Failure-handling as fundamental

[Mario]

> 
>   A "true" program is one which terminates without failure.
>   A "false" program is one which fails.
> 

This is exactly how Prolog works. And it works very nice, especially 
when combined with unification of variables (two-way pattern-matching) 
and backtracking. For example, Prolog expression

X = pair("a", Y)

where uppercase X and Y are variables and lowercase pair is a 
constructor, will:

- succeed if X is instantiated as a pair, it's first component is the 
string "a" and the second can be unified with Y.

- fail if X is not instantiated as a pair that matches the pattern

- if X is not instantiated, the expression will succeed and X will be 
unified with the pattern, i.e. it will match the pair construct for all 
the expressions that follow. If any of the following expressions fails, 
the unification of X will be "undone" before any or-alternative is tried.

	Anyway, this is what more-or-less happens on the low level. Reading 
Prolog is easier if you regard the "expressions" I mention as logical 
statements and the "sequences" of succeeding expressions as logical 
conjunctions of these statements. After you wrap your mind around it, it 
actually becomes quite easy. But I think it's easier to do if you first 
understand what really happens on the low-level, and for some reason no 
Prolog course seems to be arranged that way. After the demise of AI, the 
language went the way of Forth and APL.