On Thursday, August 22, 2002, at 01:32 AM, Richard A. O'Keefe wrote:

> "David Griswold" <David.Griswold at acm.org> wrote:
> 	You made a blanket assertion, that methods should not ever return
> 	garbage silently, and that it is their responsibility to make sure 
> that
> 	such results are prevented.
>
> Procedures, functions, methods or whatever should be so designed that 
> they
> do not return garbage silently.

It seems to me that a correct diagnosis as to whether all parameters of 
all functions fall within the computable range of every function may be 
a something of a breakthrough in recursive function theory.  We might 
use such a process to help solve the halting problem.

As a practical matter, I would always prefer that erroneous inputs be 
identified and flagged.  However, I'm not sure Richard is proposing a 
solution that satisfies his own requirements.  By using an ad hoc 
solution for a parameter not in the range of a coherent definition of 
the function, all he has done is to widen (silently) the definition from 
a general recursive function to another general recursive function of 
the form: do this, except in this special case, in which case do that.

The latter form of function may be as likely to return "garbage" 
(defined as a not necessarily expected result) as the function it 
purports to replace.

I think Richard argued passionately and credibly, but I'm with Ralph and 
David on this one.