> mouse-1 should not follow any links -- that action is too
On the contrary, _not_ following links and _not_ activating buttons as
a response to mouse-1 events is very confusing.
> mouse-1 double-click should follow a link
> in every mode when point is over a marked link or
> in a mode where a whole line is a link.
I don't have anything againt the idea of double clicks for cases where
the link/button does not look like a link or a button, and the user is
(a) likely not to expect mouse-1 to follow/activate the
(b) likely to want to use mouse-1 to set point on that particular
piece of text, for example because most of the buffer is made
up of links/buttons.
> mouse-1 should set point.
That should always be the normal case. The case when the pointer is
over a link or a button should be exceptional.
> As far as I can see, this will be OK both for novices coming and
> for experts.
In what way is it ok to highlight a piece of text as a familiar web
browser link, and then _not_ bind the familiar mouse-1 to follow it?
The same holds (probably even more so) for buttons: the concept of
clicking mouse-1 to activate a button is so basic to GUI usage that
deviating from the standard is like having an elevator with buttons
that you need to lick to activate.