When he started to write about writing (military) simulations in a functional style, I really expected a reference to this paper http://haskell.cs.yale.edu/?post_type=publication&p=366 which is one of the rare examples of a real empirical experiment in software enginering. (And it is going to turn twenty this year, we are not learning very fast as a community).
This is not the most fair paper (it could be described as: two Haskell language designers kick program-manager-struggling with-pre-STL-C++'s ass) but it's an interesting reading.
It would be great to see this paper's experiment conducted in the modern world. Haskell vs Python vs C++11 vs Clojure vs whatever.
Actually it would be great to have a site like http://benchmarksgame.alioth.debian.org/ that focuses on program readability, development speed etc.
Not an holy grail, bur take a look at Rosetta Code : http://rosettacode.org/wiki/Rosetta_Code
Interesting paper.
When I look at the outputs, I think "this looks like a very simple homework assignment". If they had provided sample input, as well as sample output, I'd be able to test my belief. I imagine I can produce functionally-equivalent Python code in a few hours (note: Python was not a reasonable option in 1993).
In any case, this is a test in prototyping, with guidance of "5 person-days of effort"), so you would certainly expect languages like C++ and Ada to fail spectacularly.
It was interesting to read about how bloody fast the lisp implementation was written. That seems insanely fast. Presumably the author was one heck of a specialist in the field. And... it is notable that with such a lower amount of documentation, it still did not score much worse than the Haskel solutions did. (I really need to learn lisp well.... Also, be in a place I could use it.)