordering that is certainly better for full mode, but is it better for tiled?

Of course, the correct answer for you is determined by what feels right to you, but I think the stack would be more intuitive for tiled mode also. It isn't stateless (that is, it has invisible state), but we humans aren't stateless either. In my experience the stack works exactly like I'd like, and exactly like I'd expect: when I close a popup window, the focus returns to where I was... which is precisely what the delete.push test checks. Of course, you could acheive the same with clever placement of new windows in the visual stack, but I doubt that'd actually be better, as you'd like to be able to actually see and respond to popup windows--which means they should be large. Of course, in mosaic mode, the stack order is ignored (except for mod-tab), so the alternate approach of placing new windows at a clever location in the stack would actually work almost identically. But I do look forward to implementing a swap focus module, which just swaps the first two windows in the focus stack. So often I want to go back and forth between two windows (e.g. emacs and terminal), it'd be so nice to be able to just once move the focus between those two using the mouse, and thereafter hit ctrl-tab or something to go back and forth. (of course, automatically avoiding giving focus to the xclock would be even better...) -- David Roundy http://www.darcs.net