Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> What do others think about making this the default?
>
> My understanding is that usage style of registers varies a lot, where
> some people use them extensively, so any change that makes them
> heavier/slower (such a requiring a confirmation) will alienate them.

My impression is that these people use a workflow based on few register
(I guess maximum 5) because it is hard to read the register preview as
it was before, so they work with for example only "a", "b" and "c", they
always overwrite them and as there is only few registers they can
always remember what they do.

But when you have many registers, the time you spend scrutinizing the
preview buffer is worst than the time you spend hitting RET or even
navigating the preview buffer with C-n/p. 

> 
> For that reason I'd opt to keep the defaults as close to the previous
> behavior, focusing on making the new features discoverable rather than
> enabled by default.

So we could have the default exactly as before (register-use-preview ==
basic) and disable the behavior of register-preview == nil, keeping only
t and 'never.

t would provide the fully featured preview buffer (with RET everywhere to confirm).
And 'never would be the same but with no preview buffer (but still available
with C-h).

As I have actually a problem unsolved with register-use-preview == nil
with kmacros, this would save me some more work.

-- 
Thierry