🚀 go-pugleaf

I agree with earlier statements that DFPD is rather useless as an approach
for or-parallel logic programming systems. It has one interesting point though:

It is hard to think of an or-parallel scheme which has a simpler implementation
than this one. The needed communication is minimal and the variable 
representation is optimal ( = Warren's for instance). The price you pay is
reexecution. In all other schemes I have seen (in our lab there is work going 
on considering a handful (sic!)) there are considerable measures taken as to
either reach a highly flexible variable representation (Haridi/Ciepielweski
/Hausman proposes hashing etc.) or trying to avoid copying through more or less
smart heuristics or even specialized hardware (Khayri/Fahlen/Karlsson).
All of these schemes involve a considerable amount of communication, at least
compared to DFPD.

An implementation of a DFPD Horn Clause prover ("pure Prolog") could show
useful in the sense that any proposed "smart" scheme has to be at least as
good as the DFPD scheme (for at least some subclass of programs and queries)
to be taken seriously. It is also interesting to notice that DFPD is
complete, whereas most of the "smarter" schemes are not.