Exercise 4.78.  Redesign the query language as a nondeterministic 
program to be implemented using the evaluator of section 4.3, rather 
than as a stream process. In this approach, each query will produce a 
single answer (rather than the stream of all answers) and the user can 
type try-again to see more answers. You should find that much of the 
mechanism we built in this section is subsumed by nondeterministic 
search and backtracking. You will probably also find, however, that your 
new query language has subtle differences in behavior from the one 
implemented here. Can you find examples that illustrate this difference?

————————————————————————————————————————————————————————————————————————

We will use `amb` to select frames from those available.

In the case of the query:

(job ?person ?position)

The stream-based implementation will first query the database, and 
return a stream of frames which match this pattern, which the REPL will 
then display to the user.

In the nondeterministic implementation, we use `amb` to select a single 
assertion from the database, and `require` it to pattern match against 
the query in the context of the existing frame.  Each time we extend the 
frame with a new binding, we do so ambiguously, so that other bindings 
can also be tried.  Rather than passing streams of frames, there is only 
one frame, but at each point where it is extended we create a branch 
point.  Rather than returning a `failed` symbol when frames cannot be 
extended, we would use `require` to kill entire branches of the 
computation.

At a glance, it seems this approach would take the same choices in the 
same order that a frame-stream implementation would take, and no obvious 
subtle differences are coming to mind...