Stefan Monnier wrote:
> Sorry, don't mind me, I was completely confused.
> 
> 
No worries.  Anyway, there's probably a much more elegant way:

(Background: I «gasp» read the docstring for x-set-selection, and
_thought_ I'd found a better way - it can take a cons of markers
to _lazily_ find the selection data as whatever's between
the markers when something requests the selection. However, it turns out
that the emacs point is _not_ in fact a marker, so you can't use
mark-marker and point-marker to find the region on-demand (point-marker
just returns a marker to the instantaneous position of the point))

*** Sooo - Here's a solution that seems generally saner, though does
wander deeper into the emacs core - allow x-set-selection to take a
function that will be funcalled on demand to return a string to use as
the selection data, not just a cons of markers.

Avoids performance issues that the moronic string-equal or hash in the
timer would introduce, and the (theoretical, for inhumanly fast users)
potential flakiness of an idle timer.