RE: Default behaviour of RET.

[Drew Adams]

> The question of RET's default behavior is not independent of the
> question of that of C-j.  It is manifestly more useful to expose
> different functionality on different keys.  (Even though users can
> customize one of them away, the defaults have their usual importance
> when you're not at your home machine/account/whatever.)

Exactly.  And that's the main point Richard was making, IMO.  The
only other thing he said was that he personally did not think that
the default behavior of RET should be changed.  The more important
point is to keep two behaviors easily accessible.
 
> Is the proposal to make RET and C-j synonymous despite this?

That is unclear to me too, but I too got that impression, and that's
why I responded.  We should keep both behaviors, IMO.

> In the abstract, it would actually make sense to interchange them:
> then C-j could just be a self-inserting character.

FWIW, that works for me.  (I already have C-j self-insert elsewhere, 
including in the minibuffer.)

> It would even make sense for lisp-interaction-mode's current
> binding: you rarely would want to eval the preceding sexp and
> add its value into an enclosing sexp (i.e., where
> indentation would make a difference).

There too, I'm with you.  (In fact, I put *scratch* into Emacs-Lisp
mode because I hate having RET evaluate a sexp.)

> For my own use case, which involves a lot of Python, I like having
> a non-indenting newline available because it saves having to dedent
> manually with DEL (or M-\) when I know I want to return to top-
> level.

I too use RET sometimes, but I can't say offhand when it is that I do.