Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!allegra!princeton!caip!seismo!rochester!bullwinkle!uw-beaver!ubc-vision!ubc-cs!ludemann
From: ludemann@ubc-cs.UUCP (Peter Ludemann)
Newsgroups: net.lang.prolog
Subject: Re: Standard behavior?
Message-ID: <253@ubc-cs.UUCP>
Date: Sat, 24-May-86 01:41:35 EDT
Article-I.D.: ubc-cs.253
Posted: Sat May 24 01:41:35 1986
Date-Received: Mon, 26-May-86 01:43:55 EDT
References: <980@watdragon.UUCP>
Reply-To: ludemann@ubc-cs.UUCP (Peter Ludemann)
Organization: UBC Department of Computer Science, Vancouver, B.C., Canada
Lines: 42

In article <980@watdragon.UUCP> rggoebel@watdragon.UUCP (Randy Goebel LPAIG) writes:
>> >Consider the following trivial predicate:
>> >a([]).
>> >a(_).
>> >
>> >Given the query :-a([]). , C-Prolog finds one match and UNSW Prolog finds two.
>> >Which is standard behavior?  How do other implementations behave?
>-------------
>Hmmm.  The reason for logic programming's existence is to dispense with
>guesses about what behaviour should be.  The formulae assert that the
>individual constant named `[]' and everthing else (i.e., `_') is in the 
>class named by the predicate `a'.  If you believe that the anonymous variable
>is a universially quantified variable, then there are two resolution proofs
>of the query a([]).
>
>Implementors' treatment of `_' can produce non-standard behaviour; non-standard
>means not consistent with the logical interpretation.

I have to disagree, Randy.  The query merely asks whether or not 
there is a proof of "a([])".  Not how many there are.  Although by 
Prolog's execution order, there are only two ways of generating the 
answer, there are of course an _infinite_ number of logical proofs.  
There is no point in trying to list them all :-).  

Furthermore, if the goal "a([])" is within a predicate (for example, 
q(X) :- a([]), b(X).), then a smart implementation would notice that 
if "b(X)" fails, there is no need to re-try "a([])".  Isn't that what 
all the work on intelligent backtracking has been about?  

Now, a general philosophical point.  Prolog is certainly _not_ logic 
programming, although it is a large step in that direction.  A true 
logic programming language would not have nor need "cut" ("!"), 
"var", "nonvar", etc.  ("=..", "name", "is" and even "call" are 
legitimate because they can be considered as an infinite number of 
rules.  "var" and "nonvar" can't be described that way).  One of the 
advantages of logic programming is that it allows us to consider 
computations without knowing (precisely) the underlying execution 
strategy.  Indeed, for "pure" programs, the execution strategy could 
change, yet the programs would still execute correctly.  Let us 
strive to produce true logic programming languages which get over the 
minor failings of Prolog.