bug#13731: 24.3.50; C-h N -- Outline navigation Fails

[Drew Adams]

> > I'd say that that is a doc bug.
> >
> > This is a _user option_.  
> 
> Well, I would say this calls for making it a variable instead of an
> option.  This is what we do for org-outline-regexp for example.

That's what I meant by this:

> I suggest we remove that recommendation from the doc string.
> Or we change the status of this variable from a defcustom to
> a defvar.
    ^^^^^^

The question comes down to the kinds of "customization" we really expect from
users.  If we expect that they will want to change the var value globally, then
a user option is appropriate.  If we expect them to change it using Lisp, and
perhaps locally, then a non-option variable is appropriate.

And if we expect both possibilities then we can have both: an option and a
separate variable.  We can decide which takes precedence and document that.

Yes, it can be reasonable (IMO, but others might disagree) to allow the
non-option var to take precedence, i.e., to allow the user option to be
overruled by code, provided this is well documented so users know what to
expect.