Re: Strong Typing, Dynamic Languages, What to do?

[Jerzy Karczmarczuk <karczma@info.unicaen.fr>]

I underlined the efficiency issues related to the typing problems which
were pushed to the second plan by Noel Welsh advocating Scheme.

Noel mildly agrees with reticencies, and now Master Matthias says:

> with all due respect, I think Noel was on to the right answer.
> 
> And why does SML's performance suck, if static types are really all that
> great for performance? Once you have polymorphism you don't need dynamic
> types to mess up your performance.
> 
> I won't ask the same question about Haskell. :-)

...

>    For people who *seriously* pose themselves the question ...
> 
> they know that Chez Scheme has all of that and runs at a pretty good speed
> for the average program. If you turn of the safety checks, you get better
> performance than SML/NJ. 

But he ends another posting in such a way:


> P.S. No, I don't pay much attention to performance. I have lived with High
> Performance Fortran people for 15 years. Enough is enough. 

====

First of all, I am the last to attack Scheme, I use it all the time, and I am
not *so* attached to the speed neither. Noel's arguments were perfectly right
and sound, just a bit incomplete! [I know about Stalin which I don't use because
I am alergic to the name, and I know about Chez, which is used here only
in its Petite version, so I can't say anything about the speed of a fully
compiled program.]

There are people who teach computer science to future computer scientists and
they have their priorities. I do it as well, but I taught all that stuff to
physics students, and now I have frome time to time in front of me a bunch
of future engineers. Questions about performance are asked all the time, they
are more frequent among younger people.

Please visit from time to time such newsgroups as comp.graphic.algorithms,
where very, very young people ask the same questions again and again. There,
and in the mentioned part of my pedagogical activities, the first time I
would dare to say: "I don't care about performance" would result in a critical
attenuation of the interest of my audience.

So, we underline the positive side of dynamic typing showing how easy is to
implement (well, conceptually) objects, event-driven stuff, etc. I do some
teaching using Matlab, and I show that in Scheme it is possible to do the
same much easier. But the performance IS important, otherwise students
suspect *really* (they told me so) that we force them to use the language we
like according to our sick pedagogical visions they don't give a damn for,
without taking into account the "market demands".

Matthias is sick of discussion of performance of functional languages, and
suggests to move to Fortran. So I state once more my point: it might be a bit
harmful to *declare* that Functional Gurus don't care about it. They do, they
do a lot!

ML sucks, and Haskell, hmmmm..., despite their strong typing?
I hope nobody tries to deduce that this proves that strongly typed
languages are inefficient because of that, hm?

Hundreds of papers have been written about the efficiency issues, and this 
is good. The principal point is not to be sectarian. 
This was my swan's song within this thread. Back to work.


Jerzy Karczmarczuk
Caen, France