Paul, I realize that defining a model for movements/selections inside mathematical environment is complex and there could be different more or less equivalently good solutions. This is job for the developers, if they consider this worth of. Maybe between the many different solutions they could simply choose the one which is easier to implement.
In any case, to me it is highly desirable a consistent behavior. Above all, in my opinion, this means the fact that the shifted version of a combination should select the same elements over which you move with the non shifted version. At present this is not the case and I find myself doing and undoing selections. Yes, I know: mainly my fault... yet this doesn't happen to me while I am in a pure text paragraph. >So, what I would find logical and consistent is the following. > > Movement: > left/right arrow move one character (where entering or exiting an inset, such as going into/out of a subscript, counts as a "character" for movement purposes); > ctrl-left/right jumps the adjacent "grouping", where "grouping" is the subtree at the current node if we were to diagram the math inset as a tree; > home/end jumps to the start or end of the entire inset; > ctrl-home/end jumps to the start or end of the buffer (document). > > Selection: shift+any navigation combination selects everything from the current cursor position to where the unshifted key combination would take you. This seems to me a good model. It leaves space for a couple of bindings: ctrl-page up/page down and ctrl-up/down which could be, for example, ctrl-page up/page down jumps to the start/end of the formula ctrl-up/down something else to be decided (or viceversa) and similarly for the shifted version.
