[RFC] Alternate approach to keeping convenience variables

[Jim Blandy]

My point was simply that life is easier when you have invariants that
hold over wider regions.  I see that someone who has the whole picture
in mind can get the effect they want, but it's simpler to have only
one case to handle in the first place (is $trace_frame -1?) than two
(is $trace_frame even set? what is it set to?).

But I think you all understand what I've been trying to say.  All the
persuading that's going to happen has already happened.  :)