> Ah, but that is what I *am* arguing for. The 'reasoning' is simply
> that 'the programmer knows best' -- when the type system doesn't.

"the programmer knows best" is one of the founding principles of BCPL!! I'm
not sure I agree, though - IMO Vincent is correct that being forced to
express things "properly" results in better code in the long run. Try
writing any substantial amount of BCPL[*]...

> The evidence is: programming languages with dynamic or unsound
> typing eg Python, C, are vastly more popular than those with 
> strong typing.

I'm not at all convinced that dynamic/unsound typing is the reason for the
popularity of C, etc - I've never met a C programmer whose eyes didn't pop
out on stalks when you explain that Ocaml cannot reference a null pointer
(not that that has ever caused a Damascene-road conversion to the language,
either!). <joke> C and Python's popularity is more down to needing ever more
programmers to debug the work of the previous programmers, right? :o)
</joke> Personally, whenever I go back to C, the novelty of the relaxed type
system is instantly worn away on the first tricky-to-repeat core dump... at
least an Ocaml exception has a cat in hell's chance of being found
systematically!


David


* With apologies and due deference to Martin Richards if he reads this list!