>>>>> "Olin" == Olin Shivers <xxxxxx@shivers.ne.mediaone.net> writes:

Olin>     I think that the whole idea of "robustness" (meaning don't break under
Olin>     any circumstances) is not in a Scheme tradition:

Olin> Depends on what you mean by robustness! Scheme has a great tradition of
Olin> the kind of robustness associated with "safe" languages.

Safety doesn't come into it.  The question is whether the system goes
into a defined state in the face of undefined behavior.   Signalling
an error is totally acceptable.  (ML does this even before run time
:-) )

Olin> What I meant by robust was "has a spec that operates on the full set of
Olin> possible input values." But let's not get hung up on a word.

So, guaranteeing that an error is signalled is not robust?

Olin> Now, you can focus on pairness, and still rule out improper lists by
Olin> writing this definition:

Olin>     (let recur ((x x))
Olin>       (cond ((pair? x) (cons (car x) (recur (cdr x))))
Olin>             ((null? x) y)
Olin>             (else (error ...))))

Olin> All this means, however, that we basically coded in a gratuitous
Olin> restriction to our definition that wasn't at all needed.

But you also get something in return: better error reporting.  The
uncommon case of the improper list is IMHE *so* uncommon that it
doesn't warrant sacrificing error reporting for it.

--
Cheers =8-} Mike
Friede, Völkerverständigung und überhaupt blabla