Re: Transient Mark Mode on by default

[Chong Yidong]

Richard Stallman <address@hidden> writes:

>     It seems that Richard is in favour of a scheme which supports both ^ in
>     the interactive spec and a command property (for external packages),
>     but IMHO, this is overkill -- 
>
> It is cleaner to have this in `interactive' than in a separate property.
>
>     If we name the property something like shift-select, then the help
>     system can easily tell people that "applying the Shift modifier to
>     this command will start or extend the active region".  
>
> It can do that based on the interactive spec, too.  That is a good idea.

I've committed to CVS the shift-selection mechanism based on the
`interactive' property.  See the new variable `shift-select-mode'.

As per Stefan's suggestion, the shift-selection is a little more
persistent here than in the old `only' system: only unshifted point
motion commands and mark-deactivating commands will deactivate the mark.
For example, something like C-x 2 won't deactivate the mark.  This
behavior extends to mouse selection too.

This should mesh correctly with the ordinary C-SPC system, as well as
drag-mouse, mouse-3, C-u C-SPC, C-x C-x, and the mark-* commands.  Give
it a try, and let me know if there are any problems.  The shift-select
property and describe-function changes remain unchanged, so far.

I've also taken the liberty of turning off shift-select-mode when
cua-mode is on, so that cua-selection behavior should be unchanged.