[Shootout-list] advocacy or curiousity?

Brent Fulgham bfulg@pacbell.net
Mon, 27 Sep 2004 17:56:35 -0700 (PDT)


--- Brian Hurt wrote:

> On Mon, 27 Sep 2004, Isaac Gouy wrote:
> 
> > Once more: advocacy or curiousity?
> 
> I vote for curiosity.  I'm an Ocaml advocate as
> well- but I learned Ocaml because of the original 
> shootout.

My vote is obviously for curiosity as well.  I know of
at least two other people who became interested in
Ocaml based on the original shootout.

> And, while I like Ocaml better than any 
> other language I've looked at, I don't want to shut
> the debate down.   There might be a language I like 
> better than Ocaml out there, and I just haven't 
> heard of it yet.  

yes -- take a look at Clean for example.  Very, very
fast.  And it has great Windows support.

> The problem I have with Brandon's proposal to limit
> the number of languages in the shootout is that I 
> can't think of a criteria which would exclude any 
> language currently on the list, but include Ocaml.  

Correct me if I'm wrong, but I don't think Brandon
proposes we remove languages, but rather provide a
core set of "Least Common Denominator" tests that
don't require certain language-specific features.  We
will shortly be launching into a massive brawl about
what constitutes LCD language elements.  Stay tuned! 
:-)

> So I say- the criteria for being included in the
> benchmark should be it should run on the target 
> system, and someone needs ot be willing to write
> the tests for it.

That's the current state of affairs.  I have two
small additional requirements:
1.  I must be able to get a free license for the
software (to run the tests).
2.  I must have sufficient disk space to install the
software.

> If someone comes along and writes the tests for
> Intercal, befunge, unlambda, or brainfuck, they
> should be included too.

And they will be, should that happen.

> Yeah, they're joke languages.  So?  They can all
> compete for the title of worst programming 
> language.  In fact, I could see this being valuable-
> for humor value if nothing else.
 
One comment I've received a couple of times is that
people suggest removing some languages.  One person
(not Brandon!) commented that I had too many
non-standard languages, and that I should get rid of
some of them to make the tests more meaningful.  He
said something like "... this is a programming
LANGUAGE benchmark, not a language IMPLEMENTATION
site."

Of course that's a silly position.  Languages only
exist as lofty mental models until someone bothers to
write a language implementation.  The only thing we
CAN study are language implementations.

We might be able to reduce visual clutter by only
keeping some of the "best" Scheme implementations, for
example.  But how do we know which are the "best"? 
Basically, by looking at the scores we've accumulated
for them.  So, in attempting to toss away some
implementations we find we can only safely do so by
using the measurements we've already made.

I think this shows the value of measuring as many
implementations as possible.  Otherwise we are just
arbitrarily selecting favorites.

-Brent