Although still I am not familiar with relation between boolean constraint (clp(B)) and finite domain one (clp(FD)), I observed again a similarity of my sat based on DNF to FD as query below. I thinks similarity comes from that a truth-assignment could be viewed as a characteristic function on a subset of the given finite domain in an evident way.
Anyway I have nothing to add to what I understand in naive way about relation between clp(B) and clp(FD). I should delete the quoted comment.
% ?- sat(a+b), sat(b+c), psa.
%@ zid = 8
%@ [b]
%@ [a,c]
%@ [a,b]
%@ -------------------
%@ true.
% ?- sat(a+b), sat(b+c), sat(-(a=b)), psa.
%@ zid = 26
%@ [-a,b]
%@ [a,-b,c]
%@ -------------------
%@ true.