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

[Matthias Felleisen <matthias@ccs.neu.edu>]

  I agree 100%.  After a script gets to 1000 lines I start *really* wishing I
  had type checking.  Before that I could usually care less.  This is equally
  true in scheme, python, perl, etc.

Just wait. It won't be true in PLT Scheme :-) 

 You obviously don't do simulations, where enough is never enough.  That's
 why we still have C++, warts and all.  It's also why cross-language
 integration is such an important problem (so you can do the time-critical
 stuff in a language optimized for that without polluting the entire code
 base).

This "enough is enough" wasn't about speed. It was about the research
topic. So: 

Keep in mind that Fortran compilers for vector and parallel machines
compile through a first-order functional intermediate language with mutable
arrays (called program dependence graph) and then do an array shape
analysis.  They deal with a mono-typed language (see above). They
explicitly ignore what programs do when they go wrong and just produce
something. (An error is turned into a 5, and the program keeps running.)

Do something like that and you get a functional language that runs like a
bat out of hell. 

There is more to say, but I am sick and tired of performance discussions on
functional languages. If you need performance, use Fortran. Bye. 

-- Matthias