[plt-scheme] Query language for writing SQL commands

I would dispute *compile time errors*, as compile time errors bring
together the worst of both worlds.  I.E. you have most likely forced
some of your uses to now call "eval" (like happened when converting
pattern matching into a macro), and you have delayed the warning
message beyond where it should be.  I.E. compile time errors should be
*edit time* errors, and they should pass through anything they can't
reason about.  I.E. it should be completely transparent to the
programmer, never blocking valid run-time code.

Corey

On 8/16/07, Matthias Felleisen <matthias at ccs.neu.edu> wrote:
>
> On Aug 16, 2007, at 5:40 PM, Corey Sweeney wrote:
>
> > I would dispute #2 [compile time errors are better than run-time
> > errors],
> > but that would lead into a much larger topic.
>
> You can't possibly dispute this fact.
>
> No producer of any product -- software included -- wouldn't love to
> live in a world where errors are discovered immediately and fixed
> before that.
>
> You can only dispute the trade-off between the cost it takes to
> get error messages at compile time (and the cost of figuring out
> how to fix them) vs the cost of just accepting a run-time error.
> Since we are all on a Scheme mailing list, I bet that we more or
> less are in violent agreement on this cost. (Otherwise we'd be
> on the ML or Haskell list or on the Python or ASM one.)
>
> -- Matthias
>
>
>
>


-- 
((lambda (y) (y y)) (lambda (y) (y y)))