> From: Maxim Kizub [mailto:[EMAIL PROTECTED]]
> Sent: Monday, October 26, 1998 10:39 AM
> To: Brian Jones
> Cc: [EMAIL PROTECTED]
> Subject: Re: Free Java compiler - Kiev
>
>
> >"Maxim Kizub" <[EMAIL PROTECTED]> writes:
> >
> >> I think can't catch John's idea... Can you write me a
> three-line example?
> >>
> >> Also, I do not think javac/guavac/etc needs analogs of require/ensure
> >> You'll have checks in debug code, if compile with kiev, and you
> >> do not have them if compile with others... I see no problems...
> >>
> >
> >It sounds like the Kiev extensions will do nothing for us unless we
> >want to build a special "debug" version of the library. I think
> >that this is what Maxim keeps saying and that this isn't a 'check for
> >null' replacement.
>
>
> I'm sorry, but work-by-contract *is* debug feature!
> What else can you expect?
>
> I state, that NullPointerException, ClassCastException and
> ArrayIndexOutOfBounds exceptions are 90% of errors
> that acts on application/applets when they are released,
> and that 2 of 3 downloaded (non-trivial) applets crashes because
> of these exceptions (at least in my experience).
> I state, that having a libary that supports explicit declaration
> of require/ensure/invariant will allow check most of
> sources of these exceptions at *compile* time
> (like other exceptions, that are not childs of RuntimeException),
> and I think, this may be a valuable feature to use
> $classpath library (at least for development).
>
> And I didn't promised anything more ;-)
>
I think he was just explaining that fact to me, not trying to convict you.
It's still very useful as a debug feature, now that I understand entirely
how it works :)
--John Keiser
